Intel Skylake SP 8176 Review vs AMD Epyc 7601

[The Benjamins]

Quote

This morning kicks off a very interesting time in the world of server-grade CPUs. Officially launching today is Intel's latest generation of Xeon processors, based on the "Skylake-SP" architecture. The heart of Intel's new Xeon Scalable Processor family, the "Purley" 100-series processors incorporate all of Intel's latest CPU and network fabric technology, not to mention a very large number of cores.

Meanwhile, a couple of weeks back AMD soft-launched their new EPYC 7000 series processors. Based on the company's Zen architecture and scaled up to server-grade I/O and core counts, EPYC represents an epic achievement for AMD, once again putting them into the running for competitive, high performance server CPUs after nearly half a decade gone. EPYC processors have begun shipping, and just in time for today's Xeon launch, we also have EPYC hardware in the lab to test.

 

This is looking like a good time to be shopping for new servers. I will post a few charts but I dont have to much time at work to summarize this big article right now. I am liking how Epyc is faster in most test and uses less power too.

 

It is nice seeing a big improvement on Skylake versus Broadwell.

 

You can see Epyc losing in compression but wins in decompression.

 

Quote

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.

 

http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade

http://www.anandtech.com/show/11615/intels-data-center-event-live-blog-830am-pt

https://www.pcper.com/reviews/Processors/Intel-Xeon-Scalable-Processor-Launch-New-Architecture-New-Platform-Data-Center

 

 

 

 

 

 

 

 

 

 

 

Pricing from PCPer

Spoiler

Intel Xeon Scalable ProcessorsPriceSocketsMemory  Capacity

8180M (38.5M cache, 28 Cores, 56 Threads, 2.50GHz (205W) 10.40 GT/sec Intel® QPI, 14nm) $13,011, S8, 1.5TB

8176M (38.5M cache, 28 Cores, 56 Threads, 2.10GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $11,722, S8, 1.5TB

8170M (35.75M cache, 26 Cores, 52 Threads, 2.10GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $10,409, S8, 1.5TB

8160M (33M cache, 24 Cores, 48 Threads, 2.10GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $7,704, S8, 1.5TB

6142M (22M cache, 16 Cores, 32 Threads, 2.60GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $5,949, S4, 1.5TB

6140M (24.75M cache, 18 Cores, 36 Threads, 2.30GHz (140W) 10.40 GT/sec Intel® QPI, 14nm) $5,448, S4, 1.5TB

6134M (24.75M cache, 8 Cores, 32 Threads, 3.20GHz (130W) 10.40 GT/sec Intel® QPI, 14nm) $5,217, S4, 1.5TB

8180 (38.5M cache, 28 Cores, 56 Threads, 2.50GHz (205W) 10.40 GT/sec Intel® QPI, 14nm) $10,009, S8, 768GB

8176 (38.5M cache, 28 Cores, 56 Threads, 2.10GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $8,719, S8, 768GB

8170 (35.75M cache, 26 Cores, 52 Threads, 2.10GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $7,405, S8, 768GB

8164 (35.75M cache, 26 Cores, 52 Threads, 2.00GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $6,114, S8, 768GB

8168 (33M cache, 24 Cores, 48 Threads, 2.70GHz (205W) 10.40 GT/sec Intel® QPI, 14nm) $5,890, S8, 768GB

8160 (33M cache, 24 Cores, 48 Threads, 2.10GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $4,702, S8, 768GB

8158 (24.75M cache, 12 Cores, 24 Threads, 3.00GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $7,007, S8, 768GB

8156 (16.5M cache, 4 Cores, 8 Threads, 3.60GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $7,007, S8, 768GB

8153 (22M cache, 16 Cores, 32 Threads, 2.00GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $3,115, S8, 768GB

6154 (24.75M cache, 18 Cores, 36 Threads, 3.00GHz (200W) 10.40 GT/sec Intel® QPI, 14nm) $3,543, S4, 768GB

6152 (30.25M cache, 22 Cores, 44 Threads, 2.10GHz (140W) 10.40 GT/sec Intel® QPI, 14nm) $3,655, S4, 768GB

