[ 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: 160 KB, 1009x1055, wa wa wa waaaa.jpg [View same] [iqdb] [saucenao] [google] [report]
60880060 No.60880060 [Reply] [Original] [archived.moe] [rbt]

It's another Intel Frylake is dead on arrival thread.

>> No.60880151

Amd unironically did it

>> No.60880204

are we unironically reliving the athlon days? is zen the k7 of our times?

>> No.60880211


>> No.60880222


Looks like it.

AMD performed a miracle.

>> No.60880240
File: 30 KB, 461x400, That Guy.jpg [View same] [iqdb] [saucenao] [google] [report]

>> No.60880246

Meanwhile in real life situations its always 30% slower than Intel

>> No.60880261


>> No.60880267

And then anon woke up.
Dream to nightmare must be a bitch.

>> No.60880282

>muh bideogames

Of course, maybe that is your only "real-life situation."

>> No.60880294

Anon, jewmelter clocks no differently from normal Ryzen. It's going to have stellar gaymen performance.

>> No.60880302

Literally another shoah. Now nvidia has to get CURRYED.

>> No.60880306


It also costs 60% less.

>> No.60880314


>> No.60880365

Fake until proven otherwise.

>> No.60880390
File: 48 KB, 600x329, TheJewsFearTheSamurai.jpg [View same] [iqdb] [saucenao] [google] [report]

>oy vey! another shoah! delid!

>> No.60880414
File: 21 KB, 300x300, 1487808235004.jpg [View same] [iqdb] [saucenao] [google] [report]

DELET THIS! How could 16x almost beat 22x!!! This is outrage!
Fake news!

>> No.60880417

Now we wait for Vega.

>> No.60880441

Are these theoretical or actual benchmarks for threadripper? Didn't think they'd actually get to 3.9ghz all core turbo

>> No.60880454

Ryzen (one Zeppelin die) consumes fuckall power. Threadripper is two Zeppelin dies. Do the math.

>> No.60880457

They didn't.

Advertised turbo is never on all cores, that's 1 core with unknown XFR(200-300MHz?)

>> No.60881062

You mean 30% faster maybe.
What most people don't get is that benchmarks in sterile systems favour intel with its slightly better single core performance. In a real world scenario with an OS fully loaded with secondary services that eat cpu the one who is better at multithreading wins. Even in "muh gayms"

>> No.60881070

Threadripper scales. They can get 3.9GHz on as many cores as they want.

>> No.60881102

3.9 is single core turbo.
3.7 is all core.

>> No.60881123

3.5 is likely to be all core turbo if they want 155W
3.7-3.8 would be all core if it was 180-200W

>> No.60881137
File: 116 KB, 1024x576, 1481758749774.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.60881153

This forum is worse than reddit. Go to school, kids.

>> No.60881171


>> No.60881180
File: 4 KB, 452x523, Mysterious Merchant.gif [View same] [iqdb] [saucenao] [google] [report]


>> No.60881239

Are you fucking serious, boi?

>> No.60881261

It's real. WTF.

>> No.60881263
File: 35 KB, 636x390, harediintel.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.60881356


>> No.60881449
File: 2.34 MB, 1920x800, 1465676427202.webm [View same] [iqdb] [saucenao] [google] [report]

>12x 3.60Ghz (4.1Ghz) HT

>> No.60881489
File: 975 KB, 852x480, 1496407733523.webm [View same] [iqdb] [saucenao] [google] [report]

>> No.60881518
File: 495 KB, 1070x601, 1494090152782.png [View same] [iqdb] [saucenao] [google] [report]


>> No.60881541
File: 14 KB, 659x382, rip (2).png [View same] [iqdb] [saucenao] [google] [report]

>> No.60881659

Almost double the 1800X

Lmao, how can Intel even compete? AMD , if worse comes to worse could price these at $600 and still get a hefty profit.
It's practically sodomization of Intel's entire server lineup besides the 22+ core parts.

>> No.60881702

The interesting question is how much a 1P epyc 32 core system will cost. In theory they should be able to sell 32 core cpus for about $1600 and still make a large profit.

>> No.60881771
File: 2.29 MB, 5098x1500, gc42aigjzh2z.png [View same] [iqdb] [saucenao] [google] [report]


>> No.60881783

so it almost shit on a 22 core intel for 1/3 of its price

>> No.60881787

intel said soldering chip fucks lifespan and stupid intelfangays belives that, like they going to use it fuckin 100 years.

