Quantcast
[ 3 / biz / cgl / ck / diy / fa / g / ic / jp / lit / sci / tg / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/g/ - Technology


View post   

[ Toggle deleted replies ]
File: 52 KB, 894x250, 1546422963983.png [View same] [iqdb] [saucenao] [google] [report]
69167609 No.69167609 [Reply] [Original] [archived.moe] [rbt]

WTF IS THAT

>> No.69167624

fake news

>> No.69167642

>low end part is already 6c 12t
true big if

>> No.69167659

>>69167609
>cores and clock speed equal performance!

is that why Coffee Lake was trashing Ryzen 1 in every real world benchmark?

>> No.69167748
File: 150 KB, 775x759, 1520971059899.jpg [View same] [iqdb] [saucenao] [google] [report]
69167748

>>69167624

pay debt

>> No.69168019
File: 123 KB, 789x800, rage pepe.jpg [View same] [iqdb] [saucenao] [google] [report]
69168019

>>69167609
Fake and gay.
>and no Mommy Su closeup to make it all better

>> No.69169090

>>69167609
I'm excited for CES desu
Either AMD delivers BTFOs Intel or AMD doesn't deliver and we'll have a very entertaining shitstorm

>> No.69169102

>>69167609
How long until new series?

>> No.69169105
File: 91 KB, 400x400, 1534893703765.png [View same] [iqdb] [saucenao] [google] [report]
69169105

>>69167659
>>69168019
>>69167624

>> No.69169119

>>69167609
The next holocaust

>> No.69169184

>>69167659
>geekbench
>real world benchmark

>> No.69169251

>>69169184
that and every game released in the past decade. I hope you have a funny meme picture to post in response, that'll show me.

>> No.69169279

>>69169090
Time to prepare "RIP intel" pics or "Wait for AMD/Navi" pics.

Or both.

>> No.69169281
File: 172 KB, 1183x754, goybench.png [View same] [iqdb] [saucenao] [google] [report]
69169281

>>69169251
here it is

>> No.69169616

>>69167609
Are the 12c and up all multiple die? Whats will most likely be the largest core count single die?

>> No.69170110
File: 188 KB, 400x600, 1545672612521.jpg [View same] [iqdb] [saucenao] [google] [report]
69170110

>>69167609
6 CORE 12 THREAD FOR THE ENTRY LEVEL HOLY SHIT ADOREDTV WAS RIGHT

>> No.69170145

>>69167609
So basically there's zero reason to buy Intel now unless you have a cuckolding fetish

>> No.69170222

Now let's see those game benchmarks. heheh

>> No.69170302
File: 71 KB, 552x661, 1501441065814 (1).jpg [View same] [iqdb] [saucenao] [google] [report]
69170302

>>69170222
only the node size and the amount of cores matter!
Real world performance doesn't!

>> No.69170321
File: 87 KB, 757x627, 1529567124804.jpg [View same] [iqdb] [saucenao] [google] [report]
69170321

>>69167609
now i need to see the R3 3300X to BTFO intels 8700k on their stock cooler. just try to imagine the shit flinging after that

>> No.69170367
File: 96 KB, 750x1250, InKEKNPC.gif [View same] [iqdb] [saucenao] [google] [report]
69170367

>>69170222
>>69170302

>> No.69170430

>>69170321
Yeah that will happen.

>> No.69170438

>>69170367
I HATE AMD AND I HATE ANTI-SEMITIC GOYIM

>> No.69170442

16 cores is too many cores.

>> No.69170447
File: 386 KB, 1000x1000, lisa_su_pepe.png [View same] [iqdb] [saucenao] [google] [report]
69170447

>>69170430
reach for the stars my dude. can't be a pessimistic asshole your whole life long

>> No.69170477

these rumors are by russian trolls.

report and move on, nothing to see.

>> No.69170496
File: 164 KB, 1200x800, 1536694146733.png [View same] [iqdb] [saucenao] [google] [report]
69170496

>>69167609
>3600X
>8c/16t
>4.8GHz and above
IT'S OVER INTLEL IS FINISHED

>> No.69170521

TICK

>> No.69170531

STOP MAKING THESE AMD POSTS!! DELETE THIS NOW CUNT!

BUY INTEL BUY INTEL BUY INTEL BUY INTEL BUY INTEL BUY INTEL BUY INTEL BUY INTEL BUY INTEL BUY INTEL BUY INTEL BUY INTEL

>> No.69170614

>>69167609
DELID THIS!

>> No.69170703

>>69167609
wtf would I need 12 cores for?
What a huge letdown, that's completely useless

>> No.69170708

>>69167748
OYYYYYY VEEEEEEEEEEEEEYYYYYYYYYYYYYYYYYYYY

>> No.69170742

>>69170531
Well memed.

>> No.69170916
File: 12 KB, 480x247, 1541246183696.jpg [View same] [iqdb] [saucenao] [google] [report]
69170916

>>69170496
>not buying the 3850X when it drops for best dies on desktop

>> No.69170966

>>69167609
>8c/16t
>65w TDP
my body is ready

>> No.69170973

How will Intel use their (((jew))) powers to make this fail????

>> No.69171008

>>69170973
More gay comedy actors on planes.

>> No.69171010

>>69170973
"brand loyalty" goyim

>> No.69171013

>>69170966
Mine to anon, mine to. 2019 is going to be a great year (if it's not bullshit that is.)

>> No.69171031
File: 29 KB, 640x640, 1501764121976.jpg [View same] [iqdb] [saucenao] [google] [report]
69171031

>>69170110
AMD isn't fucking around, they really want to bury intel alive. Interesting to see what jim killer does to make intel relevant again.

>> No.69171071

>>69171031
>tfw cpu industry is now just jim keller trying to one up himself

>> No.69171088

AMD is the new Cyrix

>> No.69171120

>>69167609
It's over guys, Intel is finished.

>> No.69171187

>>69167609
As much as I want to see AMD catch up or even overtake Intel this gen. Some things still don't add up. Can am4 support 16 cores with only two memory channels? Are mainstream desktop users going to buy a 12 core cpu, let alone 16? Is everyone suddenly going to render 3d scenes, encode video on cpu and setup VMs?

>> No.69171313

>>69171187
No its not to do with EVERYONE. It's simply making high core count chips the standard. Intel has kept people at 4 cores or less for years so that's what devs designed stuff for. HEDT systems had the "high core count" setups (6c12t) but even then, programs didn't take great advantage of everything.

But now imagine 6c12t being the bare bottom standard. Multi-threaded programing would start to become the norm as well. Games did it. (DOOM, GTA5, COD) Notice how quick Intel released their 6 core i5 and 12 thread i7's in the wake of Ryzen? Even Intel sees the end of the road with their "just keep bumping the core clock" method. While everyone may not make the absolute most of the higher core count stuff, it's only a benefit to the rest of us.

>> No.69171333

>>69171187
It will definitely be bottlenecked by 2 channels but not as much as 32c threadripper on TR platform. It could be mitigated greatly if 4 GHz RAM was supported without OC'ing on B450s successor.

Given how AM4 can use ECC RAM most of the people getting the 12-16c will probably be people interested in cheaper HEDT platforms which will strike another blow to intel's i9 meme platform.

>> No.69171359

>>69171333
>AM4 can use ECC RAM
Have a source on that? Been wanting to build an AMD system for my server but ECC is something I want. I have a Haswell Pentium, Supermicro board, and ECC ram as is.

ECC ram booting up isn't the same as actually having full ECC support.

>> No.69171361

>>69171187

Can you guys stop this memory channel on 16 cores meme?