6150 (24.75M cache, 18 Cores, 36 Threads, 2.70GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $3,358, S4, 768GB

6148 (27.5M cache, 20 Cores, 40 Threads, 2.40GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $3,072, S4, 768GB

6146 (24.75M cache, 12 Cores, 24 Threads, 3.20GHz (165W) 10.40 GT/sec Intel® QPI, 14nm) $3,286, S4, 768GB

6144 (24.75M cache, 8 Cores, 16 Threads, 3.50GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $2,925, S4, 768GB

6142 (22M cache, 16 Cores, 32 Threads, 2.60GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $2,946, S4, 768GB

6140 (24.75M cache, 18 Cores, 36 Threads, 2.30GHz (140W) 10.40 GT/sec Intel® QPI, 14nm) $2,445, S4, 768GB

6138 (27.5M cache, 20 Cores, 40 Threads, 2.00GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $2,612, S4, 768GB

6132 (19.25M cache, 14 Cores, 28 Threads, 2.60GHz (140W) 10.40 GT/sec Intel® QPI, 14nm) $2,111, S4, 768GB

6136 (24.75M cache, 12 Cores, 24 Threads, 3.00GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $2,4604768GB

6134 (24.75M cache, 8 Cores, 16 Threads, 3.20GHz (130W) 10.40 GT/sec Intel® QPI, 14nm) $2,2144768GB

6130 (22M cache, 16 Cores, 32 Threads, 2.10GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $1,894, S4, 768GB

6128 (19.25M cache, 6 Cores, 12 Threads, 3.40GHz (115W) 10.40 GT/sec Intel® QPI, 14nm) $1,691, S4, 768GB

6126 (19.25M cache, 12 Cores, 24 Threads, 2.60GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $1,776, S4, 768GB

5122 (16.5M cache, 4 Cores, 8 Threads, 3.60GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $1,221, S4, 768GB

5120 (19.25M cache, 14 Cores, 28 Threads, 2.20GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $1,555, S4, 768GB

5118 (16.5M cache, 12 Cores, 24 Threads, 2.30GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $1,273, S4, 768GB

5115 (13.75M cache, 10 Cores, 20 Threads, 2.40GHz (85W) 10.40 GT/sec Intel® QPI, 14nm) $1,221, S4, 768GB

4116 (16.5M cache, 12 Cores, 24 Threads, 2.10GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $1,002, S2, 768GB

4114 (13.75M cache, 10 Cores, 20 Threads, 2.20GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $694, S2, 768GB

4112 (8.25M cache, 4 Cores, 8 Threads, 2.60GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $473, S2, 768GB

4110 (11M cache, 8 Cores, 16 Threads, 2.10GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $501, S2, 768GB

4108 (11M cache, 8 Cores, 16 Threads, 1.80GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $417, S2, 768GB

3106 (11M cache, 8 Cores, 16 Threads, 1.70GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $306, S2, 768GB

3104 (8.25M cache, 6 Cores, 12 Threads, 1.70GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $213, S2, 768GB

8176F (38.5M cache, 28 Cores, 56 Threads, 2.10GHz (173W) 10.40 GT/sec Intel® QPI, 14nm) $8,8742768GB

8160F (33M cache, 24 Cores, 48 Threads, 2.10GHz (160W) 10.40 GT/sec Intel® QPI, 14nm) $4,856, S2, 768GB

6148F (27.5M cache, 20 Cores, 40 Threads, 2.40GHz (160W) 10.40 GT/sec Intel® QPI, 14nm) $3,227, S2, 768GB

6142F (22M cache, 16 Cores, 32 Threads, 2.60GHz (160W) 10.40 GT/sec Intel® QPI, 14nm) $3,101, S2, 768GB

6138F (27.5M cache, 20 Cores, 40 Threads, 2.00GHz (135W) 10.40 GT/sec Intel® QPI, 14nm) $2,767, S2, 768GB

6130F (22M cache, 16 Cores, 32 Threads, 2.10GHz (135W) 10.40 GT/sec Intel® QPI, 14nm) $2,049, S2, 768GB

