[ 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 ]
>> No.72961797

Explain it to us, we'll understand.

>> No.72961872

the day amd replaces intel in macbooks and imacs is when they'll be in trouble, until then you're a retarded fanboy

>> No.72961898
File: 61 KB, 550x365, 1501554667405.jpg [View same] [iqdb] [saucenao] [google] [report]

intel BTFO

>> No.72961925

like for example the 21241415 security leaks that intel has and AMD not?

>> No.72962112

sirs pls let us be reasonable, Intel still has the better chip yes?

>> No.72962465

lol intels not finished. Theyre just going to play dirty like they always do when theyre behind

>> No.72962587
File: 751 KB, 1076x1050, 20181216_030750.jpg [View same] [iqdb] [saucenao] [google] [report]

AMD has fastest and best processors in the world

>> No.72962598
File: 58 KB, 856x587, AMD-EPYC-7002-OpenSSL-Verify-Benchmarks.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72962744
File: 400 KB, 850x939, 1569763836793.jpg [View same] [iqdb] [saucenao] [google] [report]

0.1 shekel has been deposited into your account.

>> No.72963115

There is a difference between theoretical address space (16EB) and current implementations. Since we do not have the capabilities to even make so much RAM we settled on implementing lower number of bits. Both software and hardware are limited far below 64-bits - only recently Intel is pushing for five layer paging in Linux that will bring us up to 57-bits on software side.
Hardware is even "worse" and usualy does not go beyond 48-bit. Only extreme cases will go above.

>> No.72963247

>tell me how Intel is not finished
are you buying anyone of them soon?

>> No.72963260

>256mb cache
WTF? XP can run on 64mb RAM.

>> No.72963281

>More power usage is better
>In any way, shape or form, any month of any year, at all.
Are you retar

>> No.72963482


>> No.72963489

stock market is joke

>> No.72963543

Since AMD started (and Intel copied) integrating the memory controller into the CPU the chipset doesn't matter. As a fact EPYC Rome doesn't even need one - it's all integrated.
The motherboard matters only in terms of signal integrity, which is passive.
The real limit is only the IMC in the CPU.

>> No.72963636

I knew Zen was gonna rock.
Glad I skipped 2 though it's humdrum
T. 2700x fag
Zen 4 5 6 is gonna be insane 3d stacking hbm on die crazy shit.

>> No.72963857

Are u dumb anon? You really think binning is the reason? What retardation gas have you been inhaling, your own farts?

>> No.72963860

AMD has implemented a great strategy - create one small piece of silicone that can be used in every segment. This allows them to manufacture a shit ton of them cheaply, bin appropriately and sell with huge profit.
The previous generation had NUMA problems that made it tricky to use the workstation/server processors effectively. Zen 2 moved the IMCs to a separate IO hub (which is 12 or 14nm - extremely cheap considering the size) which together with improved Infinity Fabric allowed them to increase performance and remove the NUMA problem almost entirely.
Meanwhile Intel has a huge range of different die sizes for different segments. Just Skylake-X comes with 3 physical variants: LCC, HCC and XCC. They have to divide the manufacturing capacity between all those different things. Huge dies also have higher defect chances than small dies, and Intel only makes huge dies in the high-end.
Intel has also stalled their architectural advancements. We should've had Ice Lake a long time ago, but the 10nm fiasco and their stubbornness in not backporting the design to 14nm resulted in a huge failure. Instead we'll get Comet Lake AND Rocket Lake on 14nm before we even see Ice Lake in the desktop segment.
What is more Intel has had huge security problems and mitigations are having a disastrous performance costs.

>> No.72963922

Just get a Ryzen 3700X/3900X/3950X and ASRock Rack X470D4U with 128GB ECC RAM.
I don't really see a reason to even bother with the likes of Xeon Bronze/Silver at this point - you don't get the big Xeon features on them anyway.

>> No.72963959

absolute state of intel cope

>> No.72963986

shes dead anon

>> No.72964244

You're right, my power is bundled at practically a steal so it's not part of my comparison at all. Running some numbers it looks like the power difference would pay off pretty quickly. Thanks for the insight you corporate demon niggas

>> No.72964288

>As a retard,...

>> No.72966921

They are binned for power consumption, nobody cares about boost clocks in server chips, it's going to be too hot to run at high clocks

>> No.72967268

Why don't they just put gigs of cache on a cpu?

>> No.72967878

>IBM wojak
Well that's "new"

>> No.72968559



I'm not the troubled by my i9 since I have custom water cooling with one 480 and one 360 58MM thick radiators.

