Intel Skylake / Kaby Lake

[imported_ats]

Yes, but that is happening because one capable system meant going HEDT route for Intel, which was more expensive. Now, when there is Ryzen 7 on the market, you can keep only one system, one monitor, less space occupied, etc, etc.

The need for the second system was an effect, not a cause. Why keep 2 windows installs and edit on the lesser secondary system when you can benefit from the power of the better one?

For me that sounds like excuses. If you have a 6900 or 6950 PC, you don't need another. Same for R7 and Threadripper.

No the second system, wasn't effect, it is simply the best setup. Regardless of the configuration of the gaming rig, you are always better off with a separate streaming PC. Literally have streamers with 69(5)0 using secondary streaming rigs. This allows them to use a relatively cheap capture card to support both PC gaming as well as console gaming as well as have the camera and mixing entirely separate from the gaming rig. Regardless of the power of the main rig, using it for streaming, mixing, and camera input affects gaming performance. Having a second streaming rig also vastly simplifies setup and configuration.

 

[Zucker2k]

i7-7900X (@4.0 Ghz): http://ranker.sisoftware.net/show_r...d5e3daebdfe9dbfd8fb282a4c1a499a98ffcc1f9&l=en

For comparison
i7-6950X (@4.50 Ghz)= http://ranker.sisoftware.net/show_r...d5e3daebdfe9dff98bb686a0c5a09da88efdc0f8&l=en
i7-6950X (@3.50 Ghz)= http://ranker.sisoftware.net/show_r...d5e3d5e7dfe9dcfa88b585a3c6a39eae88fbc6fe&l=en

i7-7900X 10C (@4.0 Ghz)= 1386.94Mpix/s
i7-6950X 10C (@4.50 Ghz)= 897.28Mpix/s
i7-6950X 10C (@3.50 Ghz)= 746.64Mpix/s


7900x is called i7 here, possibly the i9 branding was a last minute change from Intel. Based on the scores there is no full speed AVX 512 involved I would say (as we already know). The difference is still big, +55% to a faster clocked 10C Broadwell and 86% to a lower clocked Broadwell 10C.


edit: Not sure about AVX 512 to be honest, I mean +70% at the same clock and core count seems too high.

The 7900x is listed as 10c/20t too. That performance is just insane. Just a moment.... need to change underwear...

 

[Sweepr]

i7-7900X (@4.0 Ghz): http://ranker.sisoftware.net/show_r...d5e3daebdfe9dbfd8fb282a4c1a499a98ffcc1f9&l=en

For comparison
i7-6950X (@4.50 Ghz)= http://ranker.sisoftware.net/show_r...d5e3daebdfe9dff98bb686a0c5a09da88efdc0f8&l=en
i7-6950X (@3.50 Ghz)= http://ranker.sisoftware.net/show_r...d5e3d5e7dfe9dcfa88b585a3c6a39eae88fbc6fe&l=en

i7-7900X 10C (@4.0 Ghz)= 1386.94Mpix/s
i7-6950X 10C (@4.50 Ghz)= 897.28Mpix/s
i7-6950X 10C (@3.50 Ghz)= 746.64Mpix/s


7900x is called i7 here, possibly the i9 branding was a last minute change from Intel. Based on the scores there is no full speed AVX 512 involved I would say (as we already know). The difference is still big, +55% to a faster clocked 10C Broadwell and 86% to a lower clocked Broadwell 10C.


edit: Not sure about AVX 512 to be honest, I mean +70% at the same clock and core count seems too high.

Nice find mikk. Confirms the naming and the (high) 4.5 GHz Turbo clocks speed.

Also:

ASRock X299 motherboards
Before you ask, there are pictures. What we have are entries from ASRock website, which already list four X299 motherboards:

• X299 Fatal1ty Professional Gaming i9
• X299 Fatal1ty Professional Gaming i7
• X299 Gaming K6
• X299 Taichi

 

[Nothingness]

edit: Not sure about AVX 512 to be honest, I mean +70% at the same clock and core count seems too high.

Agreed. It's perhaps AVX-512 at reduced clock frequency. And that'd be quite good as I was expecting more reduction in clock frequency.