It all depends on how the cpu is made. if its a 2990wx with numa, then no.

but it isnt. its a 16core with shared memory controller. meaning it doesnt fucking matter.

>> No.69171395

>>69171359
depends on the mobo if it supports it but ryzen itself supports ECC

>> No.69171404

>>69171359
It's mainly motherboard dependent but yeah even on zen+ mainstream B450s ECC RAM support exists.

https://www.asrock.com/MB/AMD/B450%20Pro4/index.asp#Specification

>> No.69171461

>>69167609
>105W TDP
HOUSEFIRE!

>> No.69171480
File: 109 KB, 755x561, aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9DLzkvODA1MjU3L29yaWdpbmFsL2ltYWdlMDA3LnBuZw==(1).jpg [View same] [iqdb] [saucenao] [google] [report]
69171480

>>69171461
At least AMD sticks to their TDPs. Intel clearly doesn't know the meaning of a 95W TDP on their 9900K.

>> No.69171484

>>69171187
Look at it this way - AMD needs high core count chips to complete against Intel in the lucrative server market. So if they are producing tons of those little 8-core chiplets anyways they have no reason to not to offer high core count chips to consumers; many of them will be made from dies that didn't make the server cut.

>> No.69171556
File: 58 KB, 755x561, more is better.jpg [View same] [iqdb] [saucenao] [google] [report]
69171556

>>69171480

>> No.69171604

>>69171071
>heh, not bad, Jim, you made me use 14+++++% of my power

>> No.69171619

>>69171556
>>69171480
>Watts (more is better)
Is it?

>> No.69171628
File: 14 KB, 631x310, 1542220191236.png [View same] [iqdb] [saucenao] [google] [report]
69171628

>>69171619
Yes, of course.

>> No.69171687

>>69171480
obviously you don't know what TDP means. TDP is the amount of thermal energy needed to be cooled when the base clock is applied. The idea being that you need to be able to cool 95W in order to get the guaranteed all core frequency of 3.6GHz, so OEMs know what cooler to use to guarantee the base clock.

The power consumption when applying turbo clock has nothing to do with the TDP,

>> No.69171700

>>69171687
>TDP doesn't matter

>> No.69171736

please tell me there is gonna be a 3600G

>> No.69171765

now watch intel still beat them at single core with 14nm++++++

>> No.69171782

>>69171736
Probably end of the year, APUs always come last.

>> No.69171786

>>69171765
single core is getting less and less relevant by the day, even for games. only games that run better on intel are old single threaded simulators

>> No.69171789

>>69171700
yeah that's what he said. do some research on arguing before posting again, retard.

>> No.69171801

>>69171789
t. coping Incel customer

>> No.69171808

>>69171765
I believe they have basically the same IPC. The only difference was in frequency.

>> No.69171831

>>69171628
>higher is better

>> No.69171843

>>69171480
>>69171556
>Water Cooled Witch Industrial Chiller
N-nani?!

>> No.69171964

>>69169281
NOOOOOOOOOOOOOO

>> No.69171967

>>69170966
The R7 1700 already had a 65W TDP, this one's gonna be even better.

>> No.69171973

>>69171687
Then why does AMD consume less than it's TDP at base frequencies. 2700X will only consume about 65W at 3.7GHz.

>> No.69171977

>>69171071
The STATE of cpu engineers

>> No.69171991
File: 319 KB, 800x800, 1465101708732.jpg [View same] [iqdb] [saucenao] [google] [report]
69171991

>>69171604

>> No.69172001

>>69171687
>TDP is 95W
>CPU uses 200W at base clocks at high load
You don't think this is a little off spec?

>> No.69172003

>>69171480
>>69171556
THE MORE WATTS THE BETTER, GOYIM
BUY BUY BUY

>> No.69172004

>>69171973
AMD also calculates their TDP differently, supposedly.
Whose is more accurate I'm not sure.

>> No.69172015

>>69172001
TDP doesn't matter

>> No.69172025

>>69171786
SOMEONE POST IT

>> No.69172038

>>69171031
AMD have 3 years to do the greatest possible damage to Intel until Intel are going to roll out the Keller's new arch.

>> No.69172070

>>69171480
>>69172001

According to tech Jesus it's actually the MOBO manufactorers doing whatever the fuck they want with turbo boost and baseclock frequency.

Almost every Intel motherboard doesn't give a fuck about Intel's specs and does it's own thing to turbo more, turbo longer has it's baseclock faster than it's supposed to and so on.

>> No.69172133

>>69171071
kek'd heartily

>> No.69172167

>>69171973
Because AMD isn't at the limit of their architecture and manufacturing process.
Personally I assume that they decided to go with the 95W TDP to further their claim that the Ryzen 7 is up to par with Core 7 (since both have the same TDP). A stupid person might assume that 65W TDP might mean a slower product. Maybe they went with it to not get any negative backlash like intel is getting now, since people don't understand that TDP isn't maximum powerdraw in any situation.
Using Prime you can get a Ryzen 7 to draw 150W as well. Is the TDP wrong because of that? Personally I wouldn't say so.

>>69172001
it isn't base clocks, it's turbo clock, essentially the CPU overclocks itself. The TDP only covers the 3.6Ghz Baseclock and that's achievable.

>> No.69172171
File: 16 KB, 480x289, 1535623854013.jpg [View same] [iqdb] [saucenao] [google] [report]
69172171

>>69171628

>> No.69172184

>>69172167
>Because AMD isn't at the limit of their architecture and manufacturing process.

This is all I needed to read right here, thanks.

>> No.69172186
File: 37 KB, 640x360, 1535623727413.jpg [View same] [iqdb] [saucenao] [google] [report]
69172186

>>69172171

>> No.69172210

>>69167642
The ryzen 5 3600x with 10-15% more IPC than 2000 series and a boost clock up to 4.8 is all but tied with the i7-9900K

Ryzen 3300X with same IPC gains and a 4.3 boost and 6cores/12 threads will curbstomp every i5 and run close to an i7 stock.

The 12-16core chips are the real prize though because they offer HEDT power in a non HEDT segment. This is a big deal for content creators and streamers. You need that extra +6+8 cores for rendering and for running background programs including a video stream while you are gaming. Most streamers actually have two separate computers running with one doing video capture for streams. AMD is allowing you to do both on the same platform.

>> No.69172395

>>69172210
delusion

>> No.69172405

>>69172395
autistic screeching is not an arguement

>> No.69172431
File: 205 KB, 807x745, 1545697157024.png [View same] [iqdb] [saucenao] [google] [report]
69172431

>>69172015
>>69172025
>>69171786
>Multi-core doesn't matter!
>Productivity doesn't matter!
>Price/performance doesn't matter!
>Performance per watt doesn't matter!
>Power usage doesn't matter!
>Temperatures don't matter!
>Soldered dies don't matter!
>Stutters don't matter!
>Streaming doesn't matter!
>Data centers don't matter!
>Locked CPUs don't matter!
>OEMs don't matter!
>Hyperscalers don't matter!
>Upgradeability doesn't matter!
>Anti-competitive business practices don't matter!
>Locked platform features don't matter!
>Synthetic loads don't matter!
>PCI-e lanes don't matter!
>Burnt pins don't matter!
>Heat doesn't matter!
>1771w cooler doesn't matter!
>Server space doesn't matter!
>ECC support doesn't matter!
>Free RAID doesn't matter!
>NVMe RAID doesn't matter!
>StoreMI doesn't matter!
>IPC doesn't matter!
>7nm doesn't matter!
>StoreMI doesn't matter!
>HEDT doesn't matter!
>Stock coolers don't matter!
>Security doesn't matter!
>Games don't always matter!
>Enterprise doesn't matter!
>Hyperthreading doesn't matter!
>VMware doesn't matter!
>MySQL doesn't matter!
>Unix doesn't matter!
>Linux doesn't matter!
>Waffer yields don't matter!
>Benchmarks after full patches don't matter!
>Asian markets don't matter!
>Own fabrics don’t matter!
>Chipset lithography doesn't matter!
>Cray doesn't matter!
>Cisco doesn't matter!
>HPE doesn't matter!
>AZURE doesn't matter!
>NEW 5nm doesn't matter!
>NEW TDP doens't matter!
>NEW 10nm doesn't always matter!
>NEW Cache doesn't matter!
>NEW Integrated graphics doesn't matter!
>NEW PCI-Express 4.0 doesn't matter!

