AMD EPYC and Intel Xeon

Damn, Intel is in a lot of trouble. EPYC consumes less power, can compete and outright destroy in some scenarios, has a stupid amount of possible IO and RAM support, and their yields are through the roof. What the fuck will Intel do in the upcoming years? Making a dumbhuge die will fuck them up too badly in the next 5-7 years. Zen+ is probably next year so we can expect 7nm EPYC on 2019, taking away Intel's process node advantage. What if AMD can get 6 cores on a CCX? Intel must be sweating bullets.
anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade

Other urls found in this thread:

anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade/11
mediafire.com/file/8v26cd117rcscs2/amadaDump.zip
twitter.com/NSFWRedditGif

FUD and promises of MCM Soon(tm).

More FUD, more bribes.

DELID THIS

>a-at least we still got 4S
>battle t-tested

>promises of MCM Soon(tm).
There is no other way they can compete. Their yields on those huge dies must be atrocious. AMD could undercut them further but choose not to. Also, holy shit the 1700 is at $270, I'm really tempted.

...

NOOOOO

he he
WHAT THE FUCK

But how much does it cost?

>The current Intel pricing draws the first line. If performance-per-dollar matters to you, AMD's EPYC pricing is very competitive for a wide range of software applications. With the exception of database software and vectorizable HPC code, AMD's EPYC 7601 ($4200) offers slightly less or slightly better performance than Intel's Xeon 8176 ($8000+). However the real competitor is probably the Xeon 8160, which has 4 (-14%) fewer cores and slightly lower turbo clocks (-100 or -200 MHz). We expect that this CPU will likely offer 15% lower performance, and yet it still costs about $500 more ($4700) than the best EPYC. Of course, everything will depend on the final server system price, but it looks like AMD's new EPYC will put some serious performance-per-dollar pressure on the Intel line.

obligatory

Server markets are going to end up being owned by AMD. I don't think anyone can argue with that.

A fraction of what Intel's offering does, that's all that matters. Business applications are all about profit margins, Expect Apple and every other workstation OEM to switch over to EPYC asap.

DELET

...

Apple needs to embrace RR for their next Macbook's, for the first time they will not be throttlebooks and they will finally have decent battery life.

Everything oscillates

If you are customer, you can be only glad that the intel monopol is over.

If you are amd fanboy, enjoy while you can. If you are intel fanboy, do damage control and post only single thread benchmarks.

Doeas anybody know how will the cpu coolers look like? Something as massive as ninja 4? Does even something exist for it yet?

Owned != permanently dominated

>gcc btfo
take that stalldrones

this needs to be stickied on the front page of Sup Forums

Oh dear. . . Looks like Intels TDP means jack shit yet again, with AMD using much less power under 100% load, though the idle and MySQL test looks funny for AMD, might be something going on there. Anyway, EPYC is ~25% more efficient under full load in that test (which EPYC beat Intel), that's a huge fucking deal. When you weigh TCOO, power consumption is pretty much the top of the list.

>server processor
Literally who gives a shit

I HATE AMD SO FUCKING MUCH

Leave Sup Forums.

less initial investment with less operating costs at virtually the same performance more or less.
EBYN

>MySQL test looks funny for AMD, might be something going on there
Lots and lots of hops between the dies/sockets.

Fun fact, unless you own a large-scale company you will literally never benefit from a server

Yeah, that would do it. Looks like an area for further improvement in the future for AMD. If they can get latency+overall power down from constant die/socket hops then they will really crush the market.

>you can't talk about F1 cars in a car forum because you will never need one anyway.

>not buying low end EPYC for your homeserver

that's a pretty qt corpora-tan

WHEN WILL AYYMD RELEASE SOURCE CODE?

I just want to point out that those Anandtech benchmarks are a bit biased in favor of Intel because they are using an old version of Ubuntu with a kernel that doesn't support the zen architecture.
>anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade/11
>"All of our testing was conducted on Ubuntu Server "Xenial" 16.04.2 LTS (Linux kernel 4.4.0 64 bit). The compiler that ships with this distribution is GCC 5.4.0. "
this is actually hilarious, Intel is legit about to be murdered in the server market.

entitled FOSS-tard plz jump from a cliff

Don't get me wrong AMD used to be great (I still use my librebooted d8 system with 2 AMD opterons) but now they have become just as botnet as intel.

>The power delta for the EPYC system is 177w.
>The power delta for the Xeon 8176 is 263w.
>The EPYC system performs better

This is a fucking bloodbath. AMD has pure gold on their hands.
Their 7nm parts are going to steal the entire data center market.

>Anandtech
>biased
Stop the presses! They can be dicks about it sometimes, but they called out Intel on their new AVX shit.

Well it's going to be POWER9+ versus EPYC2 in the end.
Intel's first 10nm server parts are far far far away due to yields.

>6cores/CCX
4 8 C O R E S

there aren't many sites that test 25.8k$ worth of CPU's. I agree with you but I'm just saying, if they run a newer kernel even on that old version of Ubuntu, AMD will have an even higher performance lead, and probably less power consumption as well.

one can kill a man with that brick

people in the future will look at it like we look at 80's technology

>Baidu
>Microsoft
>A&A
>Dropbox
>Dell
>SuperMicro
>Redhat
>VMware
>HPe

>80% of the server market as day 1 partners

>128 bit AES memory encryption on the fly
>32c/64t starting at $2100 for single socket support

hnnnng

>128 PCIe lanes
>2TB RAM capacity
>on all chips
>even the $500 8c/16t one
LITERALLY RAPING INTEL

REEEEE
HOUSEFIRES
AMDPOORS
PAJEET ON STREET

JUST USE INTEL DAMNIT UGHHHH

Brian, unJUST your fabs.
That 10nm node is a disaster.

pretty much, rape is the only word to describe how bad Intel just got shit on.