 

[LTC8K6]

i7-7900X (@4.0 Ghz): http://ranker.sisoftware.net/show_r...d5e3daebdfe9dbfd8fb282a4c1a499a98ffcc1f9&l=en

For comparison
i7-6950X (@4.50 Ghz)= http://ranker.sisoftware.net/show_r...d5e3daebdfe9dff98bb686a0c5a09da88efdc0f8&l=en
i7-6950X (@3.50 Ghz)= http://ranker.sisoftware.net/show_r...d5e3d5e7dfe9dcfa88b585a3c6a39eae88fbc6fe&l=en

i7-7900X 10C (@4.0 Ghz)= 1386.94Mpix/s
i7-6950X 10C (@4.50 Ghz)= 897.28Mpix/s
i7-6950X 10C (@3.50 Ghz)= 746.64Mpix/s


7900x is called i7 here, possibly the i9 branding was a last minute change from Intel. Based on the scores there is no full speed AVX 512 involved I would say (as we already know). The difference is still big, +55% to a faster clocked 10C Broadwell and 86% to a lower clocked Broadwell 10C.


edit: Not sure about AVX 512 to be honest, I mean +70% at the same clock and core count seems too high.

Crikey...didn't expect that sort of jump.

 

[jpiniero]

Those scores are consistent with full speed AVX-512 enabled. Maybe Intel toyed with the idea of leaving it on at one point.

 

[Edrick]

Those scores are consistent with full speed AVX-512 enabled. Maybe Intel toyed with the idea of leaving it on at one point.

Or maybe you can disable it in the BIOS?

 

[csbin]

Skylake i7 7700 (3.6g-4.2g) 4C AVX2 http://ranker.sisoftware.net/show_r...d5e3d5e2d1e7d2f486bb8badc8ad90a086f5c8f0&l=en

Multi-Media Integer 525.14Mpix/s
Multi-Media Long-int 190.14Mpix/s
Multi-Media Quad-int 2115kpix/s
Multi-Media Single-float 466.73Mpix/s
Multi-Media Double-float 272.81Mpix/s
Multi-Media Quad-float 10500kpix/s

------------------------------------
Skylake-X i9 7900X？(4g-4.5g？) 10C AVX2？ http://ranker.sisoftware.net/show_r...d5e3daebdfe9dbfd8fb282a4c1a499a98ffcc1f9&l=en

Multi-Media Integer 1453.74Mpix/s
Multi-Media Long-int 519.46Mpix/s
Multi-Media Quad-int 5348kpix/s
Multi-Media Single-float 1716.66Mpix/s
Multi-Media Double-float 1069.06Mpix/s
Multi-Media Quad-float 24022kpix/s

 

[Topweasel]

Wow that is a bit of a monster for a 10 core setup.

 

[blue11]

Skylake i7 7700 (3.6g-4.2g) 4C AVX2 http://ranker.sisoftware.net/show_r...d5e3d5e2d1e7d2f486bb8badc8ad90a086f5c8f0&l=en

Multi-Media Integer 525.14Mpix/s
Multi-Media Long-int 190.14Mpix/s
Multi-Media Quad-int 2115kpix/s
Multi-Media Single-float 466.73Mpix/s
Multi-Media Double-float 272.81Mpix/s
Multi-Media Quad-float 10500kpix/s

------------------------------------
Skylake-X i9 7900X？(4g-4.5g？) 10C AVX2？ http://ranker.sisoftware.net/show_r...d5e3daebdfe9dbfd8fb282a4c1a499a98ffcc1f9&l=en

Multi-Media Integer 1453.74Mpix/s
Multi-Media Long-int 519.46Mpix/s
Multi-Media Quad-int 5348kpix/s
Multi-Media Single-float 1716.66Mpix/s
Multi-Media Double-float 1069.06Mpix/s
Multi-Media Quad-float 24022kpix/s

Are SiSoft frequency numbers accurate? In single-precision, the results show a 3.67x gain in throughput from 2.5x gain in cores, which is less than the expected 5x gain. Someone earlier in the thread claimed that the "multi-media" benchmark in SiSoft was in-register, so it should reach theoretical FLOP/s.

 

[nvgpu]

https://newsroom.intel.com/editorials/envision-world-thunderbolt-3-everywhere/

With this vision in mind, Intel is announcing that it plans to drive large-scale mainstream adoption of Thunderbolt by integrating Thunderbolt 3 into future Intel CPUs and by releasing the Thunderbolt protocol specification to the industry next year.

Will Icelake be the first Intel platform with Thunderbolt integrated?

 

[jpiniero]

Will Icelake be the first Intel platform with Thunderbolt integrated?

Icelake would make sense since they could just fuse in a future revision to Alpine Ridge.

'Course this is going make it much easier on Apple to use non-Intel processors.

 

[dullard]

Are SiSoft frequency numbers accurate? In single-precision, the results show a 3.67x gain in throughput from 2.5x gain in cores, which is less than the expected 5x gain. Someone earlier in the thread claimed that the "multi-media" benchmark in SiSoft was in-register, so it should reach theoretical FLOP/s.

Why would you expect 5x gain from 2.5x more cores/threads?

The single and double float scores are much higher than I would have expected. The quad float are less (probably reaching a limitation elsewhere in the system).

