http://forums.anandtech.com/showpos t.php?p=37963428&pos

>implying AMD have even been competing @ the very high end market
Nigga are you high?

>>52421092
Imagine if the guys over at AMD had the funds at their disposal that Intel and Nvidia do.

They're gonna take the market by storm soon my friends, I swear it.

File: 1450844644634.png (39KB, 156x129px) Image search: [Google]

39KB, 156x129px

>>52421061
>5.1GHz (default clock)

jesus fuck

>>52421410
There go all the faggits who said Broadwell-E was going to clock worse than Haswell-E.

The only reason Broadwell sucked so much ass on 1150 was because of the L4 cache, I'm sure of it. Without integrated graphics holding it down, the clocks will be just fine.

>>52421398
>Imagine if the guys over at AMD had the funds at their disposal that Intel and Nvidia do.
Then they would not push technology as hard as they are, because they have a cushion to fall upon.

What has Intel invented in the past 30 years? Very little considering their budgeting.

AMD on the other hand have a very long list of technological achievements and inventions. Because they have been forced to innovate and sell the product to their competitors in order to survive.

>5.1GHz

My god.

Nice to see that they're able to push the clocks this high.
Then again it's about time that they start rising.
We've been dicking around in the sub 4GHz region forever.

Rumors are that it was designed for NSA

>>52421061
>Xeon with four cores

>>52421061
> OEM-only server chip
> high end server CPU

You almost had me there, OP.

>IT'S OVER, AMD IS FINISHED & BANKRUPT
>165W TDP

But hey, it's okay when Intel does it I guess.

I'll wait for Skylake-e or just purchase a cheap Haswell-e

>>52421499
who cares about that

5ghz bro

>>52421061
Holy shit, if I can get a Broadwell-E 6C/12T CPU that OC's around 5+GHz I'm buying that shit immediately.

>>52421500
Wasn't Skylake-E pushed to 2017?

>>52421499
>AMD 5GHz = 220W TDP
>Intel 5GHZ = 165W TDP

55W difference there senpai.

>>52421061
>4 cores
>xeon
No.

Not even Intel are so stupid as to shoot themselves in the foot like that.

Xeons target the server and workstation markets. And while single threaded performance does matter do a degree (Mostly in workstations where a heavy load is present) the ultimate goal of a server is to achieve a high core count and use properly threaded software in order to execute as many operations per second. Of which, a high core count will 99.9% of the time beat out a high performance, singlethreaded focusses product. Like this chip.

>>52421465
This is very true.

Lets just hope it can calculate primes ;)

>>52421512
Every Intel user cared when the FX9590 came along.

It's ok when Intel does it tho right famalamadingdong

>>52421532
55watt, and 14nm...a 55watt improvement is not all that impressive all things considered. Especially taking into account that the 9590 is 5Ghz over 8 cores. While this a quad-core solution.

File: overview-full.png (801KB, 960x778px) Image search: [Google]

801KB, 960x778px

>>52421475
>>52421483
It's a Xeon that means you can pair up to four of them. Maybe broadwell will being quad socket boards.

Even with two it's still a monster that will rape everything for years to come.

File: smug moeshit slut face.png (36KB, 268x237px) Image search: [Google]

36KB, 268x237px

>>52421573
>over 8 cores
>'8 cores'

>>52421475
Your point? There are plenty of quad core Xeons.

>>52421483
Intel has made some pretty crazy server CPUs in the past, most are custom made for certain clients so unfortunately they've never been released.

>>52421532
Also it'll be an order of magnitude faster than the AMD chip.

>>52421539
That's why this is a one-off thing. Someone ordered 5ghz quad-cores in enough quantity for Intel to develop such chips, this isn't by accident or on a whim.

>>52421532
>Intel still ends up twice as fast nevertheless

>>52421592
You have had the option to pair up Xeons like that for a very long time.

But it only being 4 cores is a huge detriment to server operators.
Maybe in a workstation I can see this fly, but even then, the software in use can make excellent use of available threads, so it is likely that the typical 8core/16thread Xeons will not be replaced by this clusterfuck any time soon.

>>52421532
ALL Haswell-e chips are over 140w tdp.

>>52421475
>Forgetting about the E3-1231 v3

Also: HT

>>52421592
The code starts with 2xxx which means it's a dual-socket chip, it also means it'll start from at least $2000 per chip.

The first number always refers to how many CPUs can be linked together. 1600 is single socket, 2600 is dual socket, and 4600 is quad.

>>52421636
Not really, it's obviously a specialized chip considering they have xeons up to 18 cores. It was built with high clocks in mind.

>>52421061
>4 cores 8 Threads

Octa core Xeon @ 5.1 ghz when?

I'd rather stick with 6370P Opterons.