>> No.60881820


>> No.60881833
File: 207 KB, 1280x590, sketch-1497359173805.png [View same] [iqdb] [saucenao] [google] [report]

>97,1% scaling
this kills the bingbus

>> No.60881857

Yeah, but they don't sell just chips, they sell support, longtime support(10+ years) validation and IHV/ISV certification with those.

So don't expect anything short of $6000

>> No.60881867

The xeon 2699 V4 is

The highes threadripper will probably not surpass $1200

Dothe math. It's AT LEAST 1/7 of the price. For what 97% the performance.

Intel is scared shitless with absolutely no response

>> No.60881921
File: 230 KB, 548x642, 1465519075030.png [View same] [iqdb] [saucenao] [google] [report]

well it seems the clockspeeds are different, averaging the 1800x(3.6/4.0) and the 1700x(3.4/3.8) results gives 1574.5 multicore score for a 8 core at 3.5/3.9, which makes threadripper have a 100.6% scaling.

what the actual fuck have they done

>> No.60882021
File: 46 KB, 500x300, tmp_23153-IMG_20170613_062346-948796873.jpg [View same] [iqdb] [saucenao] [google] [report]

Near perfect scaling and a happy fish

>> No.60882245


>> No.60882276

>real life



I wonder if you even thought about what you wrote kid

>> No.60882532
File: 470 KB, 1060x1569, TB2f4vIvrJmpuFjSZFwXXaE4VXa_2341477237_0.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.60882837
File: 162 KB, 633x900, 1489103005792.png [View same] [iqdb] [saucenao] [google] [report]

>Muh real life!

>> No.60882927

Sweet Jesus

>> No.60883080

which is better for gaming? intel or ryzen?

>> No.60883137

And your monitor refresh rate is?

>> No.60883143

7700K = Max performance at 1080p 144hz
1600/1600X = Max performance per dollar
1700/1700X = Lets you game while streaming

>> No.60883180


>> No.60883185

a GPU.

>> No.60883193

i don't care at all about streaming. what could make me play gta5 at +60fps?

intel cpus have a gpu too.

>> No.60883196

And your resolution is? Anyway, get 7700k of you want ~120fps here and now.

>> No.60883210

1680x1050 or 1650x1080 (can't remember)

>> No.60883222

Looks like it
K17 is the new K7

>> No.60883244

>gta5 at +60fps

Any of them

>> No.60883294

Much better frametimes

>> No.60883295

cool, thank you all for helping.

>> No.60883346

>intel cpus have a gpu too.
not x299 once, don't.

>> No.60883381
File: 6 KB, 271x186, download.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.60883400


>> No.60883431
File: 244 KB, 350x501, oyvey.gif [View same] [iqdb] [saucenao] [google] [report]


>> No.60883437


>> No.60883448


>> No.60883577

Contract Keller the Jew Killer

>> No.60883600

>slaughtering a server CPU with 4 more cores
>very close to CPU with 6 more cores
>basically perfect scaling across cores
>costs less

Jesus fucking Christ. How is this fucking possible? Its him isn't it, it must be. There is no other excuse. I don't even think Jim Keller is human.

>> No.60883634
File: 48 KB, 676x426, 1493490396015.jpg [View same] [iqdb] [saucenao] [google] [report]

It seems like AMD went from joke to total domination.

What went so right?

>> No.60883663

They hired the right people to do the R&D. I wonder whether Pajeet will deliver another R300.

>> No.60883833

Is that why Intel over heats?

>> No.60883900
File: 31 KB, 630x372, toothpaste inside.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.60883931

this Raja pooinloo has to fucking deliver Vega to wreck nvidia's shit.
Otherwise I'll be furious since I'm still running a 6870 and don't want to go team green.

>> No.60883995


>> No.60884025


>> No.60884144


>> No.60884310

it's not that AMD performed some miracle.
They just did the job - designed a proper CPU which performs exceptionally well by today's standards.
You can now truly see that for all this time, when AMD wasn't competitive enough, intel was just shrinking the transistors without making any advances in the design, releasing the same shit with barely anything changed for several years. They were just holding everything back.
Hopefully not any longer.

>> No.60884372

Wait for 2020/2021 lol.

>> No.60884484

He's a big guy

>> No.60884580

Considering his track record, he probably will.

>> No.60884697

don't, the 770 pain is still with me

>> No.60884782

>1800x beating 15 core xeon
Top kekkels

>> No.60885575


It's more than that.

