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

If you can see this message, the SSL certificate expiration has been fixed.
Become a Patron!

/g/ - Technology

View post   

[ Toggle deleted replies ]
File: 410 KB, 576x181, hYjbOv9.png [View same] [iqdb] [saucenao] [google] [report]
72227570 No.72227570 [Reply] [Original] [archived.moe] [rbt]


>one (ONE) Zion Golturd 6148 is 3500$
>TWO EBYN ROME 7452s are ~6000$

>> No.72227605

Holy fucking shit

>> No.72227612

>>>/v/ tranny

>> No.72227613


>> No.72227620

Epyc post

>> No.72227626

Intel is a value leader in delivering more seconds per socket.

>> No.72227638

Lower is better/faster, Intbecile. Cope harder.

>> No.72227645

its over for intel

>> No.72227648

Are you retarded?

>> No.72227658

It's time to completion, not FPS, Intbecile.

>> No.72227672

I'm not even one of the two, but you seem pretty retarded, those comments were making fun of intel you absolute troglodyte, go shit on a street pajeet.

>> No.72227685
File: 109 KB, 1085x948, bakaren.jpg [View same] [iqdb] [saucenao] [google] [report]

Man are these intel security bios/windows mitigations THAT bad? What makes them have such crippling performance regressions?

>> No.72227691
File: 251 KB, 868x756, dafe2991f27055fa3391a14d28f7165cec22373960b4ad7f2e90b2942ca3a8bd.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72227697

Higher is better, right?

>> No.72227701

what the fuck

>> No.72227706

...the test in OP pic is without ANY "mitigations". Not even Spectre/Meltdown. It's a "clean" Inturd poocessor there.

>> No.72227714

Too bad microsoft/google/apple and every other multi billionaire company are built around intel and it's practically impossible for them to change all their data servers.

>> No.72227715

>unit: Seconds
In other words AMD completely trashed Intel.

>> No.72227756
File: 10 KB, 633x772, dead.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72227768

Sure thing buddy.

>> No.72227784

are you kidding? people are moving to amd FAST


intel couldn't even get board vendors for the 9282 because there was so little interest

>> No.72227810
File: 202 KB, 737x664, 1558219823433.png [View same] [iqdb] [saucenao] [google] [report]





>> No.72227818
File: 603 KB, 320x240, 1541133504601.gif [View same] [iqdb] [saucenao] [google] [report]

>"""""400W TDP"""""

>> No.72227822

when they
>cost less
>perform faster
>have a fraction of the power draw
the benefits add up really fucking fast

>> No.72227836
File: 96 KB, 721x1041, 57180482139470938217403219874.png [View same] [iqdb] [saucenao] [google] [report]

>> No.72227862

how is the tdp so much lower?

>> No.72227863

>two (TWO) EBYN 7542s barely consume ~450W (and that's only under extremely stressful long-term constant full loads) in a 2S system

>> No.72227917

what is with that font rendering, friend?

>> No.72227943

time doesn't matter!

>> No.72227949


>> No.72227998

does Intel still recommend disabling hyperthreading? if so Epyc is going to absolutely dominate

>> No.72228010

for server use intel official recommends disabling hyperthreading, but with a nudge and a wink. basically they're saying "if there's a huge data breach because our processors have some horrible security flaw, we told you to not use that feature lol"

>> No.72228060

delid dis sir thank

>> No.72228081
File: 54 KB, 943x537, 1559071129317.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72228111


>> No.72228129
File: 131 KB, 720x676, 1558404979342.jpg [View same] [iqdb] [saucenao] [google] [report]

>it's STILL live

>> No.72228151


>> No.72228178

dumb butai

>> No.72228188

With how pozzed hot garbage Intel is these days even bulldozer would probably be better.

>> No.72228236

Fine wine.

>> No.72228246

Can someone explain OPs image to me? I am not that into meme wars

>> No.72228275

new amd server processors >>72227836 announced and they shit all over intel hard

>> No.72228290

What the fuck do I need 64 cores for? Dwarf fortress is a sequentially bounded process.

>> No.72228357

Zen3 design complete.
Zen4 in design phase.

Get ready for 7nm EUV Zen3 in 2020.

>> No.72228462

They're out, dipshit. It's a release, not announce.

>> No.72228472

oh ok. where can I go have one in my hands today?

>> No.72228487

>implying you got 4-7K to drop on a cpu

>> No.72228496

Wait for the end of the conference or tomorrow, I guess.

>> No.72228502

you said they're out. where?

>> No.72228509

>I don't know how enterprise sales work

>> No.72228519

so they're not out.

>> No.72228525

Have your sales rep place an order for an HP enterprise machine

>> No.72228542

oh god no
this is moving too fast

>> No.72228565

You can't stop success.
7nm EUV Zen3 2020
5nm EUV Zen4 2021
Mark Papermaster confirmed that Zen5 is in design as well

>> No.72228578
File: 675 KB, 1126x608, 94333363.png [View same] [iqdb] [saucenao] [google] [report]

>> No.72228602

>per-socket licensing (rhel, vsphere, whatever) putting intel on suicide watch: live

>> No.72228673

>we may see 32/64c 1st gen EPYC chips come up on the used market for sub $500 very soon
>Octo-channel memory
>So many PCIe lanes


>> No.72228675

They are out to oem/odm partners. From, there will make their way to first tier customers, like Google and Microsoft. After that they will move into availability through channel partners and finally to retail like newegg, or Amazon.
So yes it is a release today, if you are Foxconn, or Dell.

>> No.72228703

>for sub $500
You wish m8. And good luck with the motherboards.