>> No.69172444

>>69172431
based

>> No.69172472

How does AMD get twice as many threads as cores?

>> No.69172474

>>69172431
>pwned

>> No.69172491

>>69172472
SMT
Simultaneous multi track drifting

>> No.69172495

Didn't World of Warcraft just get an update to better utilize multicore CPUs? I think Intel is fucked, most games seem to be favoring more than one core now.

>> No.69172546
File: 99 KB, 999x432, ayy.jpg [View same] [iqdb] [saucenao] [google] [report]
69172546

>>69172472
fuck you

>> No.69172568

>>69167609
Reported for antisemitic child porn.

>> No.69172571

Jesus Christ, how does one giant company go from 800lb gorilla to shit flinging spider monkey in a matter of two years?

Oh yeah, jews.

>> No.69172601

>>69171628
The real kicker:
You can a 1950X for the price of a 9900K.
Nevermind the coars, just look at all those sexy PCI-E lanes.

>> No.69172641

They whole jew world order is crashing and burning, and Intel looks like it will go down faster than nvidia.

>> No.69172651

>>69169616
8

>> No.69172709

>>69171556
>2700x 8/16 104.7w
>8400 6/6 117.5w
wtf

>> No.69172725

>>69172651
I'm restraining my hype levels thinking this.
I don't really care if it's 8 or 16 as long as it hits 5Ghz.
But I'm afraid TDP figures would point to at least a 12 core.

>> No.69172737

>>69167609
amazing amd has caught up with technology from 3-4 years ago. well done, you useless faggots. still getting thrashed by intel at per core performance. amd could have a thousand cores on a die and not one of them would out perform a recent intel core.

AMD BTFO YET AGAIN.

>> No.69172745

>>69172709
larger datapaths in the FPU pull more power. Its an AVX torture test.

>> No.69172758

>>69171480
>(throttling)

>> No.69172763

>>69167609
is there any online site for betting for shit like this? because i can easily bet over $1000 that this shit won't happen. any takers?

>> No.69172771

So what do you guys expect from CES?
I expect some blender demo destroying Intel's HEDT platform to nothingness and literally don't wait for 9900k to finish the job.
No mention of clocks.
Coming in April, runs on b350 boards.

I'm gonna jizz all over my screen.

>> No.69172782

>>69172763
Well, you can always just buy Intel stocks.

>> No.69172783

>>69172763
that's the most expensive COPE i've ever seen

>> No.69172785
File: 147 KB, 900x600, 1545320501466.jpg [View same] [iqdb] [saucenao] [google] [report]
69172785

>>69172737
Super sour

>> No.69172807

>>69172783
amd products are always overhyped and they always under-deliver. this will be the same and you can screencap this post right now.

>> No.69172815

>>69169053

>> No.69172866

>>69172815
>engineering sample already spotted
>Now it's 3 sources giving the same naming/core clocks/core counts.
>Looked too good to be true, but maybe 7nm can do that.

>> No.69172894

>>69172807
> overhyped
Clearly.
> under-deliver
They're fine.

>> No.69172912

>>69171071
Need to shoop Jim face on the "Don't trust anyone, not even yourself" pic.

>> No.69172921

>>69172807
The only clearly over-hyped product from AMD I've seen is Vega.
Because they were supposed to have features that were never implemented driver-side.
Name something else.

>> No.69172933
File: 9 KB, 285x249, 1526975633197.jpg [View same] [iqdb] [saucenao] [google] [report]
69172933

>>69167659
yea ok

>> No.69172937

>>69172807
>pigeonshittingonchessboard.jpg

>> No.69172939

>>69172601
>Haha, these leaks are fake, there's no way AMD would sell 16 cores for $450
>1950X is $450
Holy shit ...

>> No.69172974

>>69172921
Bulldozer with its "more powerful than i7" promises.
Polaris with retards believing it'll BTFO 980 (and it did in some games, much later).

>> No.69173018

>>69172974
Yeah, Bulldozer was a turd, even though it seems to keep up better than same time period Intel things.
Polaris was OK. What killed it was cryptos, because it was no longer at the price it was supposed to be.

>> No.69173053

>>69167609

>4.7GHz turbo on the 3800X
>meaning you can manually OC all cores to 4.5GHz all cores/threads @ 125W
>32t @ 4.5GHz

sweet lord.

>> No.69173070

>>69172939
Nicely put.
It's already real.

>> No.69173075

>>69172921
remember "zen 5ghz on air?"
or "zen2 15% higher ipc?"

>> No.69173095

>>69167609
assuming this is true, what I care most about is would it be possible to oc those 3300/3300x cpus to 4.6-4.7ghz. I dont give a fuck about 8c/16t cpus, but I dont want to be forced to buy them just to get higher clocks.

>> No.69173112
File: 534 KB, 1440x2960, Screenshot_20190102-132857_Stockpile.jpg [View same] [iqdb] [saucenao] [google] [report]
69173112

AMD still a meme, pic is my portfolio it's only amd

>> No.69173122

>>69173053
I'm actually expecting all-core turbo to be vastly superior to Intel. Because better efficiency.
It will be the nail in the coffin.

>>69173075
>zen 5ghz on air
No, unless you disable all but 1 core.
>zen2 15% higher ipc
Went from 1600X to 2700X, and that's exactly what I measured.
Reviewers would tell you 5%, but a lot of the increase actually comes from better frequency memory support.

>> No.69173147

>>69173112
>$50 portfolio
wew we have next rockefeller in here

>> No.69173161

>>69173147
It could have been 90 if I sold when my gut said to but alas here I am.
Only put in 30 to start though so I'm basically still an investing genius

>> No.69173170

>>69173095

Maybe, but unlikely. The 3300/3300Xs are the worst dies on the wafer and will generally be the ones with the most defects. You're better off getting the R5s. It's not impossible to find better samples in the R3 segment, but it's like hunting for a needle in a haystack.

>>69173122

True, but even conservatively, a 4.5GHz 16c/32t desktop part would absolutely destroy Intel and their i9-9900k. They would be unable to compete.

On top of that, if fucking toasters have 12t that can all core turbo up to 3.8-4.0GHz, that's insane. Intel would either have to accept a BTFO or make their 8700k chips i3s to compete.

>> No.69173178

>>69173122
>No, unless you disable all but 1 core.
i was mentioning the hype that was generated by some pc hardware magazine back before zen's release. obviously it was bullshit, yet drones fell for it and spammed /g/ with threads similar to this one.

>> No.69173179

>>69172546
the absolute state of amdrones
>d-don't worry about performance
>just close your eyes

>> No.69173182

ITT Intfail shillcels still clinging to the last two threads of hope against big daddy AMDickings

>> No.69173184
File: 83 KB, 1600x1200, nintendo-on.jpg [View same] [iqdb] [saucenao] [google] [report]
69173184

>>69167609
It's interesting to see the /g/ equivalent of the fake console specs threads.