But that EPYC is clearly walking over the Xeon without noticing it stepped on shit.

>> No.72969603

Who would've guessed that being a mindless consumer drone required less powerful hardware?

>> No.72969624

Too expensive.

>> No.72970263

Intel reserve their best binned chips for Xeon.

>> No.72970279

>They are binned for power consumption, nobody cares about boost clocks in server chips
That's basically the same thing. Better power consumption means less heat losses which means better clocks for the same power consumption.

>> No.72970538

Apple isn't exactly a large customer for Intel m8. Servers is where it's at.

>> No.72970682


>> No.72970706

yea and? Even their "best binned" chips still have far less L3 cache, less memory support, less memory channels, and as a shit shot of PCIe lanes compared to AMD.

>> No.72971064

>you vs the L3 cache she tells you not to worry about

>> No.72971202

>I hope the best for ARM
ARM is already king anon, Android and iOS runs on ARM, and those are the most used operating systems out there. It's only a matter of time for ARM to become mainstream on professional/gaming workloads as well, and that's when x86 finally dies.

>> No.72971203
File: 41 KB, 1024x677, 1561698284076.jpg [View same] [iqdb] [saucenao] [google] [report]

>AMD managed to stay low for almost 10 years
>intel can't
Dumb cunt

>> No.72971330

So is GCP: https://cloud.google.com/blog/products/compute/amd-epyc-processors-come-to-google-and-to-google-cloud
I'm just waiting for them to be available so I can move all of my companies servers to epyc cpus

>> No.72971413

I loved how when EPYC was announced Intel and people were all like "oh the ecosystems are all built on Intel so the big companies won't buy into AMD because it will cost more than just the cost of the hardware to transition ecosystems".
Then AMD came out and absolutely DESTROYED Intel in ecosystem offerings as well and now the turns have tabled.

>> No.72971439

w-who needs more cores anyway, current intel performance is fine.
why upgrade what already works?

>> No.72971981

absolute staate of /g/

>> No.72972089

>and now the turns have tabled.

>> No.72972773

But 3950x will be binned on maximum clocks stable which may be slightly different

>> No.72972800

They have different dies for their chips, while AMD can put a chiplet anywhere

>> No.72972918

>not Watts(higher is better)

>> No.72974052

Ignore him, he's probably the guy who was claiming that two 9900Ks would out perform the 64 core EPYC.

>> No.72974337

in short: you are an antisemyte

>> No.72974489

This is your brain on Intel

>> No.72974641

Isn't that like 1KW of power consumption?

>> No.72974706

implying the 7th-9th gen sold nothing.
fucking lol.

>> No.72974880

Boosts are fixed with the ABBA Dancing Queen BIOS.
Lower clock speed when comparing different architectures doesn't matter, just ask Intel how their Pentium 4 went.

>> No.72974894


Arm server chips have been tried and the performance/watt honestly ends up not being much better than intel and amd's x86 offerings furthermore the performance/price ratio is much worse.

>> No.72974896

That's only sold by Intel in a nuclear reactor case with mandatory liquid cooling (intel.com/content/www/us/en/products/servers/server-chassis-systems/server-board-s9200wk-systems/s9256wk1hlc.html). Nobody will buy that.

>> No.72974998

Well you probably should try downgrading your BIOS then. is designed for Ryzen 3000.
Also PBO+XFR works great on 2000 while it doesn't do anything on 3000 since the chip is maxed out from the factory.

>> No.72975076

>Apple are still relevant
Didn't Google just switch all their servers over to amd?

>> No.72975084


>> No.72975380

They are at least a year behind Samsung on technology.
Only useful in a very small niche, otherwise prohibitively expensive compared to SSDs.
Dead like everything else they bought and mismanaged.
They're behind everyone else.
In fact all recent Intel networking hardware has been either a year late or did not come out at all.
If they don't nail their next CPU arch they're done for in long term because they don't have any other business where they don't suck.
>absolute staate of your retarded ass

>> No.72975453

delid this kind sir

>> No.72975527

Hm, in that case I'd reset note current settings, reset BIOS to defaults (unless you did that when you originally updated) and reconfigure. Some motherboard vendors are retarded and make Cool'n'quiet DISABLED while on Auto, enable that instead.
Remove AMD chipset drivers, reboot and reinstall them just to be as well.
Boost problems on 2000 are quite rare in comparison to 3000 so it's probably fixable.

>> No.72975539

Microsoft, Amazon, Dell, Google etc.
Don't listen to that autistic retard.