AMD has no money. They are a fraction of the size of Intel. Oh by the way, they are fighting another juggernaut in Nvidia at the same time.

They should not have been able to do this. Silicon design is extremely fucking hard.

>> No.60885622

AMD is nowhere even close to being on the same level as NVIDIA. Get real.

The only reason AMD did so well is because it realized that giving Jim Keller a fat paycheck for even a short amount of time would allow it to become king again. Look at what happened to Apple when Jim Keller worked for that company.

>> No.60885623

>14 cores
obvious fake

>> No.60885649

Are you stupid? It comes close, but that's not beating.

It's very impressive, though.

>> No.60885681

Too bad Zen was designed by Mike Clark.

>> No.60885742

Jim, you need to calm down, we just wanted competition, not murder charges.

>> No.60885941

the more cores they add, the better performance....WTF!!!!!!!!

>> No.60885998

people are really confused by this because intel doesn't work like that, they are used to really shitty scaling with core count, and negative scaling on quad socket solutions

>> No.60886009
File: 400 KB, 816x1020, Jim Keller.png [View same] [iqdb] [saucenao] [google] [report]


"When there's something's Jewish
In the neighborhood,
Who're you gonna call?"

- Jim Keller, C.S.W.

>> No.60886084
File: 15 KB, 480x270, painthuff.jpg [View same] [iqdb] [saucenao] [google] [report]

>completely ignoring single thread performance needed for megatasking.

>> No.60886085
File: 1.33 MB, 400x225, 2bqUm9L.gif [View same] [iqdb] [saucenao] [google] [report]

I've found most retards don't even understand how poorly Intel's ringbus scales and assume that they're going to get i7-7700K clockspeeds on their 16 cores.

They also have no clue how Zen works and assume it's another Bulldozer-tier housefire. Oh, the ironing.

>> No.60886119
File: 1.16 MB, 1293x1730, 1497060809496.jpg [View same] [iqdb] [saucenao] [google] [report]

>unironically using megatasking
>not realizing nobody gives a fuck about 4.1 GHz vs 4.5 GHz single core boost speeds on HEDT
>being a useless perma-BTFO tripfag that should kill itself immediately

>> No.60886127

Mike took that position after Keller left. Keller is who started the design (not hands on)

>> No.60886139

No, Clark was the one who designed Zen. Keller was behind K12 and a bag of power saving IPs.

>> No.60886170

thank mr shitwrecker

>> No.60886172

This is the same website that posted fake scores before Ryzen 7 launch.

>> No.60886184

their compass is literally:
intel = good
amd = bad

what'd you expect from someone that thinks like that?

>> No.60886226

>what'd you expect
Almost two decades of AMD getting wrecked.

>> No.60886244
File: 28 KB, 550x550, single socket holocaust.jpg [View same] [iqdb] [saucenao] [google] [report]

lmao. Kill yourself.

>> No.60886264
File: 59 KB, 548x562, 1443364338656.png [View same] [iqdb] [saucenao] [google] [report]


>> No.60886270

fuck off monkey, eat a dick

>> No.60886284

The cinebench scores for the TR CPUs are definitely simulated, but from what we know about how IF works and how Zen scales, it seems accurate.

The computex demonstration showed it running with almost 1:1 scaling with an 1800X, which usually gets 24 seconds on this blender test. https://youtu.be/NeuamYdfU4w?t=16m

>> No.60886449

has any website reverse engineered the ryzen? i want to know how they implemented hyper threading without infringing any patent.

>> No.60886478

SMT existed long before intel anon, it's not something they invented

>> No.60886484

Hyperthreading is just Intel's implementation of SMT. They don't "own" SMT. Ryzen's SMT is actually better.

>> No.60886515
File: 411 KB, 3264x1952, sad.jpg [View same] [iqdb] [saucenao] [google] [report]

i still want to know

>> No.60886539

reverse engineering processors is one of the hardest things to do, even with the proper budget it would take more than 6 months easily

>> No.60886544
File: 123 KB, 240x286, lisa su smug.png [View same] [iqdb] [saucenao] [google] [report]

Ancient Chinese secret.

>> No.60886573

i know. there is a enterprise which does that but it cost something like 250 000 USA dollars for the intel i7

hope some black hat will someday steal and leak the source code of a modern cpu like i7, rizen, or ibm power

>> No.60886599


All the patents for the technology expired in the 80s. This is really, really old tech by computer standards.

And it can take years before anyone actually reverse engineers Ryzen and releases that infromation it to the public.