>> No.72228704

I like how AMD isn't pulling punches at all. Most of these slides used to be vague. You'd see terms like "competition's leading product" and relatively vague bars. AMD is straight up announcing model numbers.

>> No.72228723

motherboards aren't hard to get. They're expensive sure, but they're all over Newegg for $450+ brand new for single socket solutions. And maybe not the 32c/64t for sub $500, but easily the 16c/32t. I'd gladly pay the premium for the ability to run Registered ECC DIMM's.

>> No.72228796
File: 136 KB, 993x583, AMD-EPYC-7002-v-2nd-Gen-Intel-Xeon-Scalable-Top-Line-Comparison.jpg [View same] [iqdb] [saucenao] [google] [report]

O shit this is almost sad now.
From servethehomes review of Rome.
"As some perspective, the $4425 list price AMD EPYC 7702P essentially puts the capability of a quad-socket Intel Xeon Platinum 8253 ($3115 each) system into a single socket. The systems have the same memory capacity, about the same raw CPU performance, and the AMD EPYC 7702P solution has slightly more PCIe I/O bandwidth. The Intel Xeon Platinum solution there has more raw memory bandwidth but now has four times as many NUMA nodes as the AMD solution, and requires a more costly quad-socket server."

>> No.72228806
File: 23 KB, 390x363, 1557242340120.jpg [View same] [iqdb] [saucenao] [google] [report]

>8/16 1900X is still 250 bucks local
>mobo is more expensive

>> No.72228811

at least intel has laptop dominance and amd isn't even close there

>> No.72228827
File: 690 KB, 1034x609, 996557.png [View same] [iqdb] [saucenao] [google] [report]

>> No.72228980
File: 55 KB, 593x767, 1559111067398.jpg [View same] [iqdb] [saucenao] [google] [report]

>thought around 10k would be cheap af
>AMD launch best processor under 8k
AMD server marketshare WILL jump to over 10% by the end of the year.

>> No.72228997

Fucking Forest getting hyped himself when on stage. Dude must love his job right now.
>just a little bit cheaper.

>> No.72229010

But anon, those cores are glued together garbage! Intel has the far superior monolithic architecture with dedicated section for exploits!

>> No.72229015

based Wendell on the right

>> No.72229054

>less than 7k
that's the most anti-semitic post I've ever seen on this cambodian sightseeing forum

>> No.72229092

what do you think AMD's yield rate is on their absolute top binned 8c16t chips to be able to make the upcoming 3950X on AM4 AND have enough for Threadripper and EPYC.

>> No.72229114

Its probably predictably low, given how Threadripper was removed from AMD's roadmaps for a while, and the clock binning spread per CCX for the consumer Ryzens is so varied. You don't need super high yields to get moderate volume, you just need to run tons of wafers, which TSMC absolutely is doing.
They might be getting around 30-40% known good candidates per wafer.

>> No.72229193

More from the STH article;
>We are also not allowed to name because Intel put pressure on the OEM who built it to have AMD not disclose this information, despite said OEM having their logo emblazoned all over the system. Yes, Intel is going to that level of competitive pressure on its industry partners ahead of AMD’s launch.
What the hell do they gain from that? Even sth who're fairly tepid when it comes to shitting on any vendors is calling intel out.

>> No.72229230

Nah, CCD KGDs are up on the moon, the problem is variance for anything higher clocking.

>> No.72229277

the fact that the 3950x is at such a price premium and the fact that it was a delayed launch means the yield rate is going to be among the lowest of all time. make sure to suck off your local store owner or you ain't getting one until 2020.

>> No.72229289

>given how Threadripper was removed from AMD's roadmaps for a while

Nigga please, threadripper always launches last. It wasn't removed from any roadmaps, just as it wasn't on any in the first place. The only fucking thing they ever said about TR3 is that its going to come after epyc. Were gonna hear about it at hot chips and its gonna come out before 2019 is over, screen cap this

>> No.72229305

if I remember right amd said threadripper would be announced q3 released q4

>> No.72229340

As the other anon mentioned, it might just be the yields for the highest performing chips. We can already see it's pretty hard to even get to the 4.5ghz+ for an appreciable time where tr3 and the 3950x are going to have the highest boost clocks. They probably have a far lower available die for that where being priced at $750 might reduce the demand as well. Thread ripper is also pretty far out of the mainstream where the top chip will be close to $4000

>> No.72229394

the thing is the 3950x is expensive compared to the rest of zen 2, but it's still damn cheap compared to any similar intel processor and way cheaper than entry level workstation builds. the demand is going to be hit from a million different angles. mark my words, the 3950x will be nearly impossible to get your hands on.

>> No.72229415

>mark my words, the 3950x will be nearly impossible to get your hands on.

Zen 3 with EUV is probably to solve that problem with better yields.

>> No.72229444

Still slower in gaming though, sadly.

>> No.72229449

It's still pushing what's considered mainstream. The 1950x and 2950x are hitting the tiers and have both done extremely well already. AMD have earned the ability to price it where they can and I'd even suggest that intel's new $480 price point has allowed them to bump the 3950x above where they initially planned the 16 core to land.
But it's still very expensive and out of reach for most consumers, then again I'm also suprised when intel managed to make $330 the most popular price point for cpus when most people only want a $200 cpu, thank god the 3600 appeared.

>> No.72229588

Loling at retsrds not buying amd stock yesterdaycat 28. Its never going tocbe that low again, i assure you.

>> No.72229594

>implying the entire world isn't on the brink of an economic collapse

>> No.72229608
File: 91 KB, 818x693, 1541538120178_0.jpg [View same] [iqdb] [saucenao] [google] [report]

>he didn't buy AMD at ~2$ per share

>> No.72229618

go away bugmen. the US is fine, the only thing that matters.