>> No.69173191

>>69167609

I'd love to see how the 3800X compares to the 3700X in gaming performance. Whether having 4 less cores and 8 less threads, but 300MHz higher all core/thread clocks will do against 4c/8t more on the 3800X with 300MHz less all over.

>> No.69173204

>>69173161
>It could have been 90
That's like what most people make in an hour damn dude.

>> No.69173206
File: 108 KB, 782x782, Smerchant.jpg [View same] [iqdb] [saucenao] [google] [report]
69173206

OY VEY WHY AREN'T MY POWERS WORKING

>> No.69173226

>>69173184
see
>>69173182

It's fun to see you No Hopes keep trying though.

COPE

>> No.69173235

>>69173184
>coming soon: 5Ghz doesn't matter!

>> No.69173254

>>69173226
>>69173235
Well, i don't think it is necessarily false, but could be.
It's just speculation, numbers tossed at the wall, fakes.
And they go both ways.

There's even the possibility mommy Su gets in the stage and announce shit that completely blows those specs out of the water.

>> No.69173260

>>69173161
wow anon, 90USD is a lot of money, you could buy almost 10 year old tech with that.

>> No.69173278

>>69173204
>Most people make $90/hr

I wish

>> No.69173358

>>69173254
>mommy Su gets in the stage and announce shit that completely blows those specs out of the water.

meme responsibly anon.

>> No.69173438
File: 23 KB, 399x393, Flame hyetard.jpg [View same] [iqdb] [saucenao] [google] [report]
69173438

>>69173358

>> No.69173452

Fuck the flagships, I don't have $2k to blow. Can someone spoonfeed me the Intel (Xeon?) price/performance equivalents of Ryzen 5 3600? I see that it's ~33% faster than the E3-1230V3 I'm using since 2013, at less $$ and TDP, which is kinda neat.

>> No.69173469
File: 137 KB, 1280x720, This cpus existence is literally impossible.jpg [View same] [iqdb] [saucenao] [google] [report]
69173469

>it fails the common sense test

Thats why its impossible for AMD to release a 8c/16t cpu on the mainstream platform for less than 1000 burgers.

>> No.69173483

>>69173452
>Can someone spoonfeed me the Intel (Xeon?) price/performance equivalents of Ryzen 5 3600?
There isn't. Now buy a platinum xeon.

>> No.69173532

>>69167609
3600 - 16cores @ 4.4GHz @ 65w
yup, time to upgrade.

>> No.69173574

So is that chart talking real possibilities of upcoming 7nm AMD chips? Because I'd sincerely like to see how 16c32t would work on AM4. I'm riding the fence right now if I should get Threadripper, or just wait for 7nm because I encode a lot of video and just need the grunt work of cores/threads. The quad channel memory and plethora of PCIe lanes of Threadripper are lost on me. Would gladly pay $450+ for the 3800X.

>> No.69173589

>>69173161
You do realize, you're gonna pay whatever benefit you got in taxes and trading commission, right?

>> No.69173608

>>69173574
Zen core is already A LOT more efficient than Intel garbage, EVEN ON AN INFERIOR NODE

Now imagine 7nm

>> No.69173619

>>69173574
Looks about right to me.
7nm is a big step, and a totally different node from what AMD had before.
I'll take a risk and say you'll actually be able to overclock those parts.

>> No.69173628

>>69173608
t. Computer engineer

>> No.69173629

>>69173589
>trading commission
what kind of commie country are you from?
>taxes
You only pay taxes from what you've earned so if he invested 30 he would only pay taxes for that 20. Learn how trading stock works before posting your bullshit.

>> No.69173646

>>69173629
Hello.
You gotta pay that guy in a suit that bundles your order with all the others.
How much do you think Cocaine and whores is worth?

>> No.69173658

>>69173628
The 14nm process on Zen is a bit behind Intel's 14nm, dumbass. And it's still more efficient.

>> No.69173717

>>69173646
I just checked like 10 different broker accounts and the worst one I found has big whooping 0.2% commison on each transaction. That's about $30*0.002.

>> No.69173728

>>69173469
It also "fails the common sense test" for AMD to be on a superior node compared to Intel's node huh?

>> No.69173763

>>69173717
I guess they just stopped having whores and drugs.
I never really meddled into stocks, but I remember 30€ fees on one bulk transaction, no matter the amount.
You must have not read the right paragraph.
Those guys make upwards of 400k a year.

>> No.69173773

>>69173728

thats the point, people thought AMD's performance on the 1800X was fictional, until it wasn't

It was also supposed to be fictional for the price to be as low as it was, but it wasn't

>> No.69173777

>>69173608
>>69173619
Really hoping their 7nm pans out the. How's Intel's 10nm looking as well? And their security flaws? We're their flaws ever patched out at the cost of performance like people were saying? Been outta the loop since I bought my 1700X.

>> No.69173793

>>69173763
>but I remember 30€ fees
>euro
ah now it explains everything, so which cucked euro country are you from? So I can avoid it in the future.

>> No.69173851

>>69173777
>Were(here, I fixed it for you) their flaws ever patched out at the cost of performance like people were saying?
Yes they were, and it's a fucking disaster.
Disable HT for a start. And then some.
Overall, they perform 40% less than they did.
Or they just keep running the security flaws.
To be frank, they're not easily exploitable. But they're still there.

>> No.69173862

>>69173793
Just sell your 'portfolio' and buy a few beers or something.

>> No.69173881

>>69173777
>Intel 10nm
A colossal fuckup. Even if they do manage to magically pull a 10nm node out of their ass tomorrow, it'll take another iteration or two to have it perform better than their current 14nm+++++++ shit. Intel admitted that themselves.

>> No.69173926

>>69173862
I'm not that guy I have close to 50k in stock, so I know how it works and if you pay commision and especially 30 fucking euro commison you're dumb.

>> No.69173954
File: 272 KB, 2302x1010, qfbgger6k2821.png [View same] [iqdb] [saucenao] [google] [report]
69173954

Looks like Jim might have been right. This is taken from Gigabyte's website.

>> No.69173958

>>69173926
I mean, his is not Crypto.
This is never going to the sky.
You'll never win more than a few bucks on it.

>> No.69173976

>>69173954
True if as big as my dick right nao.

>> No.69173992

>>69173954
N-no Gigabyte is lying there's no way they'd know that it's all fake! FAKE!

>> No.69173998

>>69167609
>TSMC:
>Our 7nm process will allow for 1.7x the transistor density
>Or 40% faster clock speeds
>or 60% less power
Then
>AMD:
>We've doubled our core density!
Finally
>Fanboys and/or shills:
>AMD doubled the cores, added 25% IPC, added 15% clocks at the same power for half the price!

Only one of these groups are making shit up.

>> No.69174008

>>69173998
What bad timing.
Cry more.

>> No.69174021

>>69173954
Notice the 32MB L3 also. What a beast in the same tdp as zen+.

>> No.69174022

>>69173992
Well as far as I can tell that picture actually is.

https://www.gigabyte.com/Motherboard/X470-AORUS-GAMING-7-WIFI-rev-10#support-cpu

>> No.69174035

>>69173954
>18C/32T
wut?