6126F (19.25M cache, 12 Cores, 24 Threads, 2.60GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $1,931, S2, 768GB

8160T (33M cache, 24 Cores, 48 Threads, 2.10GHz (150W) 10.40 GT/sec Intel® QPI, 14nm) $4,936, S8, 768GB

6138T (27.5M cache, 20 Cores, 40 Threads, 2.00GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $2,742, S4, 768GB

6130T (22M cache, 16 Cores, 32 Threads, 2.10GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $1,988, S4, 768GB

6126T (19.25M cache, 12 Cores, 24 Threads, 2.60GHz (125W) 10.40 GT/sec Intel® QPI, 14nm) $1,865, S4, 768GB

5120T (19.25M cache, 14 Cores, 28 Threads, 2.20GHz (105W) 10.40 GT/sec Intel® QPI, 14nm) $1,727, S4, 768GB

5119T (19.25M cache, 14 Cores, 28 Threads, 1.90GHz (85W) 10.40 GT/sec Intel® QPI, 14nm) $1,555, S4, 768GB

4116T (16.5M cache, 12 Cores, 24 Threads, 2.10GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $1,112, S2, 768GB

4114T (13.75M cache, 10 Cores, 20 Threads, 2.20GHz (85W) 9.60 GT/sec Intel® QPI, 14nm) $773, S2, 768GB

4109T (11M cache, 8 Cores, 16 Threads, 2.00GHz (70W) 9.60 GT/sec Intel® QPI, 14nm), $5012, 768GB

 

http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade

http://www.anandtech.com/show/11615/intels-data-center-event-live-blog-830am-pt

https://www.pcper.com/reviews/Processors/Intel-Xeon-Scalable-Processor-Launch-New-Architecture-New-Platform-Data-Center

http://www.tomshardware.com/reviews/intel-xeon-platinum-8176-scalable-cpu,5120.html

[tom_w141]

Well the highly anticipated Skylake Xeons as an answer to Epyc have come out like a wet fart... Bet they are pricey too... Is pricing available on these?

 

Looks like AMD with Epyc will be soaking up some of those server profits. With their crazy yields on the 8 core CCXs for zen the profit margin on these chips is no doubt high.

 

2017 hasn't been Intel's year at all.

[The Benjamins]

Just now, tom_w141 said:

Well the highly anticipated Skylake Xeons as an answer to Epyc have come out like a wet fart... Bet they are pricey too? Is pricing available on these?

the 8176 that is being tested it $8000~ and the 8160 is $4700. basically if you are using databases or HPC stuff (AVX 2/512) use intel otherwise AMD seems to be better.

[Jito463]

19 minutes ago, The Benjamins said:

 

Is it just me, or do those memory bandwidth numbers look really odd?  I would have expected Skylake-SP to be superior to Broadwell-EP, but it only (marginally) beats it in 8 thread tests.  Or am I missing something?

[tom_w141]

1 minute ago, The Benjamins said:

8176 that is being tested it $8000

EACH?!

 

Thats double the $4000 Epyc 7601.

[The Benjamins]

1 minute ago, Jito463 said:

Is it just me, or do those memory bandwidth numbers look really odd?  I would have expected Skylake-SP to be superior to Broadwell-EP, but it only beats it in 8 thread tests.  Or am I missing something?

Maybe the mesh vs ring plays a role, that is odd. You can also see Epyc slows down if its not using other CCXs.

 

Just now, tom_w141 said:

EACH?!

 

Thats double the $4000 Epyc 7601.

Yes each.

[tom_w141]

Just now, The Benjamins said:

Yes each.

Wow Intel's arrogance astounds me... They need to stop behaving like they have no competition. They just absolutely refuse to change gear it seems. Well let's see how burying their heads in the sand does Epyc sales

[Coaxialgamer]

Nice . For once AMD is competitive in perf/$ AND perf/w.

[Jito463]

Just now, The Benjamins said:

Maybe the mesh vs ring plays a role, that is odd. You can also see Epyc slows down if its not using other CCXs.

