AMD's Future Chips & SoC's: News, Info & Rumours.

[gamerk316]

Intel cofirms to tomshardware that Mesh is having a negative effect on skylake results:
http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-2.html

Not shocking. As core count goes up, inter-core latency is going to become more and more of a problem.

 

[juanrga]

Intel cofirms to tomshardware that Mesh is having a negative effect on skylake results:
http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-2.html

It doesn't say it has a negative effect overall, but what it affects a "handful of applications", whereas "the overall performance of the platform" increases

This is the same that I said (but with other words) when I explained that different workloads can have different requirements and that optimizing for some can introduce penalties on others. Of course, this is not anything specific to Skylake-X, the same happens with any other microarchitecture, including the ones from AMD. I mentioned some examples before.

Moreover, I was discussing the DigiCortex 1.20 bench, and the penalty here here seems to be traced to the change in the cache subsystem, not to the change on interconnect.

 

[jaymc]

Intel cofirms to tomshardware that Mesh is having a negative effect on skylake results:
http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-2.html

It doesn't say it has a negative effect overall, but what it affects a "handful of applications", whereas "the overall performance of the platform" increases

This is the same that I said (but with other words) when I explained that different workloads can have different requirements and that optimizing for some can introduce penalties on others. Of course, this is not anything specific to Skylake-X, the same happens with any other microarchitecture, including the ones from AMD. I mentioned some examples before.

Moreover, I was discussing the DigiCortex 1.20 bench, and the penalty here here seems to be traced to the change in the cache subsystem, not to the change on interconnect.

Well I was discussing the Mesh and the L3 cache.. and it appears that both are causing problems as well as heat problems major issues with cooling..

Hell basically every web site on the Internet (across the board including TomsHardware) and all the YouTubers are giving terrible reviews of the I9...

I just can't understand, how can you continuously defend such a product.. It doesn't make sense ?

It was clearly rushed ! An Intel themselves are admitting that Mesh is causing negative results.. Couple that with L3 cache.. High power draw and crazy high temps.. It's a Terrible product end of story ! An it seems most of the reviewers agree on this across the board (un-biased ones anyway that is)

 

[gamerk316]

Intel cofirms to tomshardware that Mesh is having a negative effect on skylake results:
http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-2.html

It doesn't say it has a negative effect overall, but what it affects a "handful of applications", whereas "the overall performance of the platform" increases

This is the same that I said (but with other words) when I explained that different workloads can have different requirements and that optimizing for some can introduce penalties on others. Of course, this is not anything specific to Skylake-X, the same happens with any other microarchitecture, including the ones from AMD. I mentioned some examples before.

Moreover, I was discussing the DigiCortex 1.20 bench, and the penalty here here seems to be traced to the change in the cache subsystem, not to the change on interconnect.

Well I was discussing the Mesh and the L3 cache.. and it appears that both are causing problems as well as heat problems major issues with cooling..

Hell basically every web site on the Internet (across the board including TomsHardware) and all the YouTubers are giving terrible reviews of the I9...

I just can't understand, how can you continuously defend such a product.. It doesn't make sense ?

It was clearly rushed ! An Intel themselves are admitting that Mesh is causing negative results.. Couple that with L3 cache.. High power draw and crazy high temps.. It's a Terrible product end of story ! An it seems most of the reviewers agree on this across the board (un-biased ones anyway that is)

Its *bad* because people were expecting something unrealistic. They wanted more cores on the assumption their specific workloads (games mainly) would benefit. They don't; hell, they actually get penalized due to the new mesh bus (which is necessary; a ring bus is non-viable as core count increases due to worst-case latencies).

This was predictable and unsurprising. Hell, I even predicted the outrage.

 

[jaymc]

Intel cofirms to tomshardware that Mesh is having a negative effect on skylake results:
http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-2.html

It doesn't say it has a negative effect overall, but what it affects a "handful of applications", whereas "the overall performance of the platform" increases

This is the same that I said (but with other words) when I explained that different workloads can have different requirements and that optimizing for some can introduce penalties on others. Of course, this is not anything specific to Skylake-X, the same happens with any other microarchitecture, including the ones from AMD. I mentioned some examples before.

Moreover, I was discussing the DigiCortex 1.20 bench, and the penalty here here seems to be traced to the change in the cache subsystem, not to the change on interconnect.

Well I was discussing the Mesh and the L3 cache.. and it appears that both are causing problems as well as heat problems major issues with cooling..

Hell basically every web site on the Internet (across the board including TomsHardware) and all the YouTubers are giving terrible reviews of the I9...

I just can't understand, how can you continuously defend such a product.. It doesn't make sense ?

It was clearly rushed ! An Intel themselves are admitting that Mesh is causing negative results.. Couple that with L3 cache.. High power draw and crazy high temps.. It's a Terrible product end of story ! An it seems most of the reviewers agree on this across the board (un-biased ones anyway that is)

Its *bad* because people were expecting something unrealistic. They wanted more cores on the assumption their specific workloads (games mainly) would benefit. They don't; hell, they actually get penalized due to the new mesh bus (which is necessary; a ring bus is non-viable as core count increases due to worst-case latencies).