>> No.69174046

>>69173170
It's gotta be at least 3 gens old to qualify as a toaster, fren

>> No.69174047

>>69173954
i don't see it
https://www.gigabyte.com/Motherboard/X470-AORUS-GAMING-7-WIFI-rev-10#support-cpu

>> No.69174054

I can't wait for CES, it's finally time to replace the ol 2500k

>> No.69174080

>>69174054
If you're going full upgrade, you might as well wait for zen3.
DDR5 is gonna make them cores actually work.

>> No.69174082

>>69174035
Your screen isn't big enough to display the image and the scaling artifacts make it seem like 18C

>> No.69174088
File: 74 KB, 960x818, H.jpg [View same] [iqdb] [saucenao] [google] [report]
69174088

>>69173998
They're all correct, but the devil is in the details.
AMD doubled the cores and got em cheaper by removing pretty much all the I/O from the dies, and moved all to this center, still 14nm big thing.
The next AMD offerings will ALL use the exact same 7nm chip, but the I/O chip will be different on ryzen/TR/Epyc etc..
On low end 8core chips, they will use two defective 7nm dies that only have 4 cores working on each, and on the 16 core chips you have two with 6 cores working and so forth.

>> No.69174103

>>69174008
what is price for $400, Alex?

>> No.69174104

>>69173954
>no archive link

>> No.69174106

>>69173954
I am rock hard right now.

>> No.69174109

>>69174088
IndianTV, all but admitted consumer zen2 would have a 7nm IO die.
See, it has to fit in the fucking processor.

>> No.69174134

>>69173998
The Rome CPU AMD showed off was 64c/128t, clocked at 2.35ghz(faster than all prior 32c/64t Epyc SKUs) and it has an IPC uplift over Zen1.
You're retarded.
Your post is retarded.

>> No.69174135

>>69174109
Different dies then? To do that they must be really confident in Zen2 performance (increased costs and all)

>> No.69174144

>>69174109
Indian? I thought that he was irish
Now that I'm thinking about it, he sometimes sounds indian and sometimes sounds irish

>> No.69174150

>>69174103
400$ will be 3700X and the next mainstream/enthusiast processor.
5 fucking Ghz 12 core.
Wiping the floor with 9900Ks all day long.
AMD is really going for the throat this time.

>> No.69174187

>>69174144
Are you deaf? AdoredTV is scottish.

>> No.69174188

>>69174109
They can actually saw the I/O die in four apparently.

>> No.69174190

>>69174144
Liking AMD does that to you.

>> No.69174219

Goddamn I wish I had AMD stocks

>> No.69174223

>>69173452
To answer my own question:
Ryzen 5 3600 - $220
i7-8700 - ~11% faster, $306
E-2136 - slightly lower msrp than above but not in stock, and higher tdp so just ignore this shit
I'm not actually running ECC RAM or > 32GB. E3-1230v3 was a ~$50 cheaper i7 4770 just without igpu and slightly lower tdp.
i5-8600 - ~3% faster, $237
(Ignore E-2126G as well.)
i5-8500 - "~same effective speed" off userbenchmark but significantly slower multi core benches (duh AMD), $225
Overall the Ryzen appears the solidest practical value perf/$ at 65W.

>> No.69174230

>>69174187
many people have "identified" him to be indian. stupid niggers everywhere.

>> No.69174236
File: 45 KB, 690x388, 1515197493906.jpg [View same] [iqdb] [saucenao] [google] [report]
69174236

>>69174150
FINISHED

>> No.69174237

>>69174188
Nah Jim is but your average Youtuber.
He didn't see shit, but got information.
Now, the source producing those leaks could be a fucking troll.
But it's a troll 3 different sources chose to trust.

>> No.69174248

Does anyone have that Tom's Hardware "Number of Hardware Insecurities: Higher is Better"?

>> No.69174266

>>69173954
Looks fake

>> No.69174276

>>69174187
My only reference to the scottish language is Limmy and Adored sounds nothing like him, but if he actually is scottish, my bad

>> No.69174309

AMDrones will be staying poor & low FPS this year, just like every year.

>> No.69174311

>>69174248
>>69174286

>> No.69174313

>>69174276
Not only is scottish not a language there are many accents within scotland.

>> No.69174372

>>69171480
>>69171556
HIGHER IS ALWAYS BETTER YOU AYYMDNIGGER I DON'T SEE A FUCKING PROBLEM HERE

>> No.69174444
File: 71 KB, 1200x600, hf.jpg [View same] [iqdb] [saucenao] [google] [report]
69174444

>>69171480
https://www.youtube.com/watch?v=_3eC35LoF4U

>> No.69174470 [DELETED] 

>>69173954
>32nm cache
install windows 98 on cache kek

>> No.69174472

>>69174309
screencapped

>> No.69174492

>>69174223
Why tf i7-8700 $270 on ebay newegg but not newegg newegg?

>> No.69174560

>>69174492
ebay newegg is where the smart people shop

>> No.69174565

>>69174047

Seems like they took it down.

>> No.69174567

>>69174223
Where are you getting these "faster" numbers from?

>> No.69174603

>>69169281
Only in Geekbench can a laptop i5 get twice the score of a desktop i7 because it was running Android inside VirtualBox

>> No.69174640
File: 158 KB, 712x1441, this can't be happening.png [View same] [iqdb] [saucenao] [google] [report]
69174640

>>69173628

>> No.69174711

>>69174640
Would that change if Ryzen had a full 256-bit AVX unit?

>> No.69174726

>>69174711
No. Pentium has no AVX. Check the image again.

>> No.69174731

>>69174711
>Would it change if it used proprietary Intel gimp instructions.

>> No.69174764

>>69174731
It's only gimped because Intel's implementation sucks (kek lower clocks or I'll burn your house down). I'm sure AMD can do vastly better

>> No.69174859

>>69174711
>>69174726
Intel is very bad and inefficient even on their own instructions. It's similar to how StoreMI makes Optane work better on Ryzen than Intel.

>> No.69175116

>>69174603
what about a crappy 4c/8t APU that easily outperforms 99% of CPUs in multithreaded

>> No.69175268

Hello,
I have been out of the loop for awhile. I haven't upgraded my computer in a long while. Can someone spoon feed me on what happened to the CPU, once upon a time it was understood that Intel cpus were better but now they arent? I know of the spectre exploits and so on but performance/dollar is now better when it comes to AMD?

>> No.69175285

>>69175268
Intel is still king but advancement stopped at Sandy Bridge.

>> No.69175288

>>69175268
Intel is still better
AMD is for the poor

>> No.69175316

>>69175268
Every once in a while there is a rumor that AMD has finally caught up but it never turns out to be true. What you are witnessing in this thread is just the latest iteration of that.

>> No.69175345

>>69175285

So that means people are just waiting for AMD to keep increasing their performance until it catches up to Intel? What is the point of this thread if that is the case? However should I not buy Intel with the exploits that have come to light?

>> No.69175357

Imagine buying indiantel.

>> No.69175361
File: 349 KB, 220x220, 1545936192408.gif [View same] [iqdb] [saucenao] [google] [report]
69175361

>>69172186

>> No.69175364

>>69175116
Yeah, and how somebody got a 2400g into an iMac. Or does Geekbench just assume it's an iMac because it has OS X on it?

>> No.69175393

>>69175345
You should avoid Intel pozzed products altogether. They have literal security AIDS.

>> No.69175397

>>69175345
Nothing you can do to avoid exploits. They are built into everything. One day AMD may catch up to Intel at which point Intel will just release whatever they have been working on for the last 8 years and crush them again.

>> No.69175428

>>69175397
Just wait for new uarch in 2025, Intel will surely beat AMD completely by then.