After reading the comments, it seems Anandtech used an older distro.  Maybe that's why the number seem skewed.

 

Quote

CajunArson - Tuesday, July 11, 2017 - link

"All of our testing was conducted on Ubuntu Server "Xenial" 16.04.2 LTS (Linux kernel 4.4.0 64 bit). The compiler that ships with this distribution is GCC 5.4.0."

I'd recommend using a more updated distro and especially a more up to date compiler (GCC 5.4 is only a bug-fix release of a compiler from *2015*) if you want to see what these parts are truly capable of.

Phoronix does heavy-duty Linux reviews and got some major performance boosts on the i9 7900X simply by using up to date distros: http://www.phoronix.com/scan.php?page=article&...

Considering that Purley is just an upscaled version of the i9 7900X, I wouldn't be surprised to see different results.

 

[Coaxialgamer]

35 minutes ago, tom_w141 said:

Wow Intel's arrogance astounds me... They need to stop behaving like they have no competition. They just absolutely refuse to change gear it seems. Well let's see how burying their heads in the sand does Epyc sales

it's not necessarily just arrogance . At 28 cores , that xeon die is massive (>500mm² ) and very costly to manufacture . AMD just uses 4 smaller 200mm² dies , making it considerably easier and cheaper to manufacture .That xeon likely legitimately costs more to make than that epyc cpu .

[AnonymousGuy]

 

1 minute ago, Coaxialgamer said:

it's not necessarily just arrogance . At 28 cores , that xeon die is massive (>500mm² ) and very costly to manufacture . AMD just uses 4 smaller 200mm² dies , making it considerably easier and cheaper to manufacture .That xeon likley legitemately costs more to make than that epyc cpu .

E5 and E7 Xeons are separate products from the consumer side.  New design, new litho, new testing and validation.  Epyc they basically just made a 4S system on the same package which is a far lower skill requirement.

[The Benjamins]

17 minutes ago, tom_w141 said:

Well the highly anticipated Skylake Xeons as an answer to Epyc have come out like a wet fart... Bet they are pricey too... Is pricing available on these?

 

Looks like AMD with Epyc will be soaking up some of those server profits. With their crazy yields on the 8 core CCXs for zen the profit margin on these chips is no doubt high.

 

2017 hasn't been Intel's year at all.

PCPer had pricing, now added with link to their article.

[Coaxialgamer]

Just now, AnonymousGuy said:

 

E5 and E7 Xeons are separate products from the consumer side.  New design, new litho, new testing and validation.  Epyc they basically just made a 4S system on the same package which is a far lower skill requirement.

i know xeon e7 and e5 are a seperate lineup . Intel couldn't possibly manufacture such large chips in volume , let alone at a price consumers could afford .I wouldn't call an MCM an"easier" approach , just a more elegant one . Intel just brute forced a large chip because they have the R&D money .

Also , broadwell was also produced on 14nm , so this isn't technically a new process.

[tom_w141]

29 minutes ago, Coaxialgamer said:

That xeon likley legitemately costs more to make than that epyc cpu .

Oh i absolutely agree. Intel have to accept this and put less profit % on top rather than their usual.

[tom_w141]

26 minutes ago, AnonymousGuy said:

Epyc they basically just made a 4S system on the same package which is a far lower skill requirement.

True but its going toe to toe/sometimes beating a chip that costs double. Sometimes simple is better.

[NumLock21]

So Xeon can be used for entherium mining?!

Video cards can finally go back to their regular prices! Yay!

 

 

And here is the cpu fan installation video. It's for the Xeon Phi, where the Xeon are using the same type of socket.

 

[apoclypse]

In certain instance it seems that even with AVX on Inte's side, AMD still has an advantage, So good on AMD. 

[apoclypse]

1 hour ago, Jito463 said:

After reading the comments, it seems Anandtech used an older distro.  Maybe that's why the number seem skewed.

Doesn't matter in real world scenarios imo. Anandtech used the latest LTS version of the distro which in the server space is all they are going to use. They aren't going to be running the latest version of any distro. It's either CentOS or LTS versions of Ubuntu or possibly something older. Anandtech seemed to be going for more real world scenario benchmarks. In that kind of workflow AMDs chips performed admirably. 