This was predictable and unsurprising. Hell, I even predicted the outrage.

To quote our v own Yuka "Boiler Lake" lmfao...

 

[-Fran-]

Its *bad* because people were expecting something unrealistic. They wanted more cores on the assumption their specific workloads (games mainly) would benefit. They don't; hell, they actually get penalized due to the new mesh bus (which is necessary; a ring bus is non-viable as core count increases due to worst-case latencies).

This was predictable and unsurprising. Hell, I even predicted the outrage.

Unrealistic? So the jump to Conroe and then to Nehalem and then to Sandy were unrealistic in terms of expectations? I'm having a hard time putting that into this specific context, just because Intel over promised, but this time did not deliver. The expectations, at least as I read everywhere, weren't even high. Just like with Conroe, Nehalem and even Sandy and they were good jumps in all aspects with new uArch or interesting tweaks to the governing design and additions to the uArch.

So, again, unrealistic? How?

Cheers!

 

[jaymc]

Epyc Wins, Intel Prepares To Fight Dirty:

https://www.youtube.com/watch?v=ctgAzn5Wx8o

 

[-Fran-]

Yeah. Marketing is evil. Intel's marketing defines the word "evil".

Cheers!

 

[goldstone77]

I seen this the other day, and I forgot to post it, same reference from video!

Sizing Up Servers: Intel's Skylake-SP Xeon versus AMD's EPYC 7000 - The Server CPU Battle of the Decade?
by Johan De Gelas & Ian Cutress on July 11, 2017 12:15 PM EST
Closing Thoughts
"First of all, we have to emphasize that we were only able to spend about a week on the AMD server, and about two weeks on the Intel system. With the complexity of both server hardware and especially server software, that is very little time. There is still a lot to test and tune, but the general picture is clear.

We can continue to talk about Intel's excellent mesh topology and AMD strong new Zen architecture, but at the end of the day, the "how" will not matter to infrastructure professionals. Depending on your situation, performance, performance-per-watt, and/or performance-per-dollar are what matters.

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."

Looking Towards the Future
"Looking towards the future, Intel has the better topology to add more cores in future CPU generations. However AMD's newest core is a formidable opponent. Scalar floating point operations are clearly faster on the AMD core, and integer performance is – at the same clock – on par with Intel's best. The dual CCX layout and quad die setup leave quite a bit of performance on the table, so it will be interesting how much AMD has learned from this when they launch the 7 nm "Rome" successor. Their SKU line-up is still very limited.

All in all, it must be said that AMD executed very well and delivered a new server CPU that can offer competitive performance for a lower price point in some key markets. Server customers with non-scalar sparse matrix HPC and Big Data applications should especially take notice."

As for Intel, the company has delivered a very attractive and well scaling product. But some of the technological advances in Skylake-SP are overshadowed by the heavy price tags and somewhat "over the top" market segmentation.
http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade

 

[juanrga]

Epyc Wins, Intel Prepares To Fight Dirty:

https://www.youtube.com/watch?v=ctgAzn5Wx8o

From AdoredTV, a well-known biased source. Moreover, I wonder who is playing dirty on servers:

What's extremely interesting about this announcement vs. AMD's dog & pony show for Epyc is that the only numbers AMD could post were intentionally-fudged SPEC benchmarks based on their own in-house testing of old Xeon parts compared to their own in-house-tuned Epyc scores.

Intel just shows numbers from third party customers who are actually using Skylake in real-world workloads instead of intentionally gimping Epyc in some in-house bakeoff competition.

 

[juanrga]

I seen this the other day, and I forgot to post it, same reference from video!

Sizing Up Servers: Intel's Skylake-SP Xeon versus AMD's EPYC 7000 - The Server CPU Battle of the Decade?
by Johan De Gelas & Ian Cutress on July 11, 2017 12:15 PM EST
Closing Thoughts
"First of all, we have to emphasize that we were only able to spend about a week on the AMD server, and about two weeks on the Intel system. With the complexity of both server hardware and especially server software, that is very little time. There is still a lot to test and tune, but the general picture is clear.

We can continue to talk about Intel's excellent mesh topology and AMD strong new Zen architecture, but at the end of the day, the "how" will not matter to infrastructure professionals. Depending on your situation, performance, performance-per-watt, and/or performance-per-dollar are what matters.

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."

Looking Towards the Future
"Looking towards the future, Intel has the better topology to add more cores in future CPU generations. However AMD's newest core is a formidable opponent. Scalar floating point operations are clearly faster on the AMD core, and integer performance is – at the same clock – on par with Intel's best. The dual CCX layout and quad die setup leave quite a bit of performance on the table, so it will be interesting how much AMD has learned from this when they launch the 7 nm "Rome" successor. Their SKU line-up is still very limited.