>> No.72229627

Nah, enterprise guys just buy more Intel until they're completely sure they want to move their ENTIRE infrastructure to EPYC. AMD will eat the new markets, but the old buyers will take some time. We'll see huge AMD marketshare gains if Intel is still not on chiplets in about 5 years.

>> No.72229633

>implying all the big money in the usa isn't hording cash right now in record amounts because everybody is preparing for a crash to buy everything up for cheap
>implying 9/10 of the most common crash metrics aren't flagged right now for the american economy
trump is doing everything he can to make it happen after the election but everybody can see the writing on the wall

>> No.72229658

god why did they have to fucking name it EPYC
could have easily added a prefix to zen like ryzen

>> No.72229663

the guys not considering amd are the same guys who don't upgrade often or put in many orders

>> No.72229675

I agree, epyc is a fucking lame name. and they have such badass names elsewhere too, like the infinity fabric or threadripper.

>> No.72229750
File: 206 KB, 893x459, NineDiesvsFour.png [View same] [iqdb] [saucenao] [google] [report]

Alright now this is EPYC

>> No.72229756

I had to scramble to add capacity to our farm when the perf regressions tanked our VM performance. That meant more Intel hardware just because that's all the vendor offered and we needed it right the fuck now.
We've since started testing AMD based equipment and gone to tender for the supplier.
t. Sysadmin

>> No.72229769

Sure it will. After the split. :3

>> No.72230206

You misunderstood them, deal with it fag.

>> No.72230331 [DELETED] 

Oy vey, this post is anti-semitic. Please remove it.

>> No.72230443
File: 18 KB, 637x631, 7F325118861E410F921FB4A4C55696F0.png [View same] [iqdb] [saucenao] [google] [report]

>why do datacenter CPU cost many moneys?

>> No.72230695

Wow you're right, I see spastika in his post

>> No.72230721

i should have fucking listened

>> No.72230845


>Zen3 expected to bring SMT4 to the table
>up to 4 threads per core

If they somehow manage to keep the clocks and TDP the same, and leverage the benefit of EUV's die reduction to instead put towards 4 fucking threads per core, with another 3-5% IPC uplift. Can you imagine it?

An EPYC 8842 with 64c/256t for a 1S system or 128c/512t 2S system for like ~$10-12k? Oh and 5nm EUV is expected to bring an 80% increase in transistor density despite basically being a half-node shrink. Which would mean that at 5nm EUV, Zen4 could give you 128c/512t for 1S and 256c/1024t for 2S at around the same $10-12k or even up to $14k price point with a 5-7% deficiency in IPC compared to the rumored 18% IPC uplift that Icelake will bring late next year.

Theoretically 1024 threads per rack.

>> No.72230870
File: 76 KB, 609x587, 1563842210668.jpg [View same] [iqdb] [saucenao] [google] [report]

>Theoretically 1024 threads per rack.
oooh fuck i cant wait to waste part of the company budget on something like that

>> No.72230882


>> No.72230892

> 7742 - $7,327.27
oh yes, intel is finished. LMAO.
won't happen. i'd bet my house on it. you know why? because AMD is run by incompetent communists.

>> No.72231023




Intel is fucked.

>> No.72231067

You didn't get the point at all.

>> No.72231199
File: 1.00 MB, 350x191, dis gon b gud.gif [View same] [iqdb] [saucenao] [google] [report]

>Intel even struggles to come out ahead in their own MKL-DNN software tailored for features AMD doesn't even have
Holy fuck that's got to hurt.

>> No.72231227

what the actual fuck is this font rendering
this is beyond bad

>> No.72231414
File: 281 KB, 1172x702, Supermicro-1U-WIO-AMD-EPYC-7002-and-4x-Intel-2U4N-Xeon-E5-2630-V4-Nodes.jpg [View same] [iqdb] [saucenao] [google] [report]

>The TDP of the AMD EPYC 7702P is higher than the Intel Xeon E5-2630 V4’s, but when you are replacing 6-8 sockets to 1 socket, the power savings are absolutely immense. We have not seen 6:1 consolidation ratios in under 2.5 years of technology advancement happen in the industry.

Literal piles of Intel garbage being tossed into the trash and replaced with a single rack. Holy shit.

>> No.72231535

>We have not seen 6:1 consolidation ratios in under 2.5 years of technology advancement happen in the industry.

Maybe if Intel didn't fuck up so horrendously with 10nm, the gap wouldn't be so big. Even AMD expected to compete against 10nm chips, but here we are.

>> No.72231546

I know that every architecture is different and have it's advantages, but why Zen perform so absurdly better in this job?

>> No.72231577
File: 107 KB, 825x842, intold.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72231580

SMT4 is stupidly pointless on everything not deep enough.
And Zen3 will be far more than 3-5%.

>> No.72231588
File: 58 KB, 883x536, STOPHE'SALREADYDEAD.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72231595

who the fuck calculates global illumination on CPU?

>> No.72231614

I mean shit, at the rate AMD is pushing its core counts it might be reasonable on zen3, especially if they add more even deeper avx support. This shit is pushing boundries.

>> No.72231619

Not Intel apparently

>> No.72231659

7nm EUV doesn't provide enough of a power reduction to increase core count, or likely add AVX512. They could add another ALU and AGU to the INT core and try to increase decode width though.
Either that or very marginally improve clocks.

>> No.72231668

When will we get AWS instances?

>> No.72231876

Is high end laptops the only thing where Intel isn't losing?

>> No.72231892

Wait until AMD launches Renoir next year.

