Did Passmark software with their new PerformanceTest 10 just degraded "older" CPU's???

[Zyll Goliat]

Not long ago PassMark software presented NEW PerformanceTest 10 well I downloaded just recently and notice significant difference in results compared to the their PerformanceTest 9

Both results are with the same CPU(2650V2)on EXACTLY same CPU&memory speeds...........guys from passMark software claimed that this is due to the major changes in the CPU test algorithms and that's for sure and it will be OK If the newer CPU's that have those instructions gain even more points but instead of that newer CPU's gain more points compared with the their older PerformanceTest 9 results but older CPU's get punished and lost points compared to the older PerformanceTest 9...... here some more comparison results that I found on net that some other people collected:

i7-2600---8181 5655 -----------around 30%
x5670 ----7800 7795 ---------------------0%
e5-2670--12374 10506----------around 20%
e5-1650--11736 11733--------------------0%
q9500-----3960 2961------------around 30%
i3-4130---4801 4196------------around 15%
i7-960-----5230 3858------------around 30%

and some more....

PT9 results (CPUMark, Single threaded result)
Intel Core i7-3820QM @ 2.70GHz 8,397, 1,844
Intel Core i7-3840QM @ 2.80GHz 8,759, 1,914
Intel Core i7-3920XM @ 2.90GHz 8,983, 1,963
Intel Core i7-3940XM @ 3.00GHz 9,133, 1,982

PT 10 results as at 14/Mar/2020, 3pm. (CPUMark, Single threaded result)
Intel Core i7-3820QM @ 2.70GHz 5,701, 1878
Intel Core i7-3840QM @ 2.80GHz 5,694, 1914
Intel Core i7-3920XM @ 2.90GHz 5,839, 1963
Intel Core i7-3940XM @ 3.00GHz 5,937, 1982

Also guys from Passmark posted this list:

Well I always liked performance tests from Passmark and I been using them for benchmark comparison for years....version changed in the past but results was never drastically different like this.... Is this something that you guys expected?What's your take on this?

If you want(for FUN sake)you can also DOWNLOAD and check both test and then post in here your results...THX....

 

[Zakin]

Just wanted to say this is a pretty common practice, it's why a lot of other tech websites are still running two different sets of the same program to keep consistency.

 

[Zyll Goliat]

Just wanted to say this is a pretty common practice, it's why a lot of other tech websites are still running two different sets of the same program to keep consistency.

O yeah...sure I know....but...still find this a bit "too much" compared to their previous versions differences...also it's a bit weird this heavy punishment and "taking"points to the older CPU's instead of just adding more points to the newer CPU'S....

 

[Zakin]

The only thing I find weird is most other programs will tend to make a major number change or a naming nomenclature change to the program, although I suppose version 9 to 10 at least makes it evident. But I've seen a lot of the 3D benchmark programs with wildly different number results when they change their methods, making it entirely uncomparable.

 

[windwhirl]

Passmark did quite a few changes in their CPU benchmark suite... so it's natural if scores change. CPUs are not measured the same way because the tests themselves have changed.

To be honest, this is common practice. AIDA64 tells you in every benchmark that different versions can't be compared, for example.

 

[Zyll Goliat]

Well I just downloaded very"old"PerformanceTest 8 and run the test...well...the result is almost identical and within the margin of error with the PerformanceTest 9....

So between this 2 versions it was literally almost no difference.....Just saying because Passmark CPU chart's was always fun to watch and compare the"raw"power between the CPU's through ages and now this seems like is going to change things a bit.......

 

[EarthDog]

Kind of OFN?

Passmark suddenly adding AVX512, intel ahead once again

https://nl.hardware.info/nieuws/71316/passmark-voegt-plots-avx512-toe-aan-performancetest-intel-weer-bovenaan https://www.planet3dnow.de/cms/55662-passmark-veraendert-cpu-benchmark-erheblich-zu-gunsten-von-intel/

www.techpowerup.com

Passmark is also MEH at best anyway. There was a discussion in some performance thread a guy started here.

 

[Zakin]

We also probably didn't have any major CPU instruction sets that hit between 8 to 9 I bet.

 

[Zyll Goliat]

We also probably didn't have any major CPU instruction sets that hit between 8 to 9 I bet.

Uff....don't bet on that...not sure but PerformanceTest 8 is very old probably close to the 10 years...

 

[windwhirl]

We also probably didn't have any major CPU instruction sets that hit between 8 to 9 I bet.

Apparently in version 9 (launched September 2016) Passmark added AVX and FMA at least. It's not clear what level of SSE they used.

Uff....don't bet on that...not sure but PerformanceTest 8 is very old probably close to the 10 years...

PT 8 was launched in October 2012. Almost a decade ago. The changelogs for almost all Performance Test versions are here: https://www.passmark.com/products/performancetest/history.php

 

[Zakin]

Doesn't sound like it used AVX or SSE fully, possibly because AMD at least until Ryzen didn't have all of those instruction sets, though it doesn't sound like Passmark cared about a level playing field either. Awkward.