All in all, it must be said that AMD executed very well and delivered a new server CPU that can offer competitive performance for a lower price point in some key markets. Server customers with non-scalar sparse matrix HPC and Big Data applications should especially take notice."

As for Intel, the company has delivered a very attractive and well scaling product. But some of the technological advances in Skylake-SP are overshadowed by the heavy price tags and somewhat "over the top" market segmentation.
http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade

As discussed in many places, the review is plain wrong. During their setup the performance of the Broadwell and Skylake Xeons was crippled by 40% and higher amounts

https://www.reddit.com/r/intel/comments/6n2sww/did_anandtech_cripple_xeon_results/

David Kanter thoughts:

Skylake-SP actually scores about 1.38X higher than Naples on GCC. It's 1.58X higher on a per-core basis.

See https://www.spec.org/cpu2006/results/res2017q3/cpu2006-20170627-47389.html and https://www.spec.org/cpu2006/results/res2017q2/cpu2006-20170529-47127.html

So I'm pretty sure that something is horribly wrong in Johan's testing.

Johan got EPYC core to be 30% faster than Skylake core. In reality the Skylake core is 58% faster than EPYC core, as one can see in above links. This means Johan test is crippling the performance of the Skylake core by a huge 2x factor. Indeed "something is horribly wrong in Johan's testing".

From Ricardo B.:

On March 2016's Broadwell-EP review, E5-2699v4 scored 33.3 on 403.gcc.

On the July 2017 Skylake-SP vs EPYC review, E5-2699v4 scores 23.7 on 403.gcc.

 

[8350rocks]

I seen this the other day, and I forgot to post it, same reference from video!

Sizing Up Servers: Intel's Skylake-SP Xeon versus AMD's EPYC 7000 - The Server CPU Battle of the Decade?
by Johan De Gelas & Ian Cutress on July 11, 2017 12:15 PM EST
Closing Thoughts
"First of all, we have to emphasize that we were only able to spend about a week on the AMD server, and about two weeks on the Intel system. With the complexity of both server hardware and especially server software, that is very little time. There is still a lot to test and tune, but the general picture is clear.

We can continue to talk about Intel's excellent mesh topology and AMD strong new Zen architecture, but at the end of the day, the "how" will not matter to infrastructure professionals. Depending on your situation, performance, performance-per-watt, and/or performance-per-dollar are what matters.

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."

Looking Towards the Future
"Looking towards the future, Intel has the better topology to add more cores in future CPU generations. However AMD's newest core is a formidable opponent. Scalar floating point operations are clearly faster on the AMD core, and integer performance is – at the same clock – on par with Intel's best. The dual CCX layout and quad die setup leave quite a bit of performance on the table, so it will be interesting how much AMD has learned from this when they launch the 7 nm "Rome" successor. Their SKU line-up is still very limited.

All in all, it must be said that AMD executed very well and delivered a new server CPU that can offer competitive performance for a lower price point in some key markets. Server customers with non-scalar sparse matrix HPC and Big Data applications should especially take notice."

As for Intel, the company has delivered a very attractive and well scaling product. But some of the technological advances in Skylake-SP are overshadowed by the heavy price tags and somewhat "over the top" market segmentation.
http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade

As discussed in many places, the review is plain wrong. During their setup the performance of the Broadwell and Skylake Xeons was crippled by 40% and higher amounts

https://www.reddit.com/r/intel/comments/6n2sww/did_anandtech_cripple_xeon_results/

David Kanter thoughts:

Skylake-SP actually scores about 1.38X higher than Naples on GCC. It's 1.58X higher on a per-core basis.

See https://www.spec.org/cpu2006/results/res2017q3/cpu2006-20170627-47389.html and https://www.spec.org/cpu2006/results/res2017q2/cpu2006-20170529-47127.html

So I'm pretty sure that something is horribly wrong in Johan's testing.

Johan got EPYC core to be 30% faster than Skylake core. In reality the Skylake core is 58% faster than EPYC core, as one can see in above links. This means Johan test is crippling the performance of the Skylake core by a huge 2x factor. Indeed "something is horribly wrong in Johan's testing".

From Ricardo B.:

On March 2016's Broadwell-EP review, E5-2699v4 scored 33.3 on 403.gcc.

On the July 2017 Skylake-SP vs EPYC review, E5-2699v4 scores 23.7 on 403.gcc.

David Kanter was off by about 15% on baseline Ryzen performance....why should we take what he is saying about EPYC at face value given his underestimation of Ryzen cores already?

 

[juanrga]

Spoiler

I seen this the other day, and I forgot to post it, same reference from video!

Sizing Up Servers: Intel's Skylake-SP Xeon versus AMD's EPYC 7000 - The Server CPU Battle of the Decade?
by Johan De Gelas & Ian Cutress on July 11, 2017 12:15 PM EST
Closing Thoughts
"First of all, we have to emphasize that we were only able to spend about a week on the AMD server, and about two weeks on the Intel system. With the complexity of both server hardware and especially server software, that is very little time. There is still a lot to test and tune, but the general picture is clear.

