Goty wrote:Ouch @ 400W TDP for the 56 core Cascade Lake-AP part. The spec bump from the original 48 core announcement is kind of funny, but not entirely unexpected given the assessment from some OEMs that Cascade Lake-AP wasn't going to be a compelling halo part when compared to Rome. It might end up faster than the top Rome SKU, but at probably twice the TDP, is anyone really going to care?
Anandtech wrote:While these new CPUs do not use a new microarchitecture compared to the first generation Skylake-based Xeon Scalable processors, Intel surprised most of the press at its Tech Day with the sheer number of improvements in other areas of Cascade Lake. Not only are there more hardware mitigations against Spectre and Meltdown than we expected, but we have Optane DC Persistent Memory support. The high-volume processors get a performance boost by having up to 25% extra cores, and every processor gets double the memory support (and faster memory, too). Using the latest manufacturing technologies allows for frequency improvements, which when combined with new AVX-512 modes shows some drastic increases in machine learning performance for those that can use them.
Anandtech wrote:Intel will be offering three parts, and where previously the company stated it would offer up to 48 cores in a single package, it will now offer up to 56, with the top TDP up to 400W. These processors are BGA only, and will only be sold as fundamental Intel server designs by Intel through the OEMs. The OEMs can optimize the design as they see fit, such as offering four blades in a 2U design or going between air and liquid cooling, however the motherboard/CPU configurations will be fixed by Intel.
chuckula wrote:As for Goty's usual ignorance, there are plenty of HPC customers out there for whom a 400 watt socket (not really different from a high end GPU) with 56 cores of AVX-512 is a very exciting product. Don't think for one second that even a 150 watt Epyc 2 with 64 cores (and I'm being overly generous to AMD) will win a power efficiency benchmark on the HPC and inferencing workloads that these parts will tackle. Not to mention that if a Rome processor can even boost its cores up to the 2.6 GHz base speed of the highest end AP chips it sure won't be running in your laptop either.
2.) Personal attacks are frowned upon. If you have to resort to such behavior, perhaps you belong someplace else. It is understandable that some discussions can get heated but try to maintain civility. If you cannot, you may be subject to censure or other disciplinary measures.
Goty wrote:I'm sure there are HPC customers that value performance over all else, but performance per watt is still a very important metric in that space as well.
Goty wrote:Again, I seriously doubt Intel is going to be twice as fast as AMD for twice the power, even in AVX-512 workloads
Glorious wrote:Goty wrote:I'm sure there are HPC customers that value performance over all else, but performance per watt is still a very important metric in that space as well.
He's not saying otherwise, just that there are customers who welcome this sort of thing.
Because that's true.Goty wrote:Again, I seriously doubt Intel is going to be twice as fast as AMD for twice the power, even in AVX-512 workloads
Oh, *YOU* have made your own metric of relevancy here?
To quote another poster here: "is anyone really going to care?"
Goty wrote:Just trying to apply the same scrutiny to Intel's parts that AMD's seem to receive. I apologize if that's uncomfortable.
Glorious wrote:Goty wrote:Just trying to apply the same scrutiny to Intel's parts that AMD's seem to receive. I apologize if that's uncomfortable.
It's not "scrutiny" to ignorantly declare that no one is interested in a SKU because it doesn't match your pre-conceived notions of market suitability.
Especially when those notions are *OPENLY* about "FAIR PLAY FOR AMD!" and the even more asinine "EVERYONE ELSE HAS A PROBLEM--- NOT ME!"
Yes. I'm uncomfortable. This is a cringe-post.
Goty wrote:Performance per watt is hardly my "own metric of relevancy," and the question of absolute performance is still up in the air in a variety of workloads where Intel has so far maintained a performance and value advantage
Goty wrote:I don't believe I declared anything. Maybe you didn't read my post very carefully?
Goty wrote:is anyone really going to care?
Goty wrote:I suspect it's simply your own biases causing you to read far too much into my post.
Goty wrote:Just trying to apply the same scrutiny to Intel's parts that AMD's seem to receive.
Goty wrote:Do you care to contribute to the discussion or would you rather just continue your ad hominem attacks?
Glorious wrote:Goty wrote:I suspect it's simply your own biases causing you to read far too much into my post.
You openly said that your agenda here is to level things in regards to AMD, implicitly claiming the current "scrutiny" here is unfair:
Let me quote you:Goty wrote:Just trying to apply the same scrutiny to Intel's parts that AMD's seem to receive.
So what bias do *I* have?Goty wrote:Do you care to contribute to the discussion or would you rather just continue your ad hominem attacks?
I just told you to not use rhetorical terms you don't understand.
What did I say that was ad hominem?
Ad hominem (Latin for "to the person"), short for argumentum ad hominem, is a fallacious argumentative strategy whereby genuine discussion of the topic at hand is avoided by instead attacking the character, motive, or other attribute of the person making the argument, or persons associated with the argument, rather than attacking the substance of the argument itself.
Goty wrote:If you can't get from point A to point B from that, I'm not sure how to proceed here.
chuckula wrote:A very big launch even if the HBM equipped Icelake Xeons will make more headlines when they launch next year.
Convert wrote:You really aren't benefiting the community by going to the extremes that you do.
Waco wrote:chuckula wrote:A very big launch even if the HBM equipped Icelake Xeons will make more headlines when they launch next year.
Got a roadmap for that?
chuckula wrote:This guy who bought 125 Petaflops [that's double precision petaflops, or as I call them: The awesome kind] worth of Knights Landing hardware does: https://www.hpcwire.com/2019/03/13/oil- ... 00-wafers/
Waco wrote:chuckula wrote:This guy who bought 125 Petaflops [that's double precision petaflops, or as I call them: The awesome kind] worth of Knights Landing hardware does: https://www.hpcwire.com/2019/03/13/oil- ... 00-wafers/
Forgive me for not trusting someone who claims US datacenters run on 120 volts as a reliable source.
dragontamer5788 wrote:I still remember the first time I tried to plug a computer inside of the server-room by myself. I was very confused. The sysadmin was out on vacation and I was just... trying to make due without him at that time. Eventually he introduced me to the concepts of locking-plugs and such. Unfortunately, I had to purchase some equipment while he was gone... and ended up buying an (unnecessary) transformer. Ah well, n00b mistakes. The SysAdmin more or less laughed at me when he came back.
dragontamer5788 wrote:Sounds more like this guy is Australian and doesn't understand the US market. Seems like a believable mistake if all he's got is the internet as his reference point.
dragontamer5788 wrote:I don't work at a datacenter, but L6-15 250V is what the server-room uses at my office (USA). The higher-efficiency of high-voltage would lead me to assume that most datacenter racks actually use 250V outlets. Offices typically use 120V (standard 5-15 American socket), but almost all computers will work with a 250V outlet so its not a big deal actually..
dragontamer5788 wrote:These are the sorts of things you only really learn when you walk into a room and try to plug a computer into the wall by yourself. For whatever reason, SysAdmins and electricians don't really talk about these details unless you ask them about it.
dragontamer5788 wrote:but almost all computers will work with a 250V outlet so its not a big deal actually..
Glorious wrote:Right, and Waco is pointing out that if the internet is indeed his reference-point, he's not a trustworthy or useful source.
Waco wrote:Exactly. Why would this guy who bought a lot of KNLs be any more of a valid source than any other Joe Blow on the Internet? Just buying a lot of CPUs and not understanding the market doesn't make you trustworthy (at least not in my mind).
I can't find anything on recent public roadmaps that says anything about HBM on Intel chips. Has anyone else seen otherwise? The only mentions of it in any form are for Phi, and that line is dead.