>> No.72231906


>> No.72231934

I thought it was funny, anon

>> No.72231995
File: 2.94 MB, 800x450, stray cat.webm [View same] [iqdb] [saucenao] [google] [report]

Thank you for not being retarded. Heres a stray cat I met a couple years ago.

>> No.72232171

Consequence of smaller node size. Literally a direct consequence. They could design the most retarded architecture and it would still draw less power by virtue of 7nm.

>> No.72232181

People are dropping intel as fast as they can. You have to understand that a datacentre full of intel is like a $30,000,000 investment, they're not going to dump hundreds of millions of dollars of servers all at once. They're going to get their use out of it, and then never buy intel again.

>> No.72232216


>> No.72232261

But how? is it intels double socket solution that is all fucked?

>> No.72232275

Uh, son. One AMD server processor is worth two intel ones, for half the price.
Does that make the position of someone who is looking to buy a server processor (one socket or two) easier to understand?

>> No.72232285

no I understand that, what I don't get is WHY is that. what did Intel do to make their Goldiums perform like this? Is this penalty of monolith design? Is this penalty of Intels double socket design? I can understand if intel was 2x slower but this is just ridiculousness.

>> No.72232286

>who the fuck calculates global illumination on CPU?
You might see a trend back to CPU rendering for 3d. A single 250 watt GPU will likely lose very badly to 64 core epycs in 3d rendering. GPU based rendering systems arguably only took off due to extreme stagnation in CPU core count and performance. The only thing preventing complete domination is X86 is still very overpriced, 64 core epycs are $6k and can't beat 6x $1k GPUs.

>> No.72232292
File: 60 KB, 641x431, 1349503121630.jpg [View same] [iqdb] [saucenao] [google] [report]


>2S EPYC 7742P
>replaces 12-16 sockets

So basically, a 2S AMD EPYC 7742 rack replaces 6 to 8 Intel 2S Xeon racks. We're basically looking at a 550W TDP vs 1.02kW of per/watt diff.

50% TDP reduction at equivalent cores with better IPC, double the threads, 2x memory capacity and up to 2x bandwidth on 128 PCI-E Gen4 lanes at equal pricing of simply the CPU vs CPUs.

This is the BIGGEST upset in the last 30 years. Holy fucking shit.

>> No.72232310

Double socket doesn't have anything to do with it - it's the monolithic CPU. AMD can fab their CPUs in parts and then glue them together. With big die sizes means that there's a hgiher chance that a single bit of it will be broken during the fabbing process. Imagine one of those silicon wafers, and during the fabbing process there's just flaws randomly spread over it in tiny points. Smaller chips means you can dodge the flaws while losing less silicon.
Intel on the other hand is making massive dies and then have to throw more shit away.

Throw in a more inefficient CPU design and a higher process node and you have an absolute one sided massacre.

>> No.72232313

If you play dwarf fortress you're not the target audience

>> No.72232321

You can have 64 people playing dwarf fortress on a single CPU.

>> No.72232366

Zen and Zen+ were already really efficient at lower clockspeeds, its only when they went towards 4 GHz that efficiency went mostly out of the window.
With Zen2 the design has emphasized efficiency once more and the node size does the rest, as >>72232171 said. My 3700x fixed to 3,6 GHz uses less power than my notebooks 4700HQ while delivering nearly 3x the performance.

>> No.72232458

Delayed 10nm process means house fire CPUs if they scale up any further
L3 is scalable per CCX on Zen, so you get half a gig of L3 which is has nearly the same latency now as Xeon equivalents now so it can hide DRAM latency immensely.
Huge core/thread scalability due to chiplet/IO design on EPYC is a huge win, lower cost per chip, higher volumes, better yields
Xeon's mesh is more scalable but penalizes latency
Now AMD has pure AVX256 with no FMA, nullifies some AVX256 advantage from Intel, where Intel can only pull wins from AVX512 (house fire, can't be sustained easily) or single thread boosts as they boost like 15-20% higher.
Basically its over for Intel, MKL™ advantage is literally being walked all over with AMD's improvement in IPC, AVX advancements, and process lead, not even the SPEC2006 autists are on Intel's side anymore. It's gonna be a rough year for Intel in CPU (10nm volume pushed back to 2021 btw, barely competing with Zen2+ and literally Zen3 5nm at the doorstep knocking)

>> No.72232478

Until Atom.

>> No.72232666

Basalt and rockpilled.

>> No.72232675

meant for >>72228290

>> No.72232690

>no avx-512
trash cpu, you retarded drones.

>> No.72232763

So get an intel for x265 and use zen for literally everything else? Seriously how much of a typical workload really takes advantage of avx512 instructions?

>> No.72232787

>still just as good and even better in AVX-512 workloads despite of that price/performance wise.
>don't have to make it 1000W

>> No.72232793

We need benchmarking graph to illustrate this comment and OP's picture.

>> No.72232801

>TDP 120W @ 2GHz - 3.2GHz
How does that work with much higher clocked 3950X being 105W? Not mentioning the 155W models because looking at the cache they clearly have more dies. But 7282?

>> No.72232808

i do

>> No.72232816

even workloads that can benefit from AVX-512 would run better on AMD @ the same price just thanks to more raw processing power.

>> No.72232831
File: 32 KB, 244x295, 1527208595496.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72232892

Intel is also leader in delivery of vulnerabilities per watt.

>> No.72232973

Ehm... No, not for the newest 400w CPU

>> No.72232985

Is it high-compression jpeg converted into png?
My eyes are bleeding.

>> No.72233138
File: 86 KB, 401x245, 1533901898627.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72233228

>You'd see terms like "competition can't compete"

>> No.72233235

It's called THREADRIPPER, kiddies.