We can continue to talk about Intel's excellent mesh topology and AMD strong new Zen architecture, but at the end of the day, the "how" will not matter to infrastructure professionals. Depending on your situation, performance, performance-per-watt, and/or performance-per-dollar are what matters.

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."

Looking Towards the Future
"Looking towards the future, Intel has the better topology to add more cores in future CPU generations. However AMD's newest core is a formidable opponent. Scalar floating point operations are clearly faster on the AMD core, and integer performance is – at the same clock – on par with Intel's best. The dual CCX layout and quad die setup leave quite a bit of performance on the table, so it will be interesting how much AMD has learned from this when they launch the 7 nm "Rome" successor. Their SKU line-up is still very limited.

All in all, it must be said that AMD executed very well and delivered a new server CPU that can offer competitive performance for a lower price point in some key markets. Server customers with non-scalar sparse matrix HPC and Big Data applications should especially take notice."

As for Intel, the company has delivered a very attractive and well scaling product. But some of the technological advances in Skylake-SP are overshadowed by the heavy price tags and somewhat "over the top" market segmentation.
http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade

As discussed in many places, the review is plain wrong. During their setup the performance of the Broadwell and Skylake Xeons was crippled by 40% and higher amounts

https://www.reddit.com/r/intel/comments/6n2sww/did_anandtech_cripple_xeon_results/

David Kanter thoughts:

Skylake-SP actually scores about 1.38X higher than Naples on GCC. It's 1.58X higher on a per-core basis.

See https://www.spec.org/cpu2006/results/res2017q3/cpu2006-20170627-47389.html and https://www.spec.org/cpu2006/results/res2017q2/cpu2006-20170529-47127.html

So I'm pretty sure that something is horribly wrong in Johan's testing.

Johan got EPYC core to be 30% faster than Skylake core. In reality the Skylake core is 58% faster than EPYC core, as one can see in above links. This means Johan test is crippling the performance of the Skylake core by a huge 2x factor. Indeed "something is horribly wrong in Johan's testing".

From Ricardo B.:

On March 2016's Broadwell-EP review, E5-2699v4 scored 33.3 on 403.gcc.

On the July 2017 Skylake-SP vs EPYC review, E5-2699v4 scores 23.7 on 403.gcc.

David Kanter was off by about 15% on baseline Ryzen performance....why should we take what he is saying about EPYC at face value given his underestimation of Ryzen cores already?

The early estimation of RyZen performance that David Kanter did back in 2015 was based in information given to him by AMD. My prediction of SPEC performance was very accurate. I almost nailed the score before reviews measured.

David, Ricardo, Linus, and others (including myself) are now discussing measured performance and demonstrating that AT review is wrong by huge amounts like 100%,....

 

[goldstone77]

Intel is LOSING its DIGNITY - WAN Show July 14, 2017
Linus Tech Tips
Published on Jul 14, 2017
00:20:20 - Intel says AMD EPYC processors "glued-together" in official slide deck
https://www.youtube.com/watch?v=f8sXQ6JsNu8

Intel disses AMD’s new processors in what could be an Epyc fail
By Darren Allan 2 days ago Processors
Chip giant reckons AMD hasn’t got a glue
"As you may be aware, AMD followed up its consumer-targeted Ryzen processors with Epyc CPUs (previously codenamed Naples) aimed at servers and data centres – and Intel has been busy slinging mud at the latter chips.

In a presentation slide, Intel claims that Epyc processors (which are based on the same Zen architecture as Ryzen, but go up to a mighty 32-cores) are ‘glued-together’ and a ‘repurposed desktop product for server’.

Basically, Intel is saying this is a cobbled together product, along with the glue comment, references the fact that AMD’s chips aren’t a single die, rather they are comprised of four dies put together. However, this isn’t a bad design, as Tech PowerUp, observes.

Indeed, the tech site argues that Zen cores were built from the ground-up for modularity and scalability, and to craftily maximise yields for AMD – with the Epyc chips also delivering impressive results on the power/performance front.

Intel’s slide, however, engages in other seemingly low blows, accusing AMD’s processors of having ‘inconsistent performance’ due to this ‘glued-together’ nature, and further accuses its rival of lacking in terms of its supporting ecosystem."

Broad brush
"The problem is that making criticisms like these, with broad brush strokes such as vague accusations of inconsistency – as opposed to detailed comparisons or benchmark breakdowns – doesn’t really present Intel in a good light.

Particularly when in another slide, Intel further has a pop at Epyc CPUs by bringing up Ryzen’s initial problems with optimisation for games, and notes that buyers should expect similar software optimisations to be required for the server chips.

