Andy Grave: LOL record setting performance...

Intel hasn't released Skylake server numbers yet so don't get ahead of yourself junior. And if you were a truly smart investor wouldn't you have jumped on Apple in 2000 not some cheesy AMD play that is largely still hype and stagnated for a decade. Oh wait you are still patting yourself on the back for IBMs "process technology lead". AMD even penalized Intel for 20% because Intel has a vastly superior compiler on those ridiculous competitive slides. Their benchmarking misdeeds continue to this day and the lies will be exposed. Just like yours always are. Skylake is coming fast and it is going to kick ass on frequency, IPC, and since Intel is the incumbent, they already have all the design wins. Even if you buy AMDs bluster at face value, they have a long slog ahead because ENTERPRISE IS HARD(TM). And 2GHZ is not impressive at all. Broadwell did that and they aren't competing against that by the time this thing ships. Q4???? Lot's of problems are being reported. Maybe you should wait for AMD to ship some of this crap and actually sell it and get it fairly benchmarked. This is not some tip toe through the daisies. But you're no daisy...you're not a daisy at all.

TCE: You are probably right...

Server qualification is a lot tougher than an enthusiast level chip so they wouldn't be able to do what they did with Skylake X.

Elmer: They haven't commented on that part at all...

Wffctech is sort of a rumour site and they are the source for all this. Intel has put out a roadmap with a 28 core part that seems to be pretty well documented and looks like it will be released. The 32 core Xeon part is something Wffctech discovered and says is a possibility but it may not actually be released. If they release it I would expect it to be a monolithic die but Intel hasn't even acknowledged its existence.

TCE: AMD is well aware of the thermal properties...

And they have the 32 core Epyc version for server which is apparently clocked much lower but has twice the number of cpu die. It is very likely that it is an engineering sample. And the bios is probably still being finished. So it is hard to say why it scales badly. But I also noticed that low clocked Intel high core count Xeons didn't scale as expected either. A Ryzen 1800X gets a score of 21439. Theoretically a 16 core processor should get into the 30s you would think. And yet none of the Xeons break 30k. Geekbench does not appear to scale very well above 8 cores and frequency seems to be pretty significant in the final score so it probably isn't the greatest benchmark for high core count cpus. Still it does scale and the high core count Xeons do get in to the high twenties. It is a suprising result as Threadripper is apparently clocked at 3.4GHZ and it should have done better. I guess we know why it hasn't launched yet :)

Threadripper appears on Geekbench...

Ooops.

https://browser.primatelabs.com/v4/cpu/search?q=1950x

Threadpooper 1950x 16 cores 3.394 GHZ:

4216 singe core 24723 multicore.

Intel Core i7-6950X
3.0 GHz (10 cores)

4452 Single core 29155 multicore.


https://browser.primatelabs.com/processor-benchmarks

Piledriver revisited??????? LOL

borusa: we will see soon enough...

And it is very likely that a server part with 32 cores is not going to be terribly high frequency. If it is and Wccftech is wrong, so be it. But it fits what we know about existing many core chips. And we know that AMD's process is behind Intel's. And housing 4 8 core chips together is going to impose significant thermal challenges in an environment where you want 99.999 reliability. Wccftech can be wrong but they have also been very right. And that data fits what I expected. Give me an argument based on facts that you are right instead of NUH UH.

borusa: It is good news...

But you are of too lost in your AMD fanboyness.

Threadripper is just 2 8 core dies. It is gigantic expensive package. It is already looking like it may clock significantly lower than Ryzen(see my post on UnEpych from yesterday). Honestly TR looks kind of boring. The overclocking is going to be the same or worse than Ryzen. The IPC gains over broadwell look really good and the newly redesigned cache is a significant improvement.

Grrr: If you read the article...

It is pretty much an engineering sample and the bios is still being finalized. Ryzen had far more platform flaws at its launch. And the price tag is beside the point. This is an excessive platform. If you were smart you'd just get a Core i7 4 core for 95% percent of the population. Or even a Core i5. And power consumption and temps are beside the point on this type of platform. It has always been a power hog to achieve performance. And on performance it delivers significant gains over broadwell. So I dunno what your complaining about :)

borusa: it is about marketing...