>> No.69175436
File: 72 KB, 890x400, 1535123564509.jpg [View same] [iqdb] [saucenao] [google] [report]
69175436

>>69175397
Israel cannot save them this time, schlomo. Keep coping

>> No.69175470

>>69175345
No, people are rooting for AMD to succeed so they can buy Intel and NVIDIA cheaper

>> No.69175473

i want to be lisa su's boy toy

>> No.69175481

>>69175470
Intel fabs are dead, sadly Intel will go under in less than 10 years

>> No.69175531

>>69175481
FINISHED AND BANKRUPT

>> No.69175653
File: 35 KB, 675x827, 1543412994286.png [View same] [iqdb] [saucenao] [google] [report]
69175653

Hey AMD Bros.
What's our excuse gonna be when these leaks are false?

>> No.69175657

>>69175473
lewd

>> No.69175696
File: 62 KB, 600x600, 1546411700551.jpg [View same] [iqdb] [saucenao] [google] [report]
69175696

>>69175653
>still coping this hard
Sad!

>> No.69175717

>>69175345
The point is AMD still has a niche use case for contrarians and people with little to no money.

>> No.69175725

>big chance this is true
>bought zen plus a month ago
mfw...

>> No.69175770

>>69167609
well that 3700x looks pretty good any info on the price ?

>> No.69175775
File: 1.25 MB, 1845x1923, 1501685446886.jpg [View same] [iqdb] [saucenao] [google] [report]
69175775

>>69175725
It will still be there when you need an upgrade.

>> No.69176106

>>69175653
Thanks god it was not that slow.

>> No.69176236

>>69167609
post a SOURCE or LINK you stupid faggot

>> No.69176421

>>69174021
>32MB
This is telling as the chiplets should have 32MB each. If it's not a lie then ryzen 2 has cut cache. I still think the leaks are missing the high core 65w parts. I want those 16cores at 3ghz to put into tiny low power render itx boxes.

>> No.69176526

i bought a 2600x half a year ago, did i fuck up?

>> No.69176602

>>69176526
No. It just means in another year and a half you'll upgrade to an even faster CPU.

>> No.69176612

>>69175725
I just built a 2700x system a couple months ago and I'm cool with it. The more the merrier as far as I'm concerned.

You want good, cheap CPUs to proliferate. The Intel chips were fine and all but if AMD cranks out low-end chips with twice the core count and similar or better clocks and that becomes toaster-tier that means software of all types (and games naturally) can move to being more heavily multithreaded.

Win win. If AMD can try to capture the low/mid-end market too in GPUs you'll be able to build some pretty badass toaster systems for cheap.

>> No.69177191

>>69174444
Check'd

>> No.69177509

Completely fake. They just copy + paste earlier troll "leaks" which were confirmed to be fake.

AMD isn't going to be cannibalizing their Threadrippers yet with 12-core/16-core Ryzen SKUs. Dual-channel DDR4 will not cut it for 12/16 core chips. Most Socket AM4 boards don't even have the power delivery to handle it at high clockspeeds.

It is far more likely that AMD will opt for single chipset (4/6/8 cores) + iGPU/I/O chip setup for desktop/laptop Zen2 which will uniform both line-ups and address their strategic disadvantage of not having ubiquitous iGPU support (OEMS want this)

They just need to make it performance competitive with Coffee Lake R while consuming far less power at maximum load and be cheaper to make en mass.

Intel would lose their complete shit in the mainstream/laptop market.

>> No.69177540

>>69177509
>which were confirmed to be fake
[citation needed]

>> No.69177607

>>69167609
That 3700X is looking good.

>> No.69177623
File: 564 KB, 800x430, goldface.gif [View same] [iqdb] [saucenao] [google] [report]
69177623

>>69169281
Intel shills BTFO

>> No.69177629

>>69177509
seething

>> No.69177656
File: 146 KB, 1200x1200, 803249860652.jpg [View same] [iqdb] [saucenao] [google] [report]
69177656

>>69171071
>>69171604

>> No.69177657

>>69177540

TSMC's process is new and the yields aren't going to be that stellar for AMD to use two die chips for desktop/laptop market. They also aren't to going be making a separate Zen2 chiplet (which includes I/O logic) for desktop/laptop either.

We already know that Zen2 chiplets only have CPU logic which means they a separate I/O chip either on-package (Rome has this) or chip (Northbridge). The reason is that AMD was unable to get TSMC's 7nm to work I/O stuff with acceptable yields. They are forced to use GF's 12/14nm for I/O chip.

They are going to stick with two die setup for their AM4 packaging. One is the Zen2 chiplet and other for I/O chip + iGPU (Probably little Vega). There isn't enough room for three or four die for AM4. It would require new CPU packaging. AMD is waiting until DDR5/PCIe 4 to finalize before changing desktop/laptop packing. Zen3 should be ready by then.

>> No.69177709

>>69172763
you could create a bet on augur

>> No.69177777

>>69177629

Just taking healthy doses of reality. The desktop/laptop market doesn't need or care for 12cores/16core. It is still professional/prosumer-tier. AMD already has Threadripper line-up for that market.

AMD needs to make itself attractive to OEMs in laptop/desktop market. More cores without iGPUs will fail to achieve that goal. They need to make iGPUs ubiquitous on those platforms while being a cheaper option then Intel (This is where 7nm and chiplet strategy will shine).

>> No.69177890

>>69177777
checked

>> No.69177914

>>69177777
checked
and the desktop market does care for higher amounts of cores. For example: I'm a student that needs to run a lot of VM's but don't have to money for a server or threadripper with an expensive motherboard.
You're right about the OEM part though, though low end 75 watt gpu's do help a lot.
They should make higher end APU's

>> No.69177932

>>69177777
They'll also have 8 cores with iGPU. People who rely on an iGPU don't care much about performance anyway

>> No.69177976

>>69174109
>>69174230
Can someone please explain to me how his accent even vaguely sounds Indian?

t. bong, I hear both a lot

>> No.69177999

>>69171031
Its Revenge, its another holocaust

>> No.69178062
File: 24 KB, 267x297, 1502488836418.jpg [View same] [iqdb] [saucenao] [google] [report]
69178062

AMD STOP THIS MADNESS YOU ARE NO LONGER JUST ROCKING THE BOAT YOU ARE SINKING IT

>> No.69178151

>>69167609
>>69167609
Oh look it’s this thread again, same shit different year. Another Massive Disappointment

>> No.69178161

>>69177657
>They also aren't going to be making a separate Zen2 chiplet (which includes I/O logic) for desktop/laptop either.
Does it matter? We know desktop/laptop will be using an I/O chiplet.

>The reason is that AMD was unable to get TSMC's 7nm to work I/O stuff with acceptable yields.
False. They're using 14nm there since there's not much gain to be had by shrinking an I/O die down to 7nm.

>They are going to stick with two die setup for their AM4 packaging. One is the Zen2 chiplet and other for I/O chip + iGPU
You just contradicted what you said earlier about them not making a separate Zen2 chiplet for desktop/laptop since if it's the same as Epyc it won't have an iGPU. OEMs only care for iGPUs on their 4C offerings anyways which AMD will provide.

>> No.69178173

>>69178062
It's the only way to unleash the new era in computing.
Intel sits on their ass forever until someone come and kick their asses, then they fucking innovate as fuck... and stagnate again.
It was like that on the 8bit CPUs, it was like that on the 16bit CPus, it was like that on the 32bit CPUs and so forth.

>> No.69178670

>>69172431
Based and flops-pilled.

>> No.69178702

>>69169281
>8709 MHz
Oh shit

>> No.69178733

>>69169281
Geekbench is so fucking shit, goddamn.