Which simply doesn’t follow. Exactly what does tweaking Ryzen for performance with specific games (perhaps at certain resolutions) have to do with what enterprise customers can expect from Epyc? Again, it just looks like a broad smear tactic. ‘This happened here, so it’s bound to happen again somehow…’

And this sort of marketing simply makes Intel look worried about the turf it might lose to AMD with these new Zen chips. After all, it appears that Ryzen is already making some considerable inroads into Intel territory with desktop PCs – albeit in terms of enthusiast rigs as these figures indicated earlier this month.

Intel may just have shot itself in the foot here – with a glue gun, if we can indulge in a bit of poetic licence, leaving the firm in a somewhat sticky situation regarding the tone of its marketing. It’ll certainly be interesting to see if AMD has any response."
http://www.techradar.com/news/intel-disses-amds-new-processors-in-what-could-be-an-epyc-fail

Intel slide criticizes AMD for using "glued-together" dies in Epyc processors
By Paul Lilly 3 days ago
Salty, much?
http://www.pcgamer.com/intel-slide-criticizes-amd-for-using-glued-together-dies-in-epyc-processors/

Edit: additional
Intel Says AMD EPYC Processors "Glued-together" in Official Slide Deck
by Raevenlord Wednesday, July 12th 2017 05:12
https://www.techpowerup.com/235092/intel-says-amd-epyc-processors-glued-together-in-official-slide-deck

I ran across story while reading commentary on the previous, and thought it was worth mentioning.

ECHO OF THE BUNNYMEN: HOW AMD WON, THEN LOST
by: Brian Benchoff
DAY: DECEMBER 9, 2015
http://hackaday.com/2015/12/09/echo-of-the-bunnymen-how-amd-won-then-lost/

 

[juanrga]

Intel is LOSING its DIGNITY - WAN Show July 14, 2017
Linus Tech Tips
Published on Jul 14, 2017
00:20:20 - Intel says AMD EPYC processors "glued-together" in official slide deck
https://www.youtube.com/watch?v=f8sXQ6JsNu8

Intel disses AMD’s new processors in what could be an Epyc fail
By Darren Allan 2 days ago Processors
Chip giant reckons AMD hasn’t got a glue
"As you may be aware, AMD followed up its consumer-targeted Ryzen processors with Epyc CPUs (previously codenamed Naples) aimed at servers and data centres – and Intel has been busy slinging mud at the latter chips.

In a presentation slide, Intel claims that Epyc processors (which are based on the same Zen architecture as Ryzen, but go up to a mighty 32-cores) are ‘glued-together’ and a ‘repurposed desktop product for server’.

Basically, Intel is saying this is a cobbled together product, along with the glue comment, references the fact that AMD’s chips aren’t a single die, rather they are comprised of four dies put together. However, this isn’t a bad design, as Tech PowerUp, observes.

Indeed, the tech site argues that Zen cores were built from the ground-up for modularity and scalability, and to craftily maximise yields for AMD – with the Epyc chips also delivering impressive results on the power/performance front.

Intel’s slide, however, engages in other seemingly low blows, accusing AMD’s processors of having ‘inconsistent performance’ due to this ‘glued-together’ nature, and further accuses its rival of lacking in terms of its supporting ecosystem."

Broad brush
"The problem is that making criticisms like these, with broad brush strokes such as vague accusations of inconsistency – as opposed to detailed comparisons or benchmark breakdowns – doesn’t really present Intel in a good light.

Particularly when in another slide, Intel further has a pop at Epyc CPUs by bringing up Ryzen’s initial problems with optimisation for games, and notes that buyers should expect similar software optimisations to be required for the server chips.

Which simply doesn’t follow. Exactly what does tweaking Ryzen for performance with specific games (perhaps at certain resolutions) have to do with what enterprise customers can expect from Epyc? Again, it just looks like a broad smear tactic. ‘This happened here, so it’s bound to happen again somehow…’

And this sort of marketing simply makes Intel look worried about the turf it might lose to AMD with these new Zen chips. After all, it appears that Ryzen is already making some considerable inroads into Intel territory with desktop PCs – albeit in terms of enthusiast rigs as these figures indicated earlier this month.

Intel may just have shot itself in the foot here – with a glue gun, if we can indulge in a bit of poetic licence, leaving the firm in a somewhat sticky situation regarding the tone of its marketing. It’ll certainly be interesting to see if AMD has any response."
http://www.techradar.com/news/intel-disses-amds-new-processors-in-what-could-be-an-epyc-fail

Intel slide criticizes AMD for using "glued-together" dies in Epyc processors
By Paul Lilly 3 days ago
Salty, much?
http://www.pcgamer.com/intel-slide-criticizes-amd-for-using-glued-together-dies-in-epyc-processors/

Edit: additional
Intel Says AMD EPYC Processors "Glued-together" in Official Slide Deck
by Raevenlord Wednesday, July 12th 2017 05:12
https://www.techpowerup.com/235092/intel-says-amd-epyc-processors-glued-together-in-official-slide-deck

I ran across story while reading commentary on the previous, and thought it was worth mentioning.