And currently Intel has the tastiest triple cheeseburger and AMD is still giving shit away for free.

Core i9 reviews have leaked...

The 10 core part basically destroys Ryzen.

The world’s first Core i9 processor, the 7900X, is ultimately a 10-core powerhouse offering excellent IPC performance and outstanding multi-core prowess in a single $999 chip armed with plenty of overclocking headroom. Said ingredients make it an automatic choice for power users seeking the ultimate PC experience, and the chip’s benchmarking potential makes for a fitting debut of the Core i9 brand.


http://wccftech.com/intel-core-i9-7900x-review-cpu-gaming-oc-performance/

WCCFTECH has leaked Xeon Platinum Cinebench numbers...

I know that server workloads are significantly different and Cinebench is very workstation oriented. But the remarkable thing is that the 2.7GHZ clocked 24 core platinum Xeon beats the 2.1GHZ 28 core platinum Xeon chip. But there is a pretty big distance on all the Skylakes above the older Xeons. I guess it all depends on if your server software needs the extra cores more than the frequency. But interesting stuff. Given the 2.2GHZ base clocks on the best 32 core Epych chips and Intel's IPC advantage, it looks like Intel is going to maintain the overall performance lead on most workloads. And if this is true. WOW:

The Intel Xeon Platinum 8180 is the part believed to have 28 cores and is clocked at an astounding 2.5 GHz core clock. This is very impressive for a processor with this amount of cores. It also has an L3 cache of 38.5 MB as well as a TDP of just 205 watts. The architectural improvements that set Skylake apart from Broadwell can be found throughout this lineup so it would be fair to call it Intel’s Skylake rollout to the server market.

This is more speculative but also interesting: So far Intel has four families of Xeon planned: Bronze, Silver, Gold and Platinum. The Platinum family will be the 8000 series, gold will be 6000 and 500, silver will be 4000 and bronze will be the 3000 series. Core scaling is thought to start from 10 cores all the way up to 28 cores (or even 32 cores if the processor we saw earlier ends up being part of this lineup.




http://wccftech.com/intel-28-core-xeon-platinum-8176-8168-xeon-gold-6161-6142-benchmarks/

borusa: one exclusive...

Does not change the fact that most dell's business and consumer lineup is Intel based.

So to all who were expecting 3.6 GHZ EPYC speeds...

The base is between 2.0 and 2.5 GHZ. Actually I think Intel is gonna be just fine in servers. Sure there are some niche areas where AMD can take some MSS but hardly the catastrophe the AMD pumpers are claiming. They will probably get better than this with Threadripper but clearly AMD has not invented the magic chip technology that can clock high count cpus better than Intel and defy physics like some have claimed. Anyone with server expertise want to comment on the specrate Int benchmarks? Clearly this is AMD benchmarketing so there is some misleading info. AMD is busy comparing last gen 22 cores to unreleased 32 core parts so Intel's next gen architecture should be A LOT more competitive. And by there own measurements, 22 core parts are a small part of the market apparently, so it is unclear to me that AMD has a home run here. Additionally, Alienware apparently has an exclusive with threadripper. Does that suggest production bottlenecks? Because if not AMD just pissed off HP, Falcon Northwest and a lot of indy game pc houses for nothing.




http://wccftech.com/amd-epyc-7000-series-server-cpu-family-specifications-price-performance-leak/

Andy noobsauce: Enterprise is hard...

And AMD will have to execute, which they haven't done in the past. But your posts are always useless. So keep flailing...

TCE: Actually I thought you were boring...

My mistake. But Linus Tech Tips is more deeply embedded at Intel than you might think. He has had exclusives like the Optane launch that I previously linked. AND he has contract work with Intel Retail Edge which is Intel's retail training program. He has done plenty of work for Intel. Make no mistakes...