>>52421615
>That's why this is a one-off thing
If it were a one-off ordeal, then Intel would be intelligent enough to keep this under the living room rug. As to avoid the bad publicity and difficult questions it may bring from share holders as to why they create something like this, only for it to never see the light of day.

No. I fear that this may find it's way to the marketspace. Should it really be for a large client, then I don't imagine their client will be happy with details of the processor going public. More specifically, some of it's capabilities.

>>52421573
>55watt, and 14nm...a 55watt improvement is not all that impressive all things considered
It's pretty impressive seeing how this would completely slaughter a FX 9590 in all workloads, since even a 65W Broadwell already does so.

AMD's single-threaded performance is just too shit nowadays, they really need Zen to stay competitive and even then I kind of doubt they'd reach Broadwell level IPC.

>>52421677
>It was built with high clocks in mind.
I fear to question why..
What workload would need such high single threaded performance and excessive clocks that a regular cluster or more well-rounded Xeon chips cannot handle.

>>52421636

This.

It's almost like they developed a chip for the high end consumers, but it turned out to be way too expensive to market for them, so they just decided to start selling it on the server market, with most likely an insane price tag.

>>52421687
This was the case with Haswell too though, if you look around there are 1600 V3 chips ranging up to 14-core with the 1691 V3, problem is, you can't actually buy any of the ones over 8 cores over most channels.

There was also a high-clocked 2600 V3 chip that was developed for a specific water-cooled server, the part number exists, but again it was a one off thing. It's common.

>>52421751
Common though it may appear. I find it somewhat unsettling, and I find myself incapable of explaining why.

>>52421734
The dual-socket chips are not based on the same dies as consumer parts. If it were numbered 1602 instead of 2602 that would be plausible.

>>52421790
>Be Google/Amazon/MS
>Want a custom CPU
>Go to Intel
>Pay lots of money and sign NDAs

>>52421725
idk but it looks like a gamers dream. high single thread performance for games and high core count for multitasking. maybe an Intel engineer was mad his personal binned 5960x couldn't hit 5ghz.

>>52421840
I don't know much about CPU design but I was thinking about this and was going to ask.

>>52421825
The customer may sign the NDA, but it does not appear Intel have any part in not disclosing any information.
As an analogy, and a somewhat decent one I think, would be to look at bespoke-software. When was the last time you had seen a company hired to produce software for a specific purpose, only then to release it to the general public, or even give an inkling of information of it away. It's simply not needed unless the product was developed with commercial intention.

>>52421825
Right, but the question remains; what kind of enterprise-level software benefits more from high clocks than core scale?

This is a 2P chip, so we're looking at 8 cores at 5Ghz per unit. One could also have 44 cores with the high-end Broadwell Xeon, just for reference.

This won't be less than $1500, most of Intel's high clocked, "low core" Xeons have deranged prices.

Whatever the case might be with this experiment of a chip, I really hope they release something like this for the consumers, with at least somewhat sensible price.
This is the CPU we should have gotten instead of the skylake CPUs they have so far pushed onto the market.

>>52421928
Nvidia or AMD, that how they test GPUs architectures.

>>52421943
Funny thing is that it's built on older tech than Skylake, although seeing the math problem, it was probably a good thing. Intel dodged a bullet there.

>5.1 billion cycles a second
dear god

the mega/gigahertz race is back on!

>>52421943
And what use would the everyday consumer have for such a chip?

Nothing that current Xeons do not fulfil or exceed.

To own one of these personally is to do so only out of vanity as opposed to intelligence. Should you do any work that can make proper use of a Xeon, you would never have even thought to post your post as you would realise the idiocy contained within. There is little a Single Xeon cannot do, should you come across something, you throw another Xeon at it. And continue so on, and so forth. Because the software Xeon workstation users run are specifically designed to make use of cores very well, not single threads.

>>52421061
AMD seem to be over and bankrupt every day

>>52421936
This chip will retail for $179.99

>>52422018
Great, I'll buy 3.

same

>>52421061
>Xeon
>165W TDP

Nobody will want to use this.

>>52422008
Just wait for their sandy bridge competitor in Q4 2016

>>52421992
I remember reading somewhere that FinFET is not great for high clock speeds and that the performance between each batch is inherently tighter, meaning less overclockable chips in general.
Perhaps Intels knows this and are wagging their 5GHz dong out the window to an AMD that's put in the cone of shame after their jump to FinFET

>>52422038
An overclocked 5820k is already 220w tdp. 165w is nothing.

>>52422038
That's why they won't be selling it to you, well, unless you call in directly and inquire about it. If Intel is nice enough to post SKUs of every chip they've ever made, there's no reason for them to be tight-lipped about it.

>>52422058
But that's a desktop CPU. This is an OEM server CPU, and nobody wants to run a 165W CPU in their servers, this is why nobody uses AMD's Opteron line of CPUs.