>> No.72233281
File: 405 KB, 998x1374, 1548926169288.jpg [View same] [iqdb] [saucenao] [google] [report]

>as a whole at least x2 times cheaper than "equivalent" Inturd setup

>> No.72233287
File: 4 KB, 190x190, 1542682116612.jpg [View same] [iqdb] [saucenao] [google] [report]

>tfw waiting for zen69, because waiting is more fun than buying

>> No.72233297

Is this that guy from movie mummy? He got fat.

>> No.72233306
File: 212 KB, 1800x850, 1564604313888.png [View same] [iqdb] [saucenao] [google] [report]




>> No.72233332

>I like how AMD isn't pulling punches at all
Intel used bullshit legalese and bribes to keep AMD out of major OEM desktops for years, why should they?

>> No.72233336

>what do you think AMD's yield rate
TSMC's gen1 7nm was reported to be having roughly 92% of perfect bins, 3% faulty, and the rest 5% is trash that's going into next gen consoles. Wasteless development. In comparison, gen1 Zen had perfect yields at roughly 65% and Zen+ had them at 87%. Meanwhile, Inturd's 9xxx garbage shit has them at less than 40%, kek. 7nm EUV will probably be near-100% at this point, if TSMC continues at the pace it is at right now.

>> No.72233391

IBM been doing it for quite a few years now, though. They didn't have as high PPC, IPC, and frequency, but 1024 thread CPUs been there for a while. AMD doesn't bring more cores or threads to server, rather than much better efficiency and increase in PPC/IPC/frequencies.

>> No.72233396

You forgot to enter the tripcode and log in yet once again, Francois.

>> No.72233415
File: 143 KB, 540x995, 1548351871262.jpg [View same] [iqdb] [saucenao] [google] [report]

See >>72227949.

>> No.72233426

Even IF 10nm was a success, AMD already started the Zen domination waaay back in 2017, so no, 10nm would've NOT saved Inturd's lame sorry-ass.

>> No.72233428

instead of running 2 channel memory rome can run full 8 channel, and instead of 24 gen 4 pcie lanes rome is pushing 128. that is gonna bump up the tdp

>> No.72233436


>> No.72233444

Fixed your mistakes

>> No.72233452

>a 2S AMD EPYC rack replaces 6 to 8 Intel 4S Xeon racks
Here you go.

>> No.72233470
File: 1.19 MB, 1152x924, INTURDTSXDED.jpg [View same] [iqdb] [saucenao] [google] [report]

AMD already proven it can bean amount of calculations AVX 512 does by simply broteforcing via doing AVX 256 x2 times over. And now TSX is also completely useless/irrelevant/deprecated too, according to RPCS3 dev's latest report. Keep seething, Intbecile.

>> No.72233478

>because AMD is run by incompetent communists.
i doubt there are that many heebs working in AMD tho. remember: bugmen aren't human

>> No.72233483

>being utterly slaved out by the inane alimony for 17 years
Yeah, live does that to you. Women are pure evil, fren.

>> No.72233517
File: 3.65 MB, 859x556, INTURDSLOW.png [View same] [iqdb] [saucenao] [google] [report]

A daily reminder:

>> No.72233522

jesus, that's ridiculous

>> No.72233565

Sad... I really like him, he's pretty funny.

>> No.72233599

Don't be. He's now completely finally free. The >>72233281 photos were made right outside of court after he paid off the very last of his alimony. He's getting back now. A free man. Punished Brandon no more. He's Big Boss now.

>> No.72233618

That's great. Will pirate his movie to watch this evening.

>> No.72233656


>> No.72233762

>barely competing with Zen2+ and literally Zen3 5nm
huh? thought AMD is skipping any kind of Zen2 refresh and go straight to Zen3 but on '7nm+' aka 7nm EUV from TSMC. but i only get that from roadmaps and i think a Norrad interview from a site. can you link state your source on 'Zen2+' ?

>> No.72233767

source us, anon

>> No.72233786
File: 166 KB, 1420x1497, 1519330482383.jpg [View same] [iqdb] [saucenao] [google] [report]

>Punished Brandon no more
you are fucking right.

>> No.72233881

Get Encino Man.

>> No.72233924

>64 core with 2.25GHz base
Piednoel BTFO

>> No.72233933
File: 61 KB, 612x397, 1558057941722.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72233964

Did he already delete the tweets where he claimed 64 core Rome wouldn't be able to hit 1GHz?

>> No.72233968

I hope you enjoy the mind controlling brain worms

>> No.72234007

When you have to work to survive, anything is possible.

Intel has been sleeping, doing jack shit in terms of revolutionary CPU design. They got caught with their pants down, and are thusly fucked for the next 2 years in the server market.

>> No.72234016

The sth article even mentioned they put pressure on the box vendor to not release their name, so they're already trying and it's confirmed in an article. Anand on the hard just said the name of the box maker.

>> No.72234048

>What the fuck do I need 64 cores for?
X4, Star Citizen.
X4 is Vulkan-native title and it eats up as many cores & threads as you can throw at it, due to a fuckton amount of calculations being performed in the background at all times because entire game is one big persistent instance and largely nothing at all gets culled any. It's especially noticeable after 16+ hours of single session play, when a shitload of AI is being deployed and you operate a dozen large fabs (breweries especially). It's been like this with X3 also, but X3 is a DX9 title and thus was severely limited in it's AI capacity. X4 is a fucking beast when it comes to core/thread usage. X4 also supports resolutions all the way up to 16K, which is absolute overkill and clearly intended as a future-proofing.