Intel unveils DC P4501 SSD series

Intel has introduced its new datacenter-oriented P4501 SSD series based on its second-generation NVMe controller, 3D TLC NAND and either M.2 or 2.5-inch U.2 form-factors.

Thanks to Intel's own second-generation NVMe controller, the new Intel DC P4501 SSD series should be faster than the DC P4500 and the DC P4600, despite being based on 3D TLC NAND flash memory, at least according to Anandtech.com.

The new DC P4501 SSD series will be available in 500GB, 1TB, 2TB and 4TB (U.2 form-factor only) capacities and offer sequential read and write performance of up to 3200MB/s and 900MB/s. The random 4K read and write performance peaks at 360k and 46k IOPS. The TDP is variable and can be tuned at 8W, 10W or 12.5W on 2.5-inch U.2 form-factor version or at 6W and 8.25W for the M.2 version.

http://www.fudzilla.com/news/memory/43833-intel-unveils-ssd-dc-p4501-ssd-series

Boring: You're not an insider :)

Thanks to Intel for sponsoring this exclusive look at their new Optane technology! #sponsored

Borusa: Linus Torvolds or Linus Tech tips?

Not that impressed with Linus Tech tips but it tells me a lot that you are. Fun fact about Linus though: he worked for Intel at one point LOL.

14nm++ is making a big difference in clocks. And you obviously do not understand what the Xeon socket brings to the table. But just leave it be and wait for the reviews. You can't possibly judge Core i9s because they aren't out and we know basically core counts and thats it. And yes TR has a gigantic freaking package(that adds significant cost so bye bye margin). It probably helps dissapate heat. But lets wait for the reviews. We already know your a fan boy. But give it a rest until you have some actual benchmarks and independent tests.

Borusa: I have seen the package, I know they are separated...

You shouldn't make assumptions because you are definitely no thermal engineer. It will be interesting to see the reviews. Even if TR does overclock similarly to Ryzen, 4GHZ is not terribly exciting. From an enthusiast point of view, X299 and SKYLAKEX is far more interesting and may offer some suprises. That being said, personally, I view all of this as somewhat mad. My 3 year old Core i7 4790K is still very very fast. I normally upgrade by now but this system still is pretty great. The only reason to upgrade at this point is if it dies.

borusa: you are speculating again...

We won't know the thermal characteristics until it is launched. Then we will find out the real story when reviewers see how well it overclocks. As of today all we know is that it runs at 3.6GHZ in regular mode and can turbo one core or a few cores to 3.9GHZ on the best parts. Basically the same as Ryzen. So it is either the same or worse overclocking. You are so hysterical. Slap yourself please :)

bored: what a facile and silly post.

Same old BORING. Threadripper does have more PCIE lanes but you should wait for actual reviews before jumping to conclusions on anything other than price. Quad video card solutions are falling out of favor and dual cards aren't all that popular. Maybe someone wants to build a massive raid. But Intel was the one demonstrating THAT not AMD. We already know that the Skylake architecture has better IPC than Ryzen and Broadwell. We know that typically Intel is getting more frequency from 14nm++ and that AMD is stuck at 3.9 turbo(does putting two chips on the same package impose more thermal limits????). The one year old broadwell architecture already beats Ryzen quite a bit. Do you suppose that changes when Intel gets higher frequency and a new architecture and a redesigned cache and more cores?

Joseph: When your article repeatedly cites AMDs marketing slide deck...

And you think Intel's response was embarrassing. You might assume that an AMD zealot like Andy Grave or Borusa or even Chuckie D wrote that article. That was an extremely biased and not well thought out article. There are real improvements in Intel's lineup: the cache changes, frequency from tuning the 14nm process, the skylake architecture to name a few. Skylake-X actually looks pretty good. AMD is in the same position it always is: give things away for free because it's architecture is inferior, it's brand is not respected, and it's process technology is really inferior. The best advantage they have is that they can link two 8 core chips together. Which is relevant to a tiny part of the market. And as to all the talk about HP and Dell. Well, they already use AMD chips. What remains to be seen is the margin improvement and marketshare at those accounts. But Intel has many levers to pull to combat that(wink, wink borusa: Intel Inside is alive and well.). And there are the deep relationships that have been there for decades. Let's not forget that Intel has rock solid compatibility and reliability. AMD's platform is less shaky than at launch but it is not in the same league as Intel.

mmoy: one of the few usage cases where it makes sense :)

Starting at $4999 though. Wow.