 

[windwhirl]

Doesn't sound like it used AVX or SSE fully, possibly because AMD at least until Ryzen didn't have all of those instruction sets, though it doesn't sound like Passmark cared about a level playing field either. Awkward.

AVX was supported as early as Bulldozer. AVX2 was added in Excavator. SSE up to SSE 4.2 was supported in Bulldozer and later.

My guess is that Passmark added support for those instructions (AVX and FMA) because they became more widely used over time. Regarding SSE... I'm not sure which level they use, they don't really state it anywhere, but they supported at least some set of it far back enough that MMX still mattered.

 

[bobbybluz]

It hit one of my X79 1680 V2's hard. The CPU score dropped nearly 5000 points in version 10 vs version 9.

 

[Zyll Goliat]

It hit one of my X79 1680 V2's hard. The CPU score dropped nearly 5000 points in version 10 vs version 9.

Yeah seems like X79 is really punished harsh for some reasons and I don't get why because some X58 results that I saw seems like they are been much less % diminished......

 

[Arctucas]

The hit is huge for legacy OS, never mind the hardware.

PT always sort of wonky anyway.

 

[windwhirl]

PT always sort of wonky anyway.

I sort of agree with that, but it's nice to have another thing to compare with anyway.

 

[Fizban]

Not tried Version 10 yet, but that's version 9 results for my laptop. That 14915.5 CPU score is a i7-9750H.

Some scores are higher in Version 10 (Memory and 3D Graphics). The rest dropped.

Here's the actual breakdown for CPU showing scores per category.

Beats your Xeon on Integer Math, CPU Single Threaded, Floating Point Math, Extended Instructions, and Encryption but loses to the Xeon on Prime Numbers, Physics, Compression, Sorting, and Cross-platform Mark.

Seems 10 values certain results more than 9 did. In 9 your Xeon beat my laptops CPU, in 10 my laptop wins.

 

[R-T-B]

Not long ago PassMark software presented NEW PerformanceTest 10 well I downloaded just recently and notice significant difference in results compared to the their PerformanceTest 9

View attachment 158372
Both results are with the same CPU(2650V2)on EXACTLY same CPU&memory speeds...........guys from passMark software claimed that this is due to the major changes in the CPU test algorithms and that's for sure and it will be OK If the newer CPU's that have those instructions gain even more points but instead of that newer CPU's gain more points compared with the their older PerformanceTest 9 results but older CPU's get punished and lost points compared to the older PerformanceTest 9...... here some more comparison results that I found on net that some other people collected:

i7-2600---8181 5655 -----------around 30%
x5670 ----7800 7795 ---------------------0%
e5-2670--12374 10506----------around 20%
e5-1650--11736 11733--------------------0%
q9500-----3960 2961------------around 30%
i3-4130---4801 4196------------around 15%
i7-960-----5230 3858------------around 30%

and some more....

PT9 results (CPUMark, Single threaded result)
Intel Core i7-3820QM @ 2.70GHz 8,397, 1,844
Intel Core i7-3840QM @ 2.80GHz 8,759, 1,914
Intel Core i7-3920XM @ 2.90GHz 8,983, 1,963
Intel Core i7-3940XM @ 3.00GHz 9,133, 1,982

PT 10 results as at 14/Mar/2020, 3pm. (CPUMark, Single threaded result)
Intel Core i7-3820QM @ 2.70GHz 5,701, 1878
Intel Core i7-3840QM @ 2.80GHz 5,694, 1914
Intel Core i7-3920XM @ 2.90GHz 5,839, 1963
Intel Core i7-3940XM @ 3.00GHz 5,937, 1982

Also guys from Passmark posted this list:
View attachment 158374

Well I always liked performance tests from Passmark and I been using them for benchmark comparison for years....version changed in the past but results was never drastically different like this.... Is this something that you guys expected?What's your take on this?

If you want(for FUN sake)you can also DOWNLOAD and check both test and then post in here your results...THX....

It's probably using new instructions on the newer CPUs that the older ones don't have.

In other words, the new scoring is more accurate, not less.

We also probably didn't have any major CPU instruction sets that hit between 8 to 9 I bet.

Try several.

 

[natr0n]

I gave up on that software long ago. It's not very numa friendly unlike aida64's benchmarks.

 

[Zyll Goliat]

I gave up on that software long ago. It's not very numa friendly unlike aida64's benchmarks.

Sure I also using and prefer Aida 64 it's just that online Passmark CPU CHART is easy to check and compare and it was the same methodology/scoring for ages(PT7,PT8,PT9) and I already have in my head how much points have core 2 duo or AMD FX,I7 920....etc.....and now that's just going to change....and again if they just add more points to the newer CPU's instead of taking the points to the old + adding more points to the new.....

 

[Dirtdog]

4790K here, 12861 vs 9115, 30% drop.