As for Star Citizen, even it's current unfinished state, the Persistent Universe mode (which is the MMO part of the game) already loads over 16 cores to 100% due to DP calculations and persistent instancing per-quadrant (which in on itself is several dozen thousands of kilometers in a cubic direction distance) and up to 12GB of RAM when being "maxed" (it can't be called actual maxed yet, because some effects are still unimplemented/unused due to game being in Alpha) in 4K (and the game already can do 8K).

>> No.72234058
File: 356 KB, 1771x960, 1OfELIqRkFLM9078.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72234065

>for the next 5 years
Fixed your mistake.

>> No.72234076

My bad amigo

>> No.72234174
File: 864 KB, 991x917, e8de1fa7761b239e954776ddcbc04d8289afed59fb035513b0da052cec0812d4.png [View same] [iqdb] [saucenao] [google] [report]

Can someone with good enough shooping skills change pic-related to EBYN 2?

>> No.72234223

seriously, what price should I get the stocks? I have no savings and I am broke af but I plan to spend my salary on this shit.

>> No.72234281
File: 2.88 MB, 1381x729, WTF.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72234309
File: 86 KB, 418x494, E4.png [View same] [iqdb] [saucenao] [google] [report]

>cores/threads: 16 August

>> No.72234326

don't worry, i won't last long

>> No.72234329

The chad Roman vs. the virgin Semite

>> No.72234331


>> No.72234347

>I won't last long


>> No.72234356

Even Ryzen 3xxx laptops still get only 2/3 battery life of their Intlel counterparts. Until AMD fixes idle power consumption, their laptop CPUs are gonna remain niche or low-end only.

>> No.72234376

By the time Star Citizen is gonna be more than a tech demo, we'll have 256 core quantum CPUs.

>> No.72234393

I prefer 24 December cores.
>Jingle threads, jingle threads, jingle all the way

>> No.72234403
File: 1023 KB, 2000x1417, 1543283399097.jpg [View same] [iqdb] [saucenao] [google] [report]

It says "Aug", not "August".

16 "Augmented" cores:

>> No.72234406

gween tee

>> No.72234587

what an outer destruction

>> No.72234627
File: 291 KB, 904x952, 9377cbffc72dd00730bcf52bbe1a8be7e481a3d39915881bbcb6be9a828ede2a.jpg [View same] [iqdb] [saucenao] [google] [report]

No, this is:

>> No.72234674

>meanwhile Intel consumes 600w baseload per chip

>> No.72234705
File: 89 KB, 743x928, 1558223771605.jpg [View same] [iqdb] [saucenao] [google] [report]

>recommends disabling
Wow, good thing every server OEM leaves it enabled in their firmware by default, huh?

>> No.72234828


>> No.72235048

Higher is better - (((Tom's Hardware)))

>> No.72235071

GPUs (aka nvidias reason for existing at all) will be deprecated by 2024 due to these monster cpus being able to do the job even better. I am calling it now. There is a reason AMD seems to put way more effort behind the CPU. They don't want to eat their own ass because they know GPU roadmap begins tapering very soon.

>> No.72235101

One 64c Rome could probably ray trace a modern AAA game title purely in software mode at playable fps if someone took the time to rewrite the rendering pipeline.

>> No.72235110


>> No.72235114

>Is high end laptops the only thing where Intel isn't losing?

Haven't they recently acquired someone who's put in charge of that? I wouldn't be surprised if they started an offensive at laptops as well once the Zen2 architecture goes mobile, completing the full-scale assault on Intel.

>> No.72235339

>(((((Ryan Shrout)))))

>> No.72235374
File: 3.51 MB, 732x627, RAYEBYN.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72235432

>faster than 2 top xeons

>> No.72235465

>nvidia in HIV tier

>> No.72235480
File: 3.15 MB, 732x564, EBYNTRACING.png [View same] [iqdb] [saucenao] [google] [report]

>actually reaching the absolutely maximum wall of what the current latest version of the benchmark is capable of handling
>on unrestricted Linux


>> No.72235516

>twice as faster than two top Zion Platurds

>> No.72235550

and that's on base clocks.

>> No.72235587

And I thought hard porn was banned on jewtubes...
Wew lads, that's borderline gangrape.

>> No.72235633


>> No.72235637
File: 1.83 MB, 288x377, 1537478586221.gif [View same] [iqdb] [saucenao] [google] [report]


>6 cores less than the competition
>175W more than the competition

>> No.72235859

>*8 cores less
Oops, sorry, miscounted.

>> No.72235874

Nvidia likes AMD as far as their CPUs are concerned, this comes at no surprise:


>> No.72235875

I would love to see gaming benchmarks, I just wanna know what happens

can you do more than quad sli with epyc because of the additional pci lanes?

>> No.72235896
File: 11 KB, 501x504, merchant.gif [View same] [iqdb] [saucenao] [google] [report]

>one (ONE) Zion Golturd 6148 is 3500 jews
>TWO EBYN ROME 7452s are ~6000 jews

>> No.72236112

>Ryzen Pro 3 T-series ThinkPad

Holy fuck, this is the official /g/ laptop now.

>> No.72236198
File: 48 KB, 594x460, 2019-08-08 11.04.55 twitter.com d3d372412fa2.jpg [View same] [iqdb] [saucenao] [google] [report]

Who is in the right here?
FPiednoel claims that Rome is a joke in VM performance.

>> No.72236212

>all of media and industry with benchmarks and figures
>one guy on twitter who reckons something

>> No.72236222
File: 124 KB, 301x398, lol.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72236412

Francuck Copenoel will commit suicide this weekend.

