Question What is Skymont IPC equivalent too

MS_AT · Tuesday at 7:06 PM

DavidC1 said:
Whatever it is, discrepancies clearly exist in his tests, so he is missing something, and most of the time it's due to clocks.

I mean this

DavidC1 said:
13900K's P core gets 1.98, meaning it's just 6% faster than Zen 3, which is not the case.

Is you misreading the table. The P core is 25% faster than Zen3 according to total score.

DavidC1 said:
Golden Cove is actually if anything very slightly faster than Zen 4

Which is also reflected in the table when looking at total score.

What you refer to as performance (absolute) is actually PPC. Unless your intention was to compare PPC, but then your wording confused me.

Also SPEC is a funny benchmark, as it's compiled by the user so not only you have to care about the system but also about proper compiler flags, and then maintaining up to date database is hard as compilers are evolving etc.

Funny thing, C&C was displaying abnormally low SPEC scores for one test in the FP part of SPEC2017 when they started running the benchmark, turns out it was a problem in Linux kernel, https://www.phoronix.com/news/Intel-Linux-3888.9-Performance Not sure if they noticed.

Anyway what I meant to say with this, since the benchmark is so environment dependent, reading the scores, and comparing them among different people doing the benchmark is hard, as the score may differ and not because somebody is doing something wrong.

DavidC1 · Tuesday at 11:17 PM

MS_AT said:
I mean this

Is you misreading the table. The P core is 25% faster than Zen3 according to total score.

Which is also reflected in the table when looking at total score.

No, it has a section called Perf/GHz. First, Golden Cove is NOT 25% faster per clock, and second 13900K clocks much much higher than Zen 3. You are misreading the table, not me.

5.8GHz @ 11.5 on 13900K vs 4.9GHz @ 9.15 for 5950X.

MS_AT said:
Anyway what I meant to say with this, since the benchmark is so environment dependent, reading the scores, and comparing them among different people doing the benchmark is hard, as the score may differ and not because somebody is doing something wrong.

What I am saying has NOTHING to do with environment and comparing to other people. WITHIN his own scores it's not consistent.

Io Magnesso · Tuesday at 11:51 PM

What is the problem with David Huang's benchmark?
His site was interesting and I've seen it all the time.

Wolverine2349 · Wednesday at 12:00 AM

DavidC1 said:
No, it has a section called Perf/GHz. First, Golden Cove is NOT 25% faster per clock, and second 13900K clocks much much higher than Zen 3. You are misreading the table, not me.

5.8GHz @ 11.5 on 13900K vs 4.9GHz @ 9.15 for 5950X.

What I am saying has NOTHING to do with environment and comparing to other people. WITHIN his own scores it's not consistent.

Yeah Golden Cove not quite 25% faster clock for clock than Zen 3. How much faster is it then Zen 3 clock normalized?

Am I close on these estimates.

Like 15% to 17% I think.

And Rapor Cove is 5% faster clock normalized than Golden Cove?

And is Golden Cove like 2% to 3% faster than Zen 4 clock normalized?

And Zen 5 is 15% faster than Zen 4 clock normalized in non latency non AVX512 sensitive workloads?

Zen 5 then 8 to 10 percent faster than Raptor Cove in non latency sensitive workloads but slower in latency workloads like gaming without 3d cache to mask its bas infinity fabric latency?

And Lion Cove only a 3% faster than Raptor Cove in productivity NON latency sensitive workloads?

MS_AT · Wednesday at 2:12 AM

DavidC1 said:
No, it has a section called Perf/GHz. First, Golden Cove is NOT 25% faster per clock, and second 13900K clocks much much higher than Zen 3. You are misreading the table, not me.

Sorry, but I said I am comparing values in the Total column which the table is sorted against.

Then I said you might be comparing against Perf/GHz (PPC, performance per clock) but your wording made it consfusing for me.

Yet you tell me I was comparing Perf/GHz when I clearly was not.

And 11.5 / 9.15 is 1.256 so aprox 26% difference.

Question What is Skymont IPC equivalent too

MS_AT

Senior member

DavidC1

Golden Member

Io Magnesso

Member

Wolverine2349

Senior member

MS_AT

Senior member

TRENDING THREADS