ECHO OF THE BUNNYMEN: HOW AMD WON, THEN LOST
by: Brian Benchoff
DAY: DECEMBER 9, 2015
http://hackaday.com/2015/12/09/echo-of-the-bunnymen-how-amd-won-then-lost/

A exact copy of this message was posted in the Intel thread. My reply to it can be found therein

http://www.tomshardware.co.uk/forum/id-1581001/intel-future-chips-news-rumours-reviews/page-38.html#19943768

 

[goldstone77]

Intel is LOSING its DIGNITY - WAN Show July 14, 2017
Linus Tech Tips
Published on Jul 14, 2017
00:20:20 - Intel says AMD EPYC processors "glued-together" in official slide deck
https://www.youtube.com/watch?v=f8sXQ6JsNu8

Intel disses AMD’s new processors in what could be an Epyc fail
By Darren Allan 2 days ago Processors
Chip giant reckons AMD hasn’t got a glue
"As you may be aware, AMD followed up its consumer-targeted Ryzen processors with Epyc CPUs (previously codenamed Naples) aimed at servers and data centres – and Intel has been busy slinging mud at the latter chips.

In a presentation slide, Intel claims that Epyc processors (which are based on the same Zen architecture as Ryzen, but go up to a mighty 32-cores) are ‘glued-together’ and a ‘repurposed desktop product for server’.

Basically, Intel is saying this is a cobbled together product, along with the glue comment, references the fact that AMD’s chips aren’t a single die, rather they are comprised of four dies put together. However, this isn’t a bad design, as Tech PowerUp, observes.

Indeed, the tech site argues that Zen cores were built from the ground-up for modularity and scalability, and to craftily maximise yields for AMD – with the Epyc chips also delivering impressive results on the power/performance front.

Intel’s slide, however, engages in other seemingly low blows, accusing AMD’s processors of having ‘inconsistent performance’ due to this ‘glued-together’ nature, and further accuses its rival of lacking in terms of its supporting ecosystem."

Broad brush
"The problem is that making criticisms like these, with broad brush strokes such as vague accusations of inconsistency – as opposed to detailed comparisons or benchmark breakdowns – doesn’t really present Intel in a good light.

Particularly when in another slide, Intel further has a pop at Epyc CPUs by bringing up Ryzen’s initial problems with optimisation for games, and notes that buyers should expect similar software optimisations to be required for the server chips.

Which simply doesn’t follow. Exactly what does tweaking Ryzen for performance with specific games (perhaps at certain resolutions) have to do with what enterprise customers can expect from Epyc? Again, it just looks like a broad smear tactic. ‘This happened here, so it’s bound to happen again somehow…’

And this sort of marketing simply makes Intel look worried about the turf it might lose to AMD with these new Zen chips. After all, it appears that Ryzen is already making some considerable inroads into Intel territory with desktop PCs – albeit in terms of enthusiast rigs as these figures indicated earlier this month.

Intel may just have shot itself in the foot here – with a glue gun, if we can indulge in a bit of poetic licence, leaving the firm in a somewhat sticky situation regarding the tone of its marketing. It’ll certainly be interesting to see if AMD has any response."
http://www.techradar.com/news/intel-disses-amds-new-processors-in-what-could-be-an-epyc-fail

Intel slide criticizes AMD for using "glued-together" dies in Epyc processors
By Paul Lilly 3 days ago
Salty, much?
http://www.pcgamer.com/intel-slide-criticizes-amd-for-using-glued-together-dies-in-epyc-processors/

Edit: additional
Intel Says AMD EPYC Processors "Glued-together" in Official Slide Deck
by Raevenlord Wednesday, July 12th 2017 05:12
https://www.techpowerup.com/235092/intel-says-amd-epyc-processors-glued-together-in-official-slide-deck

I ran across story while reading commentary on the previous, and thought it was worth mentioning.

ECHO OF THE BUNNYMEN: HOW AMD WON, THEN LOST
by: Brian Benchoff
DAY: DECEMBER 9, 2015
http://hackaday.com/2015/12/09/echo-of-the-bunnymen-how-amd-won-then-lost/

A exact copy of this message was posted in the Intel thread. My reply to it can be found therein

http://www.tomshardware.co.uk/forum/id-1581001/intel-future-chips-news-rumours-reviews/page-38.html#19943768

