SkylakeX teasers (aka prime95 29.5)

[ATH]

Originally Posted by petrw1

I am disappointed that the Benchmark for this is no better than my son's i7-6700 running at 3.4 with much slower RAM(2400 I recall)

Anyone else with similar hardware have any comments?

When I look at Benchmarks for 7820s I don't see a big difference.

https://www.mersenne.org/report_benc...Get+Benchmarks

Mine is running 3990Mhz.

AVX-512 should be faster try the 29.5 beta, but watch your CPU temperature that your cooling can handle it. According to Mysticial AVX512 makes the CPU run much hotter.

In the benchmark window I assume you chose Throughput benchmark, and make sure "Benchmark all FFT implementations to find the best one for your machine" is checked. There can be a big speed difference between the different FFT settings on the same FFT size.

Make sure your RAM is running dual channel, see CPU-Z Memory tab it says "Channel #" Dual:
https://www.cpuid.com/softwares/cpu-z.html

[Mysticial]

Originally Posted by ATH

AVX-512 should be faster try the 29.5 beta, but watch your CPU temperature that your cooling can handle it. According to Mysticial AVX512 makes the CPU run much hotter.

At the same frequency as non-AVX/AVX, yes it will be much hotter. But AVX512 was never meant to run without a throttle.

These chips are all supposed to be have the same TDP regardless of load, non-AVX/AVX(512). They are normalized to the same TDP by offsetting the frequencies of each workload respectively.

The only problem is that most people don't understand this. (even the mobo vendors got it wrong initially) They try to run everything at the same speed and of course things blow up. Then they blame it on Intel because AVX512 is stupid or something like that.

[TheJudger]

Originally Posted by Mysticial

These chips are all supposed to be have the same TDP regardless of load, non-AVX/AVX(512). They are normalized to the same TDP by offsetting the frequencies of each workload respectively.

The only problem is that most people don't understand this. (even the mobo vendors got it wrong initially) They try to run everything at the same speed and of course things blow up. Then they blame it on Intel because AVX512 is stupid or something like that.

So true. Throttle is when frequency is below the base frequency (which can differ for different workloads (read: non-AVX, AVX and AVX-512 for current Intel CPUs). But consumers and boardmakers think it should always run at the maximum single core non-AVX speed. Same with GPUs - if they didn't run at maximum specified frequency they call it throttle and the product gets bad reviews...

Oliver

[VictordeHolland]

The problem is also partly due to marketing and expectations.
"This chip has a max turbo of 4.x GHz" _{(but only in single thread load)}
xxxW TDP is only when it runs at it's base frequency (it can consume a lot more in turbo modes, see also: https://www.anandtech.com/show/13400...600k-review/21

[Prime95]

Originally Posted by ZK19

Hi. I have a problem with prime95 29.5b3 linux build to return a certain result

Got assignment _: PRP M40493

overall the line is a few characters >1029 in length..

I changed the SQL table to accept result_text up to 2048 characters

[Prime95]

Originally Posted by petrw1

I am disappointed that the Benchmark for this is no better than my son's i7-6700 running at 3.4 with much slower RAM(2400 I recall)

In the BIOS, make sure your RAM is running at the fastest speed (XMP profile).

For throughput benchmarks, I've found my AVX-512 CPU gets much more throughput using 8 cores/8 workers rather than 8 cores/1 worker.

[petrw1]

I've no experience in reading these but it doesn't look like the RAM is 3600 even though I watched the support guy spend 2 hours in the BIOS until he got it to say the RAM was 3600 and stable…. what do you see?

The "SPD" was the same in Slots 1, 3, 5, and 7.

Thanks

[ATH]

So on the SPD tab you can see the RAM modules are actually DDR4-2133, the 1066 Mhz is the clock rate and 2133 Mhz is the "effective" speed since they can transfer data twice per clock. You are running Quad channel RAM which is great, the best you can get without buying very expensive Xeon processors.

On the Memory tab it shows that RAM is running overclocked at ~ 1900 Mhz actual clock rate so "effective" rate of 3800 Mhz, and your processor is slightly overclocked to 3.7 Ghz (CPU tab).

I hope you only payed for DDR4-2133 (PC4-17000) modules and not something higher? The question is if it is running stable at 3800 Mhz, you will have to test that with stress test and double checks.


It is strange you are getting bad benchmarks with this setup, it could be throttling because the CPU is getting too hot, try watch the CPU temperature during benchmarks. Also try to benchmark with the 29.4b8 Prime95 (with AVX2 instead of AVX-512), and check the temperature running AVX2.

[petrw1]

Heat should not be an issue I hope with a "Liquid 360mm Freezer 360 cooler"

[ATH]

So you got 7.89 ms/iteration at 3072K FFT but how many cores is that? What are you getting with 8 cores on 1 worker? Should get < 3ms then at 3072K FFT

[petrw1]

Originally Posted by ATH

So you got 7.89 ms/iteration at 3072K FFT but how many cores is that? What are you getting with 8 cores on 1 worker? Should get < 3ms then at 3072K FFT

Someone can correct me but I believe the Benchmark timings are for 1 core/1worker.

Timings for 3072K FFT length (8 cores, 1 worker):  1.38 ms.  Throughput: 723.83 iter/sec.
Timings for 3072K FFT length (8 cores, 4 workers):  7.29,  7.23,  7.18,  7.01 ms.  Throughput: 557.47 iter/sec.
Timings for 3072K FFT length (8 cores, 8 workers): 14.50, 14.40, 14.33, 14.36, 14.34, 14.36, 14.41, 14.06 ms.  Throughput: 557.74 iter/sec.