>> No.69178755
File: 35 KB, 583x439, 1515896901860.jpg [View same] [iqdb] [saucenao] [google] [report]
69178755

>>69167609
give me a 5ghz 8 core part OR ELSE

>> No.69178784

>>69178755
Hard to do without going full Intel. Thermonuclear, that is.

>> No.69179091

My BBC is Ryzen and Incels have a TWC

>> No.69179253

>>69178784
Just use a chiller bro

>> No.69179284

Shills protecting corporations...
https://www.youtube.com/watch?v=aQJ5GvupxIs

>> No.69179368

>>69171361
>its a 16core with shared memory controller
Even AdoredTV, the true chiplet fanboy, has stated that Ryzen 3 very likely doesn't use an I/O die.

>> No.69179395

>>69172745
Not sure that explains it though, since Zen has four 128-bit FPUs, compared to Intel's two 256-bit FPUs.

>> No.69179411

Perfectly happy with my 2700x
But it's so nice to see Intel grt shit on after 10 years of mediocre quad cores rehashed over and over

>> No.69179553

>>69167609
Reminder: I was on /g/ 4 months ago and I told all of you that these were the clocks and core counts in the Ryzen 3000 SKU lineup and every single intel fag laughed and said it was IMPOSSIBURRUU!

well prepare to see intel's stock value make like the Titanic and hit a fucking iceberg.

and as an added fuck you to the nvidia fags I have even more insider info.

The reason Nvidia rushed their RTX cards to market is because they wanted to be the first to have "ray tracing" and collect mindshare. The reason they did this was because AMD's Radeon division has already been working on Compute-based ray-tracing and their RT implementation is superior to the bullshit Nvidia is peddling.

Essentially Nvidia wanted to get RT cards out on the market first and get to the game developers with their properietary Raytracing before AMD creating a repeat of Nvidia's gameworks features which broke game engines and shut out AMD graphics users.

Wont work this time because AMD is putting graphics in the consoles and no game dev is stupid enough to use Nvidia's RayTracing which barely even works on a 2080ti in their games over AMD RayTracing that will work on the new next gen Consoles with AMD GPU's

Protip: An anon posted a pic earlier showing AMD CPU's performing much higher than intel ones when paired with a 2080ti running ray tracing. That has something important to do with what AMD is cooking up in their Zen+Navi SOC's

Remember the talks they gave on heterogeneous system architectures.

>> No.69179612
File: 52 KB, 639x624, DTd5zI5VAAAlI_r.jpg [View same] [iqdb] [saucenao] [google] [report]
69179612

>>69171604

>> No.69179675

>>69179395
Buddy, Zen has two 128bit FMACs. Get the fuck out.

>> No.69179711

>all this Intel mandated FUD
I can hear them shaking all they way from Israel.

>> No.69179752
File: 266 KB, 2000x1100, s1%20Perf.png [View same] [iqdb] [saucenao] [google] [report]
69179752

>>69179675
It does in fact not have FMACs at all, strictly speaking. It has 2x128-bit MULs and 2x128-bit ADDs, and they team together in some weird way that even Agner couldn't fully explain when doing FMAs (it seems they're operating at some weird fractional capacity like 2/3). But that's still 4 128-bit FPUs, so with the right instruction mix you can get 4 128-bit ops done per cycle.

>> No.69179766

>>69178161

Wrong kid, AMD was unable to get 7nm working with I/O logic with acceptable yields and meet to launch schedules. It would be easier to get everything in a single package instead of having to split it up. AMD took a compromise so they can keep yields acceptable and meet launch schedules, once 7nm process matures more we will probably see a design that will incorporate I/O and CPU together again.

OEMs care about iGPUs for normal desktop/laptop SKUs. That's why Intel is able easily weather out Ryzen storm in the OEM world. AMD needs to match them if they to seriously snatch marketshare in this segment.

They don't care about /v/tards and wannabe enthusiast.

>> No.69179787
File: 72 KB, 589x488, ss.png [View same] [iqdb] [saucenao] [google] [report]
69179787

>>69179675
>>69179752
Also Agner's pic in case you're still in doubt.

>> No.69179791
File: 101 KB, 1920x1080, 7-1080.1262858105.png [View same] [iqdb] [saucenao] [google] [report]
69179791

>>69179752
>It does in fact not have FMACs at all,
Literally commit suicide, retard. Its literally two 128bit FMACs.
A multiply by itself is not an FPU. An add by itself is not an FPU. The FMAC is a multiply and add combined, hence why its called an FMAC.

You could have just looked at any other part of the HotChips presentation instead of trying bluff and talk out of your ass.

>> No.69179793

>>69177914

You live in fantasy world. The overwhelming majority of the desktop/laptop market does not care about having more cores. 4-cores is still perfectly fine.

VM-fags are using HEDT platforms not shitty desktop systems (Which do not have proper IOMMU support)

>> No.69179804

>>69179787
A fucking multiply by itself is not a complete FPU. Zen does not have four FPUs. It has two FMACs, you retarded little kid.

>> No.69179831

>>69179804
>It has two FMACs, you retarded little kid.
It literally does not. It has two execution units that can do MULs (along with other things), two execution units that can to ADDs (along with other things), and no execution units that can do FMAs by themselves, but does those by ganging up the MUL and ADD units. It's not that hard to understand.
>A fucking multiply by itself is not a complete FPU
Of course not, the FPU as such is all the four execution units along with the scheduling logic and register files taken together, but that doesn't change the fact that, given the right instruction mix, you can get four 128-bit ops issues, executed and retired, per cycle.

>> No.69179833

>>69179675
>>69179791
>>69179804
So in summation, Zen has two 128bit FMACs. It does not have four FPUs. Intel's Core family's latest iterations have massively wider FPU datapaths with drastically higher transistor counts. That increased width is what causes their cores to pull so much power when under heavy AVX load. 256 bit datapaths are far more power hungry than 128bit datapaths.

First year CS drop outs shouldn't talk out of their asses.

>> No.69179848

>>69179831
>AMD IS WRONG AND THEIR FMACS AREN'T REALLY FMACS!
>THOSE ZEN ENGINEERS DON'T ACTUALLY KNOW ANYTHING ABOUT ZEN!
Sure, kid.

>> No.69179853

>>69179553
They're playing the super duper long game, aren't they?

>> No.69179862

>>69179848
>build as 4 pipes, 4 Fadd, 2 Fmul
Did you even read it?

>> No.69179876

>>69179553
>The reason they did this was because AMD's Radeon division has already been working on Compute-based ray-tracing and their RT implementation is superior to the bullshit Nvidia is peddling.
Citation? I'd love it if it's true, I've been hating on nVidia's shitty dedicated-hardware solution since I first heard about it.

>> No.69179898

>>69179853
Nvidia fucked up by only having RT on their high end cards and then not having it work properly even on their flagship 2080ti

even now with fixed drivers you are barely getting 60fps at 1440p with raytracing on the 2080ti.

Its a disaster and when game devs see AMD's raytracing working flawlessly on NAVI GPU's in the console SOC's they will jump ship so fast.

The average gamer is on a console, integrated graphics or a low end (sub 250$) graphics card.

AMD has working RT for Navi 10 which is in next gen consoles and will be available in desktop cards for ~250$ and less.