Borusa: *facepalm*

Get over it already. It's been a decade.

borusa: Dell was already using AMD cpus...

Dell has not been exclusively Intel for a long time. Keep up. It is a higher end design for the first time in a long time. Let's see how it sells.

borusa: lol

AMD does NOT have a better cost structure. They have to pay GF cutting edge prices for its 14nm. AMD is NOT creating a market. Intel created the HEDT market. AMD is trying to get back in to that market. Intel has on the shelf Xeons to respond to AMDs little 16 core stunt and it did. But make no mistake. Anything above 4 core is the veritable triple cheese burger. It's all for show and relatively few people order it. And neither Intel or AMD are going to sell millions of 16 or 18 core chips there. It is all about marketing. And to that extent it does signal AMDs return to HEDT. But lets remember threadripper is not released. And that market is too small to make anything but a statement. Now if AMD manages to make a dent in servers, that has some serious volume. But that will take a lot of time and effort. And there are many barriers to entry there. In the volume markets where all the serious profits are made in retail pcs and enterprise, little has changed.

borusa: you have no idea what you are talking about...

A premium brand NEVER drops prices by 23% with a successful product just months after launch. Threadripper should be additive and require NO price adjustment. You would know that if you ever worked with a premium brand. It is a REALLY bad sign that AMD is not competing well. But then again, the press has really overhyped Ryzen and its multiple generations behind performance. I would say that Intel's recent announcment may have dried up demand. No way is this making way for threadripper. Here is how you tell: did Intel lower prices? Nope.

AMD lowering prices 23% on Ryzen

Same old 2nd rate brand. Compete on price. Rinse. Repeat.


http://wccftech.com/amd-ryzen-7-prices-drop-ahead-of-threadripper-launch/

Unkwn: While you were ignoring the PC market...

Intel established and dominated all in ones with a variety of better designs than the AMD one for the last 5 years. One design win doesn't really change that. And oh yeah, there are AMD AIO designs at Best Buy with the driver garbage. Dells appears to be a build to order deal. Lets see how many they get into retail.

TCE: Excellent. Thanks for the response!

I didn't know about the 3647 Socket. So are all the pins there for memory channels? It will be interesting to see what frequency the AMD chips are able to achieve. In theory the monolithic chips should have a significant advantage. When AMD does its competitive demos it often puts its thumb on the scales. It will interesting to see an independent review. Probably from Johan over at Anandtech. Oh and AMDs hypocrisy knows no bounds. They mocked the speed racer design of the Pentium 4 and then dove head in with the failed dozer lineup to you guessed it: a speed racer strategy after Intel went back to the braniac style cpu. And here there are back again with a braniac style cpu. Never the innovator, always the immitator.

Question for the engineers: what are the risks of doing a socket with 4094 pins

Intel Xeons have 2066. AMD finally has joined the 21st century and actually has abandoned the pin socket but what is the net net here? If I believe the hype, 2 Zen cores are going to get 155W TDP. How can they get to high GHZ at that rate???? Maybe a turboboost on one core but all 16 cores running together will be running how much slower? Chuckles the clown has already decreed that Threadripper will destroy all Intel chips but I am not so certain because Charlie lives in his own ego and AMD is god ecosystem. They will have to pay the physics piper for 2 big chips fairly close together. That means lower frequency if TDP is not 2x95W to me. Or am I missing something here? I noticed they have not released many specs. As a marketer, that tells me they are hiding something.