>> No.72236523

>GPUs (aka nvidias reason for existing at all) will be deprecated by 2024 due to these monster cpus being able to do the job even better. I am calling it now. There is a reason AMD seems to put way more effort behind the CPU. They don't want to eat their own ass because they know GPU roadmap begins tapering very soon.
Looking at blender benchmarks it looks as though 64 core epyc is pretty close to a 2080ti and radeon VII while using less power. If they really wanted to they could probably write 3d API drivers for it, the only reason not to is they can sell them as server only products for 4x the price of a high end GPU. I guess the 2020 consoles will probably be the last traditional GPU platforms after that just heterogeneous CPUs with big/little cores like ARM does.

>> No.72236567

>get shitstomped by Nvidia in GPUs

They're only focusing on CPUs because Intel was sleeping and was fine with incremental upgrades.

If Intel were as aggressive as Nvidia they would put AMD out of business.

>> No.72236619
File: 535 KB, 877x943, 1554676335994.png [View same] [iqdb] [saucenao] [google] [report]

Stop buying Epyc goy, it will set your server farm on fire!

>> No.72236625

Waiting for Vulkan Explicit Multi-GPU and Threadripper Gen3 to see.

>> No.72236637

I mean a gpu is just a different kind of cpu, it's always possible the standard becomes some sort of blend of the two

>> No.72236666

PC graphics already went through a couple of cycles of "dedicated hardware -> software rendering" race. I guess it's time for a new one again.

>> No.72236710

>I would love to see gaming benchmarks, I just wanna know what happens
>can you do more than quad sli with epyc because of the additional pci lanes?
Unlikely it would improve anything. Discreet GPUs and rasterized graphics are rapidly approaching the end of their usefulness.

>> No.72236779

imagine a motherboard with like 5 cpu sockets and you can populate as many as you feel like

>> No.72236862

>one guy on twitter
Literally the guy behind housefireburst at Intel and likely his life savings in Intel stock that he needs to defend.

>> No.72236914

>fudzilla taking bribes again
You'd think they'd have learned from last time.

>> No.72236926

Netburst is literally boomer era, why is he still loyal to Intel despite no longer being employed there? From the looks of it he's been out of Intel for eons.

You don't see Jim Keller being a hardcore AMDrone.

You don't hear anything from the folks who came up with Core and they're the ones who actually deserve to be hardcore Intel loyalists, being the ones who actually trashed AMD.

>> No.72236990

>why is he still loyal to Intel despite no longer being employed there
lots of intel stock options as per standard employment contract at intel

>> No.72237062
File: 456 KB, 480x262, Brushing Eyes.webm [View same] [iqdb] [saucenao] [google] [report]

>dat font

>> No.72237093

That's because the man doesn't give a fuck about company loyalty, dude just chads about at whatever company he works at. What do you expect from the dude who married the sister of Jordan "Wash your penis" Peterson?

>> No.72237122

>rest of the sane world
>Who's in the right here?
Gee, schlomo...

>> No.72237133
File: 1.63 MB, 2560x1440, 1562539562164.png [View same] [iqdb] [saucenao] [google] [report]

>shitstomped by Nvidia

>> No.72237344

Intel is a sect that brainwashes its members.

>> No.72237572

Any game that does 3D can be set to any resolution, ie. Quake 1 at 16K. It means nothing..

>> No.72237872

How long until intel replaces their datacentres with AMD?

>> No.72237896

There was a Rome sample leaked with a clocked speed of 2.4 GHz and he assumed that was the highest Rome could clock, but then it turned out the actual CPUs could boost up to ~3.4GHz, so he ended up getting BTFO.

>> No.72237920

>I posted it again
always makes me laugh

>> No.72238101

This guy is a former Intel employee and definitely mentally deranged.

>> No.72238149

They probably already have

>> No.72238175

>higher is better

>> No.72238388
File: 64 KB, 1414x782, 1562549113755.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72238408

Piednoel being a moron "on purpose" to get insulted and claim the high ground.
Conveniently ignores all the good counter arguments. He is jut a troll.

>> No.72238526

No he's legitimately mentally ill.

>> No.72238551


>> No.72238604

As Epyc as hot chamber hahaha

>> No.72238751

I love how you guys are actually thinking this through now due to the the new reality of Rome. I tried to start the "GPUs will die off" conversation on /g/ about 6 months ago and was mocked out of town. There are already some projects out there that seek to fully implement the APIs of OGL/D3D in x86. Also, imagine the debugging experience for a developer on these types of projects:
>Nvidia driver throws some exception, oops I guess I'm fucked, ill fiddle with these shaders for a bit maybe that will help.
Glorious Software-Based Rendering Future:
>Oh I got a nice clean stack trace here... ok its on L371 of source file X...
>[3 minutes elapse]
>ok nice, next problem please

>> No.72238786
File: 52 KB, 590x599, 2019-08-08 14.43.51 twitter.com c286f355b641.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72238841

Jeder Stoß ein Franzos'.

>> No.72239142

i'm confused...
did he just pretended to be retarded?

>> No.72239164 [DELETED] 

He pretended to be prentending that he was retarded.

>> No.72239260

What a retard

>> No.72239403

the fuck

>> No.72239703


>> No.72240280

Watch Airheads

>> No.72240405

In other words: autistic no-lifer.

>> No.72240471 [DELETED] 

Jews are literally sociopaths

>> No.72240908


>> No.72241528
File: 10 KB, 192x246, 1542380403591.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72242193


>> No.72242250
File: 983 KB, 370x339, DED.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72242957

Holy shit, you fucking retard.

>> No.72242973

What is 24-Dec then?

>> No.72243088

That's how many decrementor units it has.

>> No.72243502

Wait for intel revolutionary 7nm in 2077 AMD btfo !

>> No.72244122


Jesus, it is the K7 launch all over again. Back in the day, Intel was so scared of the K7 launch that motherboard vendors were forced to launch Slot A motherboards in plain cardboard boxes (not joking) with simple SKU coding.