>> No.69179918

>>69173469
they were so whipped by Intel that AMD came in, shot Intel, then smacked the media across the face

>> No.69179942

>>69179862
>trying to grasp at straws after talking out of your ass to someone who knows far more than you,
You tried to argue that Zen had 4 FPUs. You started this off saying that say had four 128bit FPUs. You can't back track this.
You referenced a basic bitch programming reference that doesn't support the claim you tried to make and only shows that you don't actually know what any of these words mean.
You then tried to argue that Zen doesn't have FMACs despite AMD explicitly referring to them as such, because thats exactly what they are.

How do you live with this degree of autism?
The Zen FPU is two 128bit FMACs. intel's FPUs have larger datapaths, and they draw more power. Thats all there is to it.
Its time to stop posting.

>> No.69179944

>>69179898

It is pixel/vertex shading at its infancy again.

>> No.69179972
File: 1.28 MB, 1920x1080, Screenshot_20181204_16.12.40.png [View same] [iqdb] [saucenao] [google] [report]
69179972

>>69179766
>AMD was unable to get 7nm working with I/O logic with acceptable yields
[citation needed]

>OEMs care about iGPUs for normal desktop/laptop SKUs.
Exactly. That's where the *G parts come in.

>> No.69180033
File: 103 KB, 1122x631, die stacking.jpg [View same] [iqdb] [saucenao] [google] [report]
69180033

>>69179766
Die stacking as a concept exists not just because it lowers costs, which is one of the biggest reasons its being pursued. The other half of it is that different types of ICs do not all have optimal performance/power on the same process. Different circuits can attain better power and performance on their own specialized processes.

Its unlikely that AMD would have tried to make a 7nm IO given the cost and complexity of it on such an expensive and immature process. Its likely the case that TSMC doesn't even have a full portfolio of IP blocks ready to support everything the Bridge Chip handles in 7nm. AMD most likely began pursuing a special vt of 14LPP years ago for this purpose. They've been working on die stacking for years.

>> No.69180035

>>69179942
>You tried to argue that Zen had 4 FPUs.
Is the terminology issue all that you're complaining about? Because if so you really should have been able to read that from the fact that, obviously, no CPU core whatsoever has more than one FPU per the strictest possible definition. In the same sentence, I said that Intel has 2 256-bit FPUs, which by the same standard is just as false, yet you didn't complain about that. It really should have been clear from the context that what I meant was four 128-bit FP execution units for Zen vs two 256-bit FP execution units for Intel, and if that's all you're complaining about, you're being quite autistic.
>You then tried to argue that Zen doesn't have FMACs despite AMD explicitly referring to them as such, because thats exactly what they are.
Those are marketing slides, you dumbass, obviously they're not going to be using the strictest possible terminology and be vetted by scientists (just like my first post which you autistically replied to). As I pointed out, the same slide even points out that it's "built as 4 pipes", implying that the FMAC concept they're referring to is more abstract, rather than the physical implementation, which of course aligns with everything I've said and also the reference to Agner that I posted, in that Zen can dispatch two FMA ops per cycle. Plus other ops as well.
>The Zen FPU is two 128bit FMACs. intel's FPUs have larger datapaths, and they draw more power.
And there lies the core issue. Intel has two 256-bit FP execution units, Zen has four 128-bit FP execution units. 2*256 == 4*128.

>> No.69180068

>>69180035
>hurrr durr everything I said is retarded but you didn't nitpick every single thing I said wrong!
>HotChips technical presentations are just meaningless marketing slides! They're wrong! AMD is wrong about their own designs!
Imagine being you.

>> No.69180113

god the drivel in this thread is nasty

AMD still has the worst latency and throughput for instructions as basic as pext and pdep and can't even implement avx2 let alone avx512 or any notable instructionsets for that matter

there's a reason why this shit is always in budget builds

>> No.69180116
File: 916 KB, 245x285, okay-1.gif [View same] [iqdb] [saucenao] [google] [report]
69180116

>>69180068
Whatever it takes to get over your failures, Anon.

>> No.69180132

>>69179972

The circumstantial evidence is overwhelming with Rome's official debut. AMD couldn't get a 7nm CPU design with build-in under acceptable yields. That's why it has huge 14nm I/O chip instead of splitting up on each on the chiplet like they did with Zeppelin.
>>69180033
That's marketing non-sense trying to rationalize their short-term decision.

>> No.69180134

>>69180113
>instructions as basic as pext and pdep
lol. I assume your income and livelihood depends on running GlassFish 24/7, right Anon?
>can't even implement avx2
Zeb 1 does in fact implement AVX2.

>> No.69180145

>>69180134
>GlassFish
Durr, Stockfish.

>> No.69180148

>>69171359
>full ECC support
Ryzen CPUs do have full ECC support but the AM4 platform as a whole does not. Some motherboards do, some don't. You have to check. And don't assume that a high-end x470 motherboard has ECC support, sometimes they don't. It's quite random, there's low-end B350 boards with support and high-end x470 board without.

Also: ECC memory support differs from non-ECC, motherboards have listings like
>Support for ECC Un-buffered DIMM 1Rx8/2Rx8 memory modules
>Support for non-ECC Un-buffered DIMM 1Rx8/2Rx8/1Rx16 memory modules

>> No.69180171

>>69180145
it ""implement"" avx2 by literally running its 128-bit units twice, and not actually having 256-bit units
it's TWICE as slow

>> No.69180181

>>69180116
Says the CS drop out that demonstrably doesn't understand what an FPU datapath is, since you're trying to argue that Zen and Skylake have the same throughput, ignoring actual flops, ignoring transistor count, ignoring die area.

>>69180132
Die stacking is the path the entire industry is going to take in coming years.

>> No.69180214

>>69177777
Fair point, AMD would probably be better suited to having a 6 core 12 thread cpu with like a 1050 level GPU built in to it.

>> No.69180231
File: 469 KB, 1432x2085, Screenshot_20190103-115341~01.png [View same] [iqdb] [saucenao] [google] [report]
69180231

Get ready for buttfuck!!!!

>> No.69180253

>>69179553

I just came

>> No.69180294

>>69179944
Ray Tracing is the in-between step to the ultimate end goal of Path traced rendering which will ultimately replace rasterization

AdoredTV did a video on Path-Tracing a while back which will explain it fully.

Path tracing with enough ray calculations per pixel will give us true next generation graphics but it requires an insane amount of raw compute performance. A flagship card can only do limited path tracing with a couple of rays per pixel at about a single frame per second.

Ray Tracing is a solid in-between step that will improve light-reflections and also light-source positioning and light diffusion to increase visual quality from a lighting perspective. But ultimately its still Rasterization as the base rendering of the frame.

>> No.69180306
File: 64 KB, 650x645, tightdress.jpg [View same] [iqdb] [saucenao] [google] [report]
69180306

>>69179972
just looking at that graph makes me smile, fuck I love AMD.
>tfw I haven't had an AMD since the Opteron 165 S939 days.

>> No.69180350

>>69180181
Reminder that Zen is 8 FLOPS/cycle
Top end Skylake-SP is 32 FLOPS/cycle
Zen2 will be 16 FLOPS/cycle

FLOPS aint free, intel pays for it in power.

>> No.69180975

>>69172210
!!!

>>69172395
Implying I didn't notice the massive usability drop when I went from a 8350 to my current 4790k. Game for game i got better FPS but when i tried recording games while playing, i felt that shit.

Didn't Overclock either, factory speeds, same 32GB of ram used in both setups.

I'll probably make the move to Ryzen if I can stop blowing money on building cars/trucks.

Turbo Vortec BBC ain't cheap yo.

>>
Name (leave empty)
Comment (leave empty)
Name
E-mail
Subject
Comment
Password [?]Password used for file deletion.
Captcha
Action