You also have 4 or 2 cpus on package with all the required wires to connect on Epych and Threadripper. That is a massive amount of pins. The chip is GIGANTIC. What the are pros and cons of linking chips this way? Will reliability be worse by any significant margin? Feel free to speculate :)

fastpath: Server is a bit more conservative...

So given the nature of things AMD will be able to brag about its 4 chip monstrosity. But there would seem to be possible reliability concerns and I do wonder if the infinity fabric really is 100% scaling. Time will tell...

Bored: single core IS important...

This about marketing.

18 core madness...

Intel did a decent job not leaking this. The L3 cache changes and Turboboost 3 features are also interesting.

The Core i9-7980XE sits atop of the stack as the halo part, looking down on all those beneath it. Like an unruly dictator, it gives nothing away: all we have is the core count at eighteen, the fact that it will sit in the LGA2066 socket, and the tray price at a rather cool $1999 (~$2099 retail). When this processor will hit the market, no-one really knows at this point. I suspect even Intel doesn’t know.

www.anandtech.com/show/11464/intel-announces-skylakex-bringing-18core-hcc-silicon-to-consumers-for-1999

Intel Cascade Lake “Next-Generation Xeon Scalable Platform” Family Demoed – 59% Faster, Utilizes Up To 6 TB of Optane DIMMs, Launching in 2018

"Intel Cascade Lake-SP is Skylake-SP Refresh Based on 14nm+ Node – Up To 6 TB Optane DIMM Support

The Intel Cascade Lake family of processors, also to be known as Cascade Lake-SP will be a refresh based on Skylake-SP. The new chips will keep intact the same architecture but based on a slightly new 14nm+ node that will increase power efficiency, clock speeds and we can also see a core count bump."


http://wccftech.com/intel-cascade-lake-xeon-scalable-platform-optane-dimm-2018/

borusa: it's ok, we know most of it is over your head

Get out your pom poms and go cheer your little heart out for AMD. And given your posting history, vacuous is a word you should avoid. Oh my god the irony lol.

fpg: you keep suggesting that Ryzen yields are just super...

And Elmer makes a great point that NOONE is going to be getting super yields with a big 8 core chip like Ryzen. I don't know much about yields but I did see the banner headlines on Intel's internal website when Intel got a perfect wafer and I know they are WAY better than GF. I think right now it is VERY complicated to make 14nm and smaller chips. Yields are going to be a problem for the entire industry if you haven't figured that out yet. And Elmer laid out why. And foundries are worse off than Intel. Because they are running generic inferior processes and they can't customize like Intel has done so effectively with 14nm+. And AMD's lack of profits so far is proof.

Intel plans to integrate Thunderbolt into future CPUs

Partially because Thunderbolt 3 relies on a discrete controller chip at the moment, however, manufacturers have reserved the feature for premium notebooks and motherboards—if it's implemented at all. This morning, Intel announced plans to spur adoption of Thunderbolt by integrating support for the protocol directly into its future CPUs and making the Thunderbolt protocol specification available under a "nonexclusive, royalty-free license" in 2018. That means third-party peripheral manufacturers can start developing their own Thunderbolt-compatible controllers, as well.

http://techreport.com/news/31959/intel-plans-to-integrate-thunderbolt-into-future-cpus

Oh boring: Intel never gave Dell anything for free...

But your vast misunderstanding of the Intel Inside program is well known. I wouldn't bother to explain it to you. It is above your paygrade as Apple pie maker at McDonalds.

New Surface Pro is fairly impressive...

Oh wait AMD didn't win this socket? SHOCKING! And gambling is taking place in Rick's Cafe Americain...

"This isn't the Surface Pro 5. Taking a page from Apple's MacBook naming scheme, the latest version of the class-leading detachable-hybrid 2-in-1 tablet is simply named Microsoft Surface Pro ($799). It has the same dimensions as the Surface Pro 4, and shares peripheral compatibility with that tablet and the Surface Pro 3. What's new? It has a set of Kaby Lake processors, potentially longer battery life, and a revamped Type Cover and Surface Pen."





http://www.pcmag.com/review/353810/microsoft-surface-pro-2017