2x EPYC 7601 are only 8.4k
2x Xeon 8176 are 17.4k

7601 beating it by 40% in some tasks, while being on par in others, with an older kernel like this user said , all while using 100w less power. simply EPYC.

Tyan too. They've got the whole market as partners.

HOLY SHIT!!!

It is ogre. Intel will be bleeding money for the coming years. Let us see what will their (((diversity))) hires bring them now.
Fucking kikes.

holy fuck Intel what are you even doing...

"Power draw is recorded every second in our enterprise lab. During the Linux-Bench tests, we captured multiple high-draw bursts that only appeared for one second, cresting at 711W. The same granularity is used during our Linpack tests, but because of the 8176's lower AVX frequencies, we only recorded a 670W peak. "

...

>all while using 100w less power. simply EPYC.
Using 100w less power, while supporting more PCIe lanes, and more memory, all of which's controllers require power to run.
Honestly, I'm still shocked by that 7551p, fucking 32/64 for $2100

I'm crying brahs

yeah that's going to be one beast-mode workstation CPU that Intel has absolutely no chance of competing against with their shitty 40-44 lanes.

What the...

We need to go wider.

Yeh, between the 7551p and the 7251, Intel have trouble at both ends of the market. Then you've got shit like the 7351, and 7301, which improve on Intel's offerings there significantly, 7301 by up to 70% in SPECint, and that's where a good 60% of the server market alone lies is at that price point

It's actually more efficient

intel actually says they already sold 500k of new xeons

sure, but only due to it having more cores than the other models tested, its a silly non-factor metric. total system power draw is the only important metric to care about in the server space for obvious reasons.

Where can I find more of this?

Holy shit AMD's FPUs are fucking monsters
If these ever get 256bit pipes for AVX they're gonna eat insane amounts of power.

And this is why I don't give a fuck about AMD's GPU division, their CPUs will make them more money in a year then their GPU made them in 5 years

AMD should only care about making a few pro GPU SKUs and whatever goes into their APUs, rest is just noise.

And what are you going to keep on the server?
Terabytes of furry porn you autistic fuck?

That's how you get $2800 Nvidia GPUs.

I'm interested in gpus for the unified compute platform with hsa and IF. There are rumours floating around of a node with how ever many epyc cores and how ever many vega cores being visible to an external host/controller/software as a singlular compute unit.

Look at me give a fuck.

AAAAAAAAAY 7nm EPYC will devour Intel considering their 10nm part is in 2020/2021 at best and canceled at worst

You're an idiot

Stop assuming I care about fucking graphic cards, I'm running a fucking integrated GPU and if I wasn't I'd be running some

delid
dis

mediafire.com/file/8v26cd117rcscs2/amadaDump.zip

STOP MAKING THESE SHILL THREADS AMD

WE HAVE TO KEEP THE ISRAEL PLANT OPEN

SHREEEEEEEEEEEEEEEEEEEE

Oy to the vey.

not the guy you replied to, but you're a saint.
she's so cute, thank you so much.

>SHREEEEEEEEE
holy shit im ded

RAPING INTEL EVERY THREAD

Hosting furry porn and dank Trump memes

>SHREEEEEEEEEEEEEEEEEEEE

>AMD releases 3ghz base clock 48c96t 7nm 2EPYC4U chip
>Brian sits alone in his basement beating off with a belt around his neck silently wishing that tonight will be the night he doesn't wake up after he finishes

Let it happen.

I'm sad they didn't test any VMs the CCXs are purpose built for them.

How so?
I was looking into buying the 16-core EBYN and dedicate 2 physical cores from each CCX to 4 different VMs because I didn't know how they'd perform with cores in different CCXes

>2 Socket AMD Starship servers will have 96c/192t of actual fat x86 cores
I can only imagine the amount of RAM and IO these fuckers are going to have

>wah wah i don't care about gpus therefore therefore amd, being the only other company competing in the market should quit, leaving thousands upon thousands people who do care fucked

Y'all cherry picked hard.
That duel Threadripper system lost in a lot of tests.

The lower price and power usage is it's strongest point.

amd64 is not good choice for server.
POWER 9 much faster, cheaper and power-efficient.

Probably because they got tired of Intel raping them with prices.

Because CCX are mostly self contained and very low latency and can be called dedicated VM silicon, 4 cores per VM is optimal, but you'd get a similar core performance and core bandwidth if you use 2 or even 1 core.

5.4 is fucking ancient though I haven't kept track of how much Zen tuning they've done in 7.1.

Latency between CCX in a single die is still pretty high. How much higher is the latency between CCXes in other dies?

That's why you keep a VM in a single CCX; which is really fast.
Though if you're running something that doesn't depend on other threads, ccx latency, die-die latency and socket latency are all irrelevant.

To complete the victory, "Snowly Owl" will rip Xeon D a new asshole. Pic related.

...most likely for Google, nobody else will not bother with AVX-512.

To Google, back in December.

Tleadulippelu sugoiii!

It seems like EPYC is indeed another unmitigated disaster for the server arena, it should have been obvious from the lackluster desktop processors

>this is the current state of (((intel)))

t. Brian JUSTnich

>State of Intel shitposting in 2017
Sad!

Seething.