>> No.60886614

for what end? it's just bullshit assembly that has no use without the rest of the processor

>> No.60886660

Literally who cares about kikeripper and skyshit-X? GloFo just announced 7nm mass availability in 2H 2018 with low volume in 1H 2018

These 7nm Zen chips will have no issues getting clocks over 4.5GHz and have a IPC increase to boot, if you desperately need to upgrade, do it now.
Otherwise wait for the real coming of K8

>> No.60886682

you can buy threadripper now and then upgrade for the 7nm version when it comes out next year

>> No.60886701

>it's just bullshit assembly that has no use without the rest of the processor
to see what algorithms they are using. it's more high level than assembly, actually.


intel is now using verilog


isn't ryzen the first amd cpu with true smt? (because the original patent has expired)

>> No.60886897

>Ancient Chinese Art
She must know some Kung Fu fighting

>> No.60886954

intel actually holds a lot of interesting patents on smt. hyper threading seems to be an automation of time sharing multi threading based on the instruction flow. for example, intel hold a patent on how to switch thread based on jump instructions.

>> No.60887049

no intel has only certain patents regarding hyper threading and hyperthreading alone they cant patent anything related to smt

>> No.60887072

Get a 1600 and then buy a good GPU (1070, 1080, etc)
Those extra threads on the Ryzen CPU will be beneficial for multitasking and games down the road as they get proper multithreading support.

>> No.60887074

Jim worked on k12 and infinity fabric, not ryzen

>> No.60887118

he worked on ryzen via the low power IP's

>> No.60887165


>> No.60887739

>you can waste money now instead of just waiting a year and spending it then
>you don't want to pay double the costs for no reason?

>> No.60887906

>infinity fabric
And that's what allows these processors to scale so fucking great.

>> No.60887976

Threadripper is actually a "viral" campaign to attempt to "rip" apart people's views on thread performance.

It only works on retards :3

>> No.60887987

Look everyone, it's the court jester.

>> No.60888018
File: 19 KB, 682x380, untitled-1.png [View same] [iqdb] [saucenao] [google] [report]

>muh gayming

Use a bit of the cash you save on a cheaper CPU and mobo on better ram

There your gaymen run just as fast now

>> No.60888050

So THIS is the power of Intel(r)(tm)(c).

>> No.60888170

4000 Mhz is now avaiable anon for ryzen on new bios.

>> No.60888304

Kill yourself, tripfag.

>> No.60888780

Has anyone seen any vmark benchmarks for this? I don't give a fuck about this for HEDT or whatever cinebench or whatever gay shit people are into, I'm interested in using these in my cluster to replace my older dell R630s

>> No.60890528

>Making multiple threads on the CPU core pipeline.

Windows and Linux do this every day.

>> No.60890552
File: 88 KB, 650x762, 1493230895604.jpg [View same] [iqdb] [saucenao] [google] [report]

>Meanwhile in real life situations
what did he mean by this?

>> No.60890613


>> No.60890698
File: 5 KB, 210x230, 1493572397537.jpg [View same] [iqdb] [saucenao] [google] [report]

Is it even possible to reverse engineer 14nm silicon with billions of transistors? Like how would you even do that?

>> No.60890712

>mfw only America can buy 99 dollar dildos without worrying about eating afterwards
Suck it rest of the world.

>> No.60890740

you open the processor.

then you look at it.

>> No.60890762

I wonder if we'll ever get more than 2 threads per core. When Hyperthreading was introduced there was spoken of HT2 which would increase threads to 3 per core.

>> No.60890782
File: 433 KB, 360x408, 1496324077672.png [View same] [iqdb] [saucenao] [google] [report]

I just can't imagine the ungodly abominations of dark arts hidden inside the nodes, aren't there about 12 interconnect layers above the transistors?

>> No.60890829

It's not humanly feasible to reverse engineer this shit.
Design has been done by software for decades now

>> No.60890855

>ITT Intel shills and AMD shareholders

>> No.60890929


>> No.60890932

You need more resources in the core for that, power8 for example has 64kb of l1 data cache to support 8 threads per core.

>> No.60891244


>> No.60892503

>Threadripper is actually a "viral" campaign to attempt to "rip" apart people's views on thread performance.
>Almost two decades of AMD getting wrecked.
>completely ignoring single thread performance needed for megatasking
kys eternal tripfag See >>60886119

Name (leave empty)
Comment (leave empty)
Password [?]Password used for file deletion.