>> No.72244667

well, why is amd so much better, going to take a little explaining.

1) specter/meltdown mitigations are not to be ignored, this hits intel fairly hard for certain loads.
2) 12 more cores per cpu, 24 cores total.
3) on an avx load, amd doesn't need an offset and intel cant function without one ,though avx was a bit hamstrung on zen1 so the point with that one is moot.
4) in applications that work well with amd, amd has around 25% higher ipc even on zen 1
5) amd clock for clock, even on normal applications that are not geared for it at all, saw amd matching if not exceeding intel and zen 2 pulls ahead a quite a bit.

my best guess here is that what is being stressed here is shit that got neutered by specter and meltdown mixed in with amd not needing an avx offset, and running at the same clock speed as intel.

>> No.72244792

a gpu on x86 can work, in fact last time it was done it did very well... but intel shelved it for knights landing.

you would need to severely restrict the x86 instruction set, but doing a gpu on x86 is possible just not doable as a main cpu either.

>> No.72245952

>50% TDP reduction at equivalent cores with better IPC, double the threads, 2x memory capacity and up to 2x bandwidth on 128 PCI-E Gen4 lanes at equal pricing of simply the CPU vs CPUs.
So Intel basically has just lost the next 2 years in terms of server sales.
No wonder Intel hired Jim Keller after Zen came out.

>> No.72246979


>> No.72246986

You meant K6~K8.

>> No.72247021

It's actually way better, because back then AMD had it's shares jump to 32$ during the launch, but nowadays they've clearly completely surpassed that. Not only did they repeat the moment, but they've actually gone beyond and then some.

>> No.72247466
File: 1.29 MB, 1450x950, E98998674927A5F75D0FCBC77904405FD438225B.png [View same] [iqdb] [saucenao] [google] [report]

>le reddit buzzwords

>> No.72247678

Oy vey, its jews vs Rome all over again. When will the messiah come?

>> No.72247988

delid dis sirs yes

>> No.72248439

jewtel jizz tim pozzed 14nm+++++++++++++ housefire ring -3 rootkit delidded lapped chiller AIDS speed holes israel botnet

>> No.72249221

Kek. It's growing.

>> No.72249311
File: 56 KB, 810x516, AMD-EPYC-7002-KVM-STFB-Virtualization-Workload-1-Benchmark-1.jpg [View same] [iqdb] [saucenao] [google] [report]

>Rome is a joke in VM performance

>> No.72249432


>> No.72249459

>random unknown virtualization software with random unknown sla that they had to put in for balance

>> No.72249558

For balance between what and hwat? This comes in at the point in the article where EBYN2 already thrashed Xeon at every price and TDP point and it was suggested top Xeon parts have to be discounted to 1/10 MSRP to stay competitive. This was just insult to injury.

And it's not even some synthetic bench with some random graphics engine nobody actually uses, as usual with gaymen benchmarks where Intel "wins" by single digit percentage. It's customer's proprietary line-of-business software, and the customer obviously cares about its performance enough to pay STH serious monies to bench their shit on at least 10 different HW configs to get best performance for price in production, so there's little doubt it's already optimized to its tasks and will be further optimized to the chosen arch. This is the kind of shit that earns people real money. Not "build gaymen pc" money, but "buy a yacht and a sportscar" money.

Literally this is why Momma Su was smiling on her first slide deck. 4* performance per $ of TCO is what gets you all of the contracts. 10% lead on performance per core (((in some applications))) is what gets you fanboys who don't matter.

>> No.72249596
File: 231 KB, 1446x808, the only areas where Cascadelake (green-ish line) is beating AMD's Rome is with their own MKL-DNN software.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72249634

What's up with the MariaDB outlier? Is it heavily singled-threaded, or just unoptimized code?

>> No.72250064


>> No.72250202

>they've clearly completely surpassed that
Adjusted for inflation?

>> No.72250292

Even more so.

>> No.72250312

Fuck these colors. Which blue is on top?

>> No.72250320

of course it will, just wait until GPU release
AMD rises on Ryzen, falls on anything else

>> No.72250396
File: 29 KB, 1500x1477, SRSLY 001.png [View same] [iqdb] [saucenao] [google] [report]


>> No.72250466

Well, i can't distinguish l1 and h3, so fuck you.

>> No.72250506

Zen 2 is on top.

>> No.72250580

I'd love to see what test doctoring they needed to pull those figures off because our hosting farm got completely fucked by the context switch penalties on every Intel SKU in our racks.

>> No.72250725

For now, can't wait for the moment where AMD takes those fat CPU stacks and pump them in their GPU division and blow nvidia out of the water.

If they can do it with Intel they can do it with Nvidia.

>> No.72250773

>Twice as fast
You might wanna go back to school mate

>> No.72250865
File: 288 KB, 996x600, 687ityu8568675.jpg [View same] [iqdb] [saucenao] [google] [report]


>> No.72250917

But why?
Cleaning pub loo is so much easier

>> No.72250924

Did that intel Twitter shill comment on it yet?

>> No.72251024

>what is GPGPU

>> No.72251174

Yes. You're late, he already pulled the "I was only pretending to be retarded" card.

>> No.72251367

> random unknown virtualization software
This fucking board...

>> No.72251501

buy intel sirs................

>> No.72251652


>> No.72252311


>> No.72252343

This person is mentally ill: https://twitter.com/FPiednoel/status/1159625209160192000

>> No.72252471
File: 351 KB, 2500x1282, 3700X.jpg [View same] [iqdb] [saucenao] [google] [report]

This isn't market disruption. This is market DESTRUCTION!

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