For the enduser it doesn't matter because they won't be able to buy it and they won't get the mainboard with the right socket

>>52422117
>This is an OEM server CPU, and nobody wants to run a 165W CPU in their servers
You're full of shit.

Whoever doesn't want this chip in their sever reply to me and I'll deal with you personally.

>>52422051
It's been a long time since Intel switched over to FF, they probably have all the bugs worked out by now. Xeons also pass tighter QC, so you're more likely to get a nice chip overall.

>>52422138
It's 2011-3 socket, just like Haswell-E aka X99. Broadwell is backwards compatible.

That's nice and all but I can overclock a skylake chip to 4.9.

>>52421725
>What workload would need such high single threaded performance and excessive clocks

I know right, who'll ever need more than a gig of storage. That amount of space is just insane.
Don't get me started on the "need" for anything more than 256MB of RAM. What could you possibly use it for?

>>52422151
Datacenters have thousands of CPUs, it takes money to pay for electricity and cooling. Higher clocks doesn't mean high performance per watt for all workloads.

>>52421061
Can someone in this thread please enlighten me. What is the point of these new "Xeon" processors that don't support ECC and have high clock speeds. The whole purpose of a Xeon originally was for server use and multiple CPU configurations and to be able to endure heavy work loads and still deliver 24/7 uptime. These things were supposed to have lower clock speeds so they delivered stability and low heat output under full load.

File: Here+s+a+more+derpy+version+for+you+_02ffbf769fc330664eeddd5ae75fed5b[1].jpg (4KB, 200x194px) Image search: [Google]

4KB, 200x194px

>>52422221
>Taking it out of context by removing half of the sentence.

Nice job!
Have a sticker!

>>52422117
More suited to be a professional workstation CPU rather than server.

>>52422251
Xeon's chipsets do support other features that are only useful in a high-workload or server environment. The CPU themselves are practically similar to the regular Core series.

File: mcafee.jpg (18KB, 619x350px) Image search: [Google]

18KB, 619x350px

So it's a crippled Xeon processor? It's going to use more energy and expell more compared to normal Xeons clocked at 2-3GHz.

>>52422284
Alright, I can understand that.

>>52422263
Obviously no one should ever use newer technology in favor of older hardware.

>>52422251
Many workstation tasks, like pro rendering machines for mid-range workshops or complex number crunching that GPUs can't handle.

>>52422196
If this is a server/ws chip, it's designed to run 24/7 under full load at those frequencies and not produce a single error, I highly doubt your OC is nearly as stable.

>>52422221
We're not talking about RAM though. It's easy to justify even 2TB of RAM for oil and gas surveillance.

>>52422251
Where does it say it doesn't support ECC? Besides that they feature pass-through hardware virtualization.

>>52422284
See (>>52421999)
>There is little a Single Xeon cannot do, should you come across something, you throw another Xeon at it. And continue so on, and so forth. Because the software Xeon workstation users run are specifically designed to make use of cores very well, not single threads.

You will never come across a piece of professional software now that is not multi-threaded.
Single threaded performance does matter, I will not argue, but that offered by current Xeons is acceptable. The more important factor here is core count.

The professional industry realised a long time ago that parallel computing is the way to go.

You can have on really strong dude digging a big hole, or you can have 4 moderately strong people digging a hole of identical size.
The 4 are going to finish sooner. It makes sense.
Division is labour is no new idea, it just took a long time for the computing world to catch on.

>>52422333
>We're not talking about RAM though.

I know, I was being facetious about the progression of computer hardware in general. The question "why would anyone even need higher clock speeds" has been answered with every technological breakthrough, much in the same way that increasing storage and RAM have as well.

>>52422368
OLTP workloads are single threaded, not everything in a SQL chain is multi-threaded either, basically a lot of webserver workloads are of a mixed variety.

>>52422368
Multithreading is a must concept but not every workload can take advantage of each added core after a threshold, you hit diminishing returns for various reasons from algorithmic constraints to interthread comm. overhead. The workloads that aren't limited by those (like embarrassingly parallel algos) use GPU computing rather than CPUs nowadays.

>>52421532
32nm PD-SOI
vs
14nm Trigate FinFET

>>52422762
nm is mostly a meme, architecture is the most important thing

>>52422368
>The professional industry realised a long time ago that parallel computing is the way to go
We moved from PATA (parallel) to SATA (single)

Check mate. Go back to reddit you dumbfuck; obviously know nothing about tech.

File: 1451036125371.png (214KB, 600x732px) Image search: [Google]

214KB, 600x732px

>5.1
gotta go fast

What I don't get is,
you people here all like to complain about clickbait articles.
Yet clickbait threads are the ones with the most replies.

>>52422895
>look how much of a tech illiterate retard I am
Thats cute.