 

Now if we are going to compare the latest and greatest against the latest and greatest, let's see what happens when Threadrippers arrives. 

[The Benjamins]

50 minutes ago, apoclypse said:

In certain instance it seems that even with AVX on Inte's side, AMD still has an advantage, So good on AMD. 

Ya things are looking really good for AMD cpu wise.

[captain cactus]

More expensive than Epyc AND it probably needs 3 nuclear reactors for each CPU going off of Skylake-X's power consumption numbers.

 

Seems like a no-brainer if you ask me.

[PCGuy_5960]

Now that I look at the scores a bit more carefully, they are quite competitive. It's good to see that we finally have competition in the server market

Edited July 11, 2017 by PCGuy_5960

[Drak3]

3 minutes ago, PCGuy_5960 said:

Am I the only one who thinks that Skylake Xeons are a bit disappointing compared to EPYC?

Yes, which has me asking what are each clocked at? Is SLSP different from Sky that it requires some optimization? If so, is that also the case with SLX? Did AMD have "Ryzen" dies that weren't cut back in AVX, that were always meant just for EPYC?

[MageTank]

27 minutes ago, lots of unexplainable lag said:

More expensive than Epyc AND it probably needs 3 nuclear reactors for each CPU going off of Skylake-X's power consumption numbers.

 

Seems like a no-brainer if you ask me.

Yeah... except you won't be running these at nearly that high of a voltage or clock speed. Remember, when it comes to thermals and power consumption, voltage scales quadratically. Running these high-core count CPU's at much lower clock speeds, allows them to do so at much lower voltages. While power consumption will still be pretty high, it won't be as drastic as you think it will be. 

 

The 18c/36t Xeon 2697 V4 pulled just about 470w under Linpack AVX2 load, which isn't bad at all. That was Broadwell as well. Skylake brings a superior turbo boost technology that should help improve efficiency across the board, and there should be plenty of headroom left for them to throw a few more cores on it without causing that drastic a difference in power consumption. From what I've seen, the stock voltage of Skylake-X is pretty aggressive to maintain the stock clocks, which is why you see high thermals out of the gate. The power consumption will depend on exactly which kind of load you throw at it, along with other settings that remove current limitations and the type of instruction sets you use (alongside memory overclocking, which can make AVX far more stressful).

 

I wouldn't be worried about these Xeon's power consumption too much. The 7980XE on the other hand... that's worthy of all kinds of fear, lol. 

[Trixanity]

4 hours ago, Jito463 said:

After reading the comments, it seems Anandtech used an older distro.  Maybe that's why the number seem skewed.

 

 

I don't think that particular commenter is a reliable source. If you read his other comments he claims the Zen architecture is a Haswell copy + paste job. If that was the case, one would assume Intel would be very vocal about it and that they would share the same characteristics (which they don't considering the peculiarities of Zen). And non-LTS distros aren't exactly going to be used in enterprise environments although I do concur that it might cause the Xeon to not run on all cylinders. However, might the same not be said for EPYC? AMD isn't exactly ubiquitous in the server space. Given everything we've seen so far, an older distro is probably optimized (as in developed with that kind of system in mind) for Haswell.

[MageTank]

1 minute ago, Drak3 said:

Yes, which has me asking what are each clocked at? Is SLSP different from Sky that it requires some optimization? If so, is that also the case with SLX? Did AMD have "Ryzen" dies that weren't cut back in AVX, that were always meant just for EPYC?

AVX scales with memory bandwidth. Having 8 channels for copious amounts of bandwidth, that will certainly brute force some AVX. As for whether they actually changed their AVX pipeline to be able to handle more bit-ops/clock, that's highly unlikely. If AMD would give us real control over Ryzen's memory controller, I'd wager I could make it's AVX performance at least on par with Haswell's, but those new registers completely ignored tREFI, 2DPC tertiary timings and multi-rank timings. I specifically need the 2DPC and multi-rank timings for tRDWR and tWRRD, the timings that directly influence AVX performance.