>> No.72975643

Performance doesn't matter
Efficiency doesn't matter

>> No.72975654

Imagine 4tb RAM... That makes my peepee unflacid

>> No.72975700

Let's not forget about Puma6 :^)

>> No.72975740

The more data you can keep in RAM the less time you spend waiting on IO. In-memory databases are at thing too.

>> No.72976344
File: 20 KB, 466x466, 611SiUgEl2L._SX466_.jpg [View same] [iqdb] [saucenao] [google] [report]

AMD is shit. Our office went AMD only and it's like 5 degrees colder now. Had to go get a literal space heater to keep us warm.

>> No.72976393

8280 is clearly larger than 7742 meaning more better, anon.

>> No.72976562

>not openly advertising half of your products isnt being a fucking cuck

>> No.72976576

>by the end of the year
I guess they still have time.

>> No.72976756

The Virgin Xeon Vs. The Chad Epyc (dot) jpg

>> No.72976794

I think we are focusing to much on the raw numbers. Raw numbers sure show EPIC beating out intel. However running applications is more than just numbers. You want an experience, and only intel has that offering for real world applications.

>> No.72976808

Being a brand fan is the highest level of human stupidity beyond being a sports team fan

>> No.72976820

I'd like a 6-8 core one for a small home server. Seems ideal to be honest.

>> No.72977000

It is ideal, just get some nice 2666MHz ECC Samsung RAM to go along with it.

>> No.72977022
File: 135 KB, 1148x1892, avx512.png [View same] [iqdb] [saucenao] [google] [report]

Here's your real world usage. Even with AVX512 secret sauce they just barely match EPYC at roughly TWICE the power usage. They get utterly destroyed in any integer load. Both pieces of software on this are developed by Intel themselves.

>> No.72978601

Yeah, let's talk watercooling in servers.

>> No.72981015

>Apple is developing their own custom ARM chips to ditch Apple.
Your average AMD indian shitstain right here.

>> No.72981243

Jim Keller, the guy who designed Zen (as well as a ton of other important architectures, is now working for Intel.

Give them a couple years and they'll catch up.

>> No.72981316
File: 10 KB, 241x209, images.png [View same] [iqdb] [saucenao] [google] [report]

>amd is son of a goatbitch

>> No.72981331

can't run a simple game.

>> No.72981639

But the four of them required to actually beat this EPYC processor WOULD.
And if you start going "clock for clock it's faster" or "it's better in single core performance and that's all that matters!" then your 9900k gets slain by an i3 at half the wattage, so fuck off.

>> No.72981815

AMD will do good for 2-3 years.
Then Intel will catch up
Release an amazing chip and annihilate AMD and we will be back to the status quo.
Wash and repeat every 10 years or so.
This has happened like 4 times already.
For those young folk here who weren't born yet.

>> No.72982140
File: 84 KB, 653x726, 1491958654249.png [View same] [iqdb] [saucenao] [google] [report]

>AMD will do good for 2-3 years.
>Then Intel will catch up
>Release an amazing chip and annihilate AMD and we will be back to the status quo.
>Wash and repeat every 10 years or so.
>This has happened like 4 times already.
>For those young folk here who weren't born yet.

>> No.72982620

>I hate this shitty meme
They are just that anon, shitty memes. This Laosian basket weaving forum is full of them. Don't upsetti your spaghetti.

>> No.72984005

people have been shitposting about EPYC for years but I've never seen anyone talking about any actual systems shipping with them, you'd think the marketers would pause to thank their customer base at least...

>> No.72984010

Wait. The Intel one is $10,009 for 1k units. That works out to $10.01 per unit, which is really very cheap.

>> No.72985307

Honestly i dont think so, AMD really has intel fucked.
Look at all of intels road maps, including leaked ones. They have nothing in response to the new chips. By the time they even get to 7nm AMD will likely be on 5nm with a newer revision to ryzen.

I honestly think we are entering another golden age of cpu performance, they will continue to leapfrog each other every 2 or so years imo.

>> No.72985469

In case you're not trolling, that's $10,000 per unit if you buy 1k units.

>> No.72987450
File: 1.25 MB, 1002x1480, 1492962036185.png [View same] [iqdb] [saucenao] [google] [report]

its even worse than that, the server Icelake chips that are in testing right now have less cores then Skylake did

>> No.72988550

Intels can usually manually OC high on all cores, the built in boost only does a few cores at a time I think.

>> No.72988728

the minimum sku count you can get prices on are 1000 for xeons iirc.

>> No.72988778

There's a story about the very first ARM prototypes. They'd arrive from the fabbers, get connected to some testing setup, and just run without being connected to a power supply.