 

[Timmah!]

Nice find mikk. Confirms the naming and the (high) 4.5 GHz Turbo clocks speed.

Also:

1386 for SKL-X vs 897 for BW-E? 54 percent more? Sounds legit....

 

[jpiniero]

1386 for SKL-X vs 897 for BW-E? 54 percent more? Sounds legit....

SS Multimedia is for the most part a vector unit test. AVX-512 is literally double the throughput...

 

[blue11]

Why would you expect 5x gain from 2.5x more cores/threads?

The single and double float scores are much higher than I would have expected. The quad float are less (probably reaching a limitation elsewhere in the system).

If SKL-X has full-speed AVX-512, it should have 5 times the throughput of 7700K, since 2.5 * 2 = 5.

SS Multimedia is for the most part a vector unit test. AVX-512 is literally double the throughput...

I looked up what SiSoft "multimedia" actually does, and it turns out to be Mandelbrot. The test appears to mainly measure vector unit width and not actual throughput, as can be seen from this Zen vs Broadwell result, which shows nearly equal performance between 1800X and 6900K, despite the former having half the throughput. My guess is that the implementation in SiSoft is of poor quality and ends up being latency instead of throughput bound. Therefore, the most we can gather from the SiSoft result for the 7900X is that it supports AVX-512 instructions, which we already knew, but whether the AVX-512 unit is full-speed or not is still unknown.

 

[Edrick]

Therefore, the most we can gather from the SiSoft result for the 7900X is that it supports AVX-512 instructions, which we already knew, but whether the AVX-512 unit is full-speed or not is still unknown.

I agree with that assessment. I wonder when we will have a real answer.

 

[dullard]

If SKL-X has full-speed AVX-512, it should have 5 times the throughput of 7700K, since 2.5 * 2 = 5...Therefore, the most we can gather from the SiSoft result for the 7900X is that it supports AVX-512 instructions, which we already knew, but whether the AVX-512 unit is full-speed or not is still unknown.

I asked because just last week these threads were saying over and over that it did not support AVX-512.

 

[Timmah!]

SS Multimedia is for the most part a vector unit test. AVX-512 is literally double the throughput...

Will this be reflected in real world apps or at least in something like Cinebench?

 

[blue11]

I asked because just last week these threads were saying over and over that it did not support AVX-512.

It was being said that low-end Skylake-SP and Skylake-X would not have full-speed AVX-512. In other words, they would accept the instructions but run them at half speed like AVX on Zen. Unfortunately, the SiSoft "multi-media" benchmark tests instruction latency and not throughput, so it is not clear how exactly AVX-512 is performing.

 

[LTC8K6]

IIRC, the 79XX have it and the 78XX don't? I'm pretty sure I read that somewhere.

http://ranker.sisoftware.net/show_r...d5e3daeadeeadbfd8fb282a4c1a499a98ffcc1f9&l=en

 

[wildhorse2k]

To me it doesn't make sense to support AVX-512 from Intels perspective on Core i9-7820X or Core i9-7800X as Intel may be planning to sell them at competitive prices (change of names, PCIe lanes) and based on frequencies and IPC they should perform better than Ryzens. With AVX-512 it could ask for more and not everybody needs that. But Core i9-7900X and Core i9-7920X are a different story. AVX-512 would allow Intel to keep higher prices.

 

[LTC8K6]

To me it doesn't make sense to support AVX-512 from Intels perspective on Core i9-7820X or Core i9-7800X as Intel may be planning to sell them at competitive prices (change of names, PCIe lanes) and based on frequencies and IPC they should perform better than Ryzens. With AVX-512 it could ask for more and not everybody needs that. But Core i9-7900X and Core i9-7920X are a different story. AVX-512 would allow Intel to keep higher prices.

Well, I'm not 100% certain, just to be clear. I can't find any reference to it now.

 

[Edrick]

To me it doesn't make sense to support AVX-512 from Intels perspective on Core i9-7820X or Core i9-7800X as Intel may be planning to sell them at competitive prices (change of names, PCIe lanes) and based on frequencies and IPC they should perform better than Ryzens. With AVX-512 it could ask for more and not everybody needs that. But Core i9-7900X and Core i9-7920X are a different story. AVX-512 would allow Intel to keep higher prices.

If Intel starts segmenting CPUs via AVX-512, I would really be pissed off. So the 7800 series would have fewer cores, fewer pcie lanes and gimped avx-512?

 

[Bouowmx]

Will this be reflected in real world apps or at least in something like Cinebench?

Up to doubling of throughput of AVX-512 will be reflected in real-world apps, or Cinebench, when the app's time-critical sections are updated to use AVX-512 instructions.