FTC Settles Charges of Anticompetitive Conduct Against Intel
Provisions are Designed to Foster Competition in the Computer Chip Business
August 4, 2010
"Under the settlement, Intel will be prohibited from:

conditioning benefits to computer makers in exchange for their promise to buy chips from Intel exclusively or to refuse to buy chips from others; and
retaliating against computer makers if they do business with non-Intel suppliers by withholding benefits from them.
In addition, the FTC settlement order will require Intel to:

modify its intellectual property agreements with AMD, Nvidia, and Via so that those companies have more freedom to consider mergers or joint ventures with other companies, without the threat of being sued by Intel for patent infringement;
offer to extend Via’s x86 licensing agreement for five years beyond the current agreement, which expires in 2013;
maintain a key interface, known as the PCI Express Bus, for at least six years in a way that will not limit the performance of graphics processing chips. These assurances will provide incentives to manufacturers of complementary, and potentially competitive, products to Intel’s CPUs to continue to innovate; and
disclose to software developers that Intel computer compilers discriminate between Intel chips and non-Intel chips, and that they may not register all the features of non-Intel chips. Intel also will have to reimburse all software vendors who want to recompile their software using a non-Intel compiler."
https://www.ftc.gov/news-events/press-releases/2010/08/ftc-settles-charges-anticompetitive-conduct-against-intel

Suit: Intel paid Dell up to $1 billion a year not to use AMD chips
http://money.cnn.com/blogs/legalpad/2007/02/suit-intel-paid-dell-up-to-1-billion_15.html?source=yahoo_quote

 

[juanrga]

A exact copy of the above message was posted in the Intel thread. My reply to the above message can be found here

http://www.tomshardware.co.uk/forum/id-1581001/intel-future-chips-news-rumours-reviews/page-38.html#19944010

 

[aldaia]

Processor Arithmetic
2x AMD EPYC 7601 32-Core Processor (4N 32C 64T 3.2GHz, 1.33GHz IMC, 32x 512kB L2, 8x 8MB L3) = 706.18GOPS
2x Intel(R) Xeon(R) Platinum 8180 CPU @ 2.50GHz (28C 56T 3.8GHz, 2.4GHz IMC, 28x 1MB L2, 38.5MB L3) = 1425.82GOPS

http://ranker.sisoftware.net/show_device.php?q=c9a598aad2f2b3feba9adf8fd695b582b484b595a694b9fa95e782a4c3fed3e2c4b68bba9cf5c8f8deb68bbb9de5d8e8ceabcef3c3e596ab93&l=en

Something is wrong here. I don't know if you accidentally picked best case for intel/worst case for AMD or is just a typo. See the Average Score below. Also comparable intel processor to 7601 pricewise is Xeon 8160.

$24000 2x Intel Xeon Platinum 8180 1345.76GOPS -> 56 Mops/$
$9400 2x Intel Xeon Platinum 8160 851.84GOPS -> 90 Mops/$
$8000 2x AMD EPYC 7601 32-Core 1242.00GOPS -> 155 Mops/$

http://ranker.sisoftware.net/top_device.php?q=c2ffcfe984e9d4e4c2b08dbd9bf2cff8deb68bbb9de5d8e9cfaacff2c2e497aa9a&l=en

 

[aldaia]

As discussed in many places, the review is plain wrong. During their setup the performance of the Broadwell and Skylake Xeons was crippled by 40% and higher amounts

The review is not wrong, it seems very specific to 403.gcc subtest that represents what? 2-3% of the review?
I agree that 403.gcc has raised some concerns that should be clarified, but that doesn't invalidate the whole review.

David Kanter thoughts:

Skylake-SP actually scores about 1.38X higher than Naples on GCC. It's 1.58X higher on a per-core basis.

See https://www.spec.org/cpu2006/results/res2017q3/cpu2006-20170627-47389.html and https://www.spec.org/cpu2006/results/res2017q2/cpu2006-20170529-47127.html

So I'm pretty sure that something is horribly wrong in Johan's testing.

David Kanter is reporting spec rate for the $12000 much higher clocked 8180 instead of the $8000 lower clocked 8176 used in the review. Big difference!!!. The review shows 1400 rate for epyc on 403.gcc and 1300 rate for 8176, only 8% difference. Nothing to be concerned considering AT used gcc to compile the benchmarks. Spec rate on 403.gcc for epyc with Open64 is 1680 vs 1740 for 8176 with Intel compiler, the differences are negligible, and far from 38%.
https://www.spec.org/cpu2006/results/res2017q3/cpu2006-20170626-47330.html
https://www.spec.org/cpu2006/results/res2017q2/cpu2006-20170529-47127.html

Johan got EPYC core to be 30% faster than Skylake core. In reality the Skylake core is 58% faster than EPYC core, as one can see in above links. This means Johan test is crippling the performance of the Skylake core by a huge 2x factor. Indeed "something is horribly wrong in Johan's testing".

1) Your argument is built on top of a wrong argument.
2) Single thread performance cannot (and should not) be estimated from spec rate. Use single thread spec scores instead.

From Ricardo B.:

On March 2016's Broadwell-EP review, E5-2699v4 scored 33.3 on 403.gcc.

On the July 2017 Skylake-SP vs EPYC review, E5-2699v4 scores 23.7 on 403.gcc.