>> No.72988820

built for sneed

>> No.72989056

Depends on your budget, but as long as you avoid Intel pozzed housefire trash you're good to go.

>> No.72989064

>i do a lot of cgi shit like rendering and simulations. what's the best value cpu
You want GPU power for that shit, son.

>> No.72989072

okay, what's a good gpu for that shit then? radeon vii? 2080ti?

>> No.72989107

it's hard to say. there's opencl acceleration on the houdini side for a lot of stuff but with maya there's not much info on it. it seems to be mostly cpu based for bifrost and so forth. for rendering i think i'm screwed though since there aren't any opencl renderers supported for houdini. i can use rpr on maya though. i'm hoping karma will change that. i was planning to maybe get the 3rd gen threadripper with a radeon vii for a primarily houdini based build but i don't know how that would perform in retrospect to just an i9 with 2080ti since a lot of benchmarks point to that being better on a lot of maya tasks? due to single threading or some shit i'm not sure, though there's no way bifrost and ncloth are not multi-threaded fuck that noise.

>> No.72989130

For the actual workload of rendering images? Moar Coars > *.
Nobody does single threaded rendering anymore, this isn't the 90s and Lightwave on the Amiga.

>> No.72989178

yeah that's what i figure. would a ryzen 16 core be sufficient at balancing between multicore and singlecore performance compared to threadripper? i doubt a lot of tasks could use more than 32 threads, although i would like to push simulations harder for things like oceans. that would be interesting.

>> No.72989457

>i do own a 3900x

its like an excuse for intel shills to justify their stupidity

>> No.72990236

The AMD chip didn't exist as an option and still doesn't as an option for System builders like Dell and HP enterprise, ect.
I'm saying people who already bought Intel are not going to care that 3 months later there's now a faster option for cheaper.

>> No.72990348

>still doesn't
Bruh. Go watch Horizon prez. Dell, HPe and Lenovo people literally came onstage to present their new, awesome Rome-based systems as READY FOR SHIPMENT.

People who already bought Intel are stuck with their electricity bills and performance-devouring mitigations until the end of their refresh cycle, pulling out their ass hair in impotent envy at those who could afford to wait. The companies with 3-4-5 year old infra that's coming up on refresh cycle right now, or companies who are scaling out for current or expected growth need only to look at flops/$ and flops/W numbers to make the right choice.

The IT managers and contractors building out all-new infra today or next year are the ones who'll benefit the most from Zen 2 release, as more and more systems are coming to market right now, from boutique shops like Cray and bargain bin vendors like AsRock and everything inbetween. They can just take the budget they negotiated around Xeons and get 2-4x perf in the same rackspace and power envelope, or they can slash the rackspace in half and lay aside half of projected power and cooling costs into future upgrades budget. Or they can just get equivalent flops for half the money and pocket the rest via "rebates", like some corporate assholes often do, and it still won't compromise the project.

>> No.72990387

>pulling out their ass hair in impotent envy at those who could afford to wait.
Enterprise customers don't fucking care.

>> No.72990516

Because they have the same number of cores and are cheaper than the workstation counterparts, leaving me more options for multi card setups. The hbm is one thing I value for caching and there isn't much that offers it anyway. The 2080ti beats a few Quadros actually in some render benchmarks, only losing to the RTX 8000 which is over five grand.

>> No.72990821

Stop kvetching goy

>> No.72990969


>> No.72991035
File: 12 KB, 528x505, civ VI.png [View same] [iqdb] [saucenao] [google] [report]

Nothing matters no one else

>> No.72991058
File: 121 KB, 1016x613, re2.jpg [View same] [iqdb] [saucenao] [google] [report]

Simply nothing more to give

>> No.72991079
File: 39 KB, 650x450, 111263.png [View same] [iqdb] [saucenao] [google] [report]

There is nothing more for me

>> No.72991089
File: 44 KB, 650x450, 111287.png [View same] [iqdb] [saucenao] [google] [report]

Need the end to set me free

>> No.72991102
File: 59 KB, 1019x569, WWZ.jpg [View same] [iqdb] [saucenao] [google] [report]

No one but me can save myself, but its too late

>> No.72991114
File: 64 KB, 1015x578, PUBG.jpg [View same] [iqdb] [saucenao] [google] [report]

Now I can't think, think why I should even try

>> No.72991153
File: 213 KB, 1920x1080, 1562534167877.jpg [View same] [iqdb] [saucenao] [google] [report]

Death greets me warm, now I will just say goodbye

>> No.72991590

based amd redpilled

>> No.72991610
File: 319 KB, 632x970, Capture.png [View same] [iqdb] [saucenao] [google] [report]

Left is actually a "8+8+8+8+8+8+8+8" core processor whereas Right is a true 28 core

>> No.72991627

>salty much?
holy BASED

>> No.72991634
File: 134 KB, 785x1000, DD6D1632DC4C434DB0651598E09548B1.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72991661

Because is server hardware, not for the masses you retard

>> No.72991675

The jews asshurt is so delicious

>> No.72991828

>the absolute state of shitkikes

>> No.72991847
File: 28 KB, 200x300, 1535463927715.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72991903

>Servers don't matter
I would say go back to >>>/v/, but even there would call you a retard for this. So consider suicide

>> No.72991937

Already on the list.

>> No.72991964
File: 42 KB, 426x510, 1558582009289.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72992001

Fuck you goy

>> No.72992027
File: 225 KB, 1800x850, ehjRyTG.png [View same] [iqdb] [saucenao] [google] [report]

Intelbros...we're not shilling hard enough...

>> No.72992099
File: 70 KB, 500x487, your-tears-are-saltier-than-these-delicious-crisps-my-dear-941766.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72992149

You will pay for this you stupid goy

>> No.72992184
File: 47 KB, 636x420, 1313387093113269314.jpg [View same] [iqdb] [saucenao] [google] [report]

I only want to see the world burn

>> No.72992210

>he eats salt and vinegar

>> No.72992212


>> No.72992237

Cache doesn't matter

>> No.72992253

Just perfect for my salad

>> No.72992284
File: 37 KB, 474x415, 1539296460877.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72992430
File: 10 KB, 270x190, 1569564692290.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72992484

Just because you couldn't make it work back in the day, doesn't mean it's an inherently shit solution, Intel.

>> No.72992673

oy vey bad goyim

>> No.72993006

KEK I'm not even mad for "wasting money" buying a 1800x at release.
At least i can sleep well knowing i didn't get jewed and my money helped the industry

>> No.72993046

1800x is still pretty capable even today, 8 cores are optimal actually even for games.

>> No.72993064

Can't say the same about quad corelets

>> No.72993083
File: 507 KB, 578x644, amdramadan.png [View same] [iqdb] [saucenao] [google] [report]

>At least i can sleep well knowing i didn't get jewed and my money helped the industry

I can guarantee you didn't.

>> No.72993119
File: 232 KB, 605x1466, 1540319177583.jpg [View same] [iqdb] [saucenao] [google] [report]

>the official Intel roadmap doesn't even have anything competitive in it in the future either, unless I've missed something.
They really don't, not even ice lake looks that mindblowing. The only thing that looks interesting is sunny cove shit but that's for 2022

>> No.72993137

You are confusing sikhs with shiah muslims.

>> No.72993181

Naw they just think 10 years out and when they are wrong it takes them a year or two to pivot.

Hammer is coming down in 2020 lol.

>> No.72993186


Intel will be kept alive just so that AMD doesn't get in trouble for monopoly reasons, but their fabs are pretty much finished.

>> No.72993232

Whatever you say, pajeet

>> No.72993419

Cute pussies

>> No.72993432


>> No.72993784

what ever,no different for old ryzen.
and more and more intels cpus going win,wait end of year.

tdp is just that high what is should be with more cores or clocks...none upgrade.!!!

looks same effect ike gpu sector, fury,vega and radeon vii...no different
well radeon vii was 7nm gpu, but not belive it! so lausy efficiency,and still slow!

ok, cputdp also same...there is anyhting new,just different name....i wonder how long amd can sell product with no profit..well looks that no longer...sooner or later its eat hole company and debt givens not tolerance more...end of 2020 is my guess.

hmm. i try really see different from ryzen 2000 series,but no....oh yeah... more cores..but is it only trend or commercial way.
you dont need more than total max 6 cores for gaming,even 4-core is enough. easily.
more than that is waist of cash, no matter how much its cost.

and,i still see not any goodies from amd 32-core cpu...erhh i mean where to use thouse??!! tell me..

>> No.72993885

>more cores useless
nothing you said makes sense if ryzen is beating the i9 9900k in games when it's supposed to be the highest single threaded performing chip out there.

>> No.72994036

RAM doesn't matter

>> No.72994231

>Intel is pretty much dead. They will become a memory


>> No.72995171

stop moving the goalpost, at first it was servers and now it's laptops what's next Pentiums?

>> No.72995513

This but unironically.

>> No.72996066
File: 177 KB, 1200x900, DhO7WXhWkAA3OQV.jpg [View same] [iqdb] [saucenao] [google] [report]

Both of them will bring AMD's end.

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