That is indeed a good point, and the one that in my opinion deserves further exploration. Something strange happens with 403.gcc, and not only in Johan's review. Official SPEC submissions for Xeon 8176 have scores varying by up to 40%.Maybe there is something in broadwell/skylake that produces big changes in 403.gcc scores depending on the compiler-kernel-library and flags used. Some people point to AVX2 AVX-512, but so far I've seen nothing conclusive.

Ideally someone should use the same platform, OS, and compiler and try to reproduce the results. That is the only way of validating/refuting the review. Anything else is only speculation.

 

[juanrga]

It is not only the gcc subtest, NAMD scores in the review are also weird.

The gcc subtest is a fundamental that represents very good server performance and that is why David, Linus, Ricardo, and rest of people at RWT is discussing it. Once Linus devoted some posts to explain why the gcc subtest is so important, but I missed the link.

The huge discrepancy of 100% between the official spec database and AT results cannot explained by the change of model and/or cost. Don't forget as well that cost includes the AVX512 units that increase performance by 50--100% above Broadwell.

Same comments about OS and compiler used. Those variables can affect the gcc subscore by some low percents, not introduce a discrepancy of about 40--100%.

The score for the Xeon 8176 is 1740 whereas it is 2020 for the Xeon 8180. The difference is a 16% which matches very well the difference in all-core turbo between both models. Again this doesn't explain the discrepancy. gcc subtest doesnt use AVX, again this doesn't explain discrepancy.

The point is that both Broadwell and Skylake Xeons have performance crippled by huge amounts in the AT review. This is not speculation but technical analysis of the review. Power consumption figures are also incorrect. I don't have link here, but only info someone tested both EPYC and Xeon gave me.

 

[BonScott]

Junagra, EPYC is crushing Skykake SP and you can't handle it.

 

[juanrga]

Barclays analysts have confirmed that almost no one want EPYC

We have conducted regular channel checks since the fall and are still hearing little to no material traction at the ODMs [original design manufacturers of computers], which is the ultimate gauge of success.

http://www.cnbc.com/2017/07/18/barclays-predicts-shares-of-amd-will-plunge-35-percent.html

I predicted this was going to happen and I explained why. EPYC is slower, hotter, and scales worse.

 

[TMTOWTSAC]

Seems a little early to pass judgement. Especially in an extremely slow moving and conservative market segment. Even if it were an incontrovertibly superior product, datacenters just don't shift platforms spur of the moment.

I think the first place Epyc might gain some traction is the small business market. Specifically new businesses that don't yet have a large investment in their IT infrastructure yet. The price point forces consideration, and the performance aspects aren't as relevant since their overall demand is low. Of course, the stumbling block here is the rise of cloud services. But not everyone is willing to cede control of their hardware so there should still be opportunities.

 

[gamerk316]

Seems a little early to pass judgement. Especially in an extremely slow moving and conservative market segment. Even if it were an incontrovertibly superior product, datacenters just don't shift platforms spur of the moment.

I think the first place Epyc might gain some traction is the small business market. Specifically new businesses that don't yet have a large investment in their IT infrastructure yet. The price point forces consideration, and the performance aspects aren't as relevant since their overall demand is low. Of course, the stumbling block here is the rise of cloud services. But not everyone is willing to cede control of their hardware so there should still be opportunities.

Those types of business will just purchase the lowest cost product that meets their needs. More then likely, that's the Pentium/i3 lineup.

 

[goldstone77]

Barclays analysts have confirmed that almost no one want EPYC

We have conducted regular channel checks since the fall and are still hearing little to no material traction at the ODMs [original design manufacturers of computers], which is the ultimate gauge of success.

http://www.cnbc.com/2017/07/18/barclays-predicts-shares-of-amd-will-plunge-35-percent.html

I predicted this was going to happen and I explained why. EPYC is slower, hotter, and scales worse.

Intel Plays Defense: Inside Its EPYC Slide Deck
by Paul Alcorn July 17, 2017 at 12:00 PM
"EPYC changes the game. During AMD's launch event, representatives from several major companies took to the stage and expressed support for the platform. Baidu, Microsoft, Supermicro, Dell, Xilinx, HPE, Dropbox, Samsung, and Mellanox were all there. Notice the Super Seven+1 members? Surely there are other high-profile names being courted behind the scenes, so we expect more partner announcements in the future."
"These CPUs are a threat to Intel's margins because they give Xeon customers another option. Consequently, Intel might have to get more price-competitive in key portions of its product stack, especially with high-volume customers. That means EPYC could affect Intel's bottom line, even if it doesn't gain significant market share."
"There is little doubt that AMD's EPYC will find some measure of success in the data center, and Intel wants to get ahead of any potential adoption."
Gaming Data Center Performance & Final Thoughts
"AMD does have a smaller group of partners, but if it offers the right value-to-performance ratio, we expect it could enjoy similar long-term success."
http://www.tomshardware.com/reviews/intel-amd-die-fabric-slides,5125.html?utm_source=th-newsletter&utm_medium=email&utm_campaign=20170718-th