Interpreting PassMark Results

[tagadvance]

According to cpubenchmark.net the i7-3930K has an average CPU mark of 13,576 while the FX-8150 has an average mark of 8,166. Assuming 100% load across all cores, will the i7-3930K really be 60% faster than the FX-8150, or am I misinterpreting the results?

 

[jaguarskx]

No, I don't think so.

The best way to interpret PassMark results is to ignore them. Synthetic benchmarks have a tendency to be inaccuracy... and that's on a good day. On a bad day, they are grossly incorrect.

The best way to find out how well CPUs performs against one another is to read reviews and look at the various benchmarks. Anandtech has a quick benchmark database, just select the two CPUs that you are interested in.

http://www.anandtech.com/bench/CPU/2

See following reviews for details and more info:

http://www.xbitlabs.com/articles/cpu/display/core-i7-3770k-i5-3570k.html

http://www.bit-tech.net/hardware/cpus/2012/05/01/intel-core-i5-3570k-cpu-review/1

http://hexus.net/tech/reviews/cpu/38421-intel-core-i5-3570k-22nm-ivy-bridge/

http://www.madshrimps.be/articles/article/1000289/Intel-Ivy-Bridge-i5-3570K-i7-3770K-Review#axzz24EDMiDiS

http://www.guru3d.com/article/core-i7-3770k-and-3750-review-with-z77/

 

[PassMark]

I just cross checked a few results.

The Intel chip was benchmarked as follows on CPU based tasks,

SysMark 2012: 61%
3DSmax R9: 47%
Cinebench R10: 50%
Excel 2007: 77%
Visual Studio: 51%
PassMark: 66%

So the PassMark result looks accurate to me.

Having said that, different applications will preform differently on different systems. So it makes sense to look at a range of different benchmarks from different sources.

 

[tagadvance]

I forgot to reply after jaguarskx's post. I decided to go with the Intel. Thanks!

 

[rds1220]

Passmark is a joke and I wouldn't put much into their "benchmarks."I would look at benchmarks like Anandtech, Tomshardware and xbitlabs. The benchmarks they do are more accurate.

 

[amuffin]

I just cross checked a few results.

The Intel chip was benchmarked as follows on CPU based tasks,

SysMark 2012: 61%
3DSmax R9: 47%
Cinebench R10: 50%
Excel 2007: 77%
Visual Studio: 51%
PassMark: 66%

So the PassMark result looks accurate to me.

Having said that, different applications will preform differently on different systems. So it makes sense to look at a range of different benchmarks from different sources.

Yeah....thanks but no thanks "passmark" :lol:

 

[PassMark]

I would look at benchmarks like Anandtech, Tomshardware and xbitlabs. The benchmarks they do are more accurate.

Completely agree that all these sites do a great job. You should use them when you can.

The problem however is that,

- You can't easily & quickly reproduce the tests they use on your home PC, for comparison. My guess is that Tom's takes a couple of days per new CPU they benchmark. Most people don't have this kind of time and expertise to runs these tests at home. Further they don't have access to the same data sets used on the review sites.

- Their charts are often not up to date. I just had a look at Tom's CPU charts. At the moment there isn't any 2012 (or even 2011) chart.
http://www.tomshardware.com/charts/processors,6.html
(yes I know that newer results are available, spread around all over the place, but it is useful to have it consolidated into one page)

- These sites don't cover the full range of CPUs on the market. Tom's 2010 desktop chart included 33 AMD processors and 44 Intel products. At the moment there are more that than 1,600 different X86 CPU available. So only ~5% of the CPUs are covered. If you are interested in comparing one of the other 1500 CPUs then you need to look elsewhere.

- These sites don't offer any long term comparisons e.g. comparing a CPU from 2006 against a current 2012 CPU using the same test suite. For example, how to answer the question, exactly how much faster is Core i7-3770K compared to a AMD Sempron 2100+. You'd be surprised how many people research questions like this before an upgrade. For some people the answer that the i7 is 'way faster' isn't detailed enough.

- The sites almost never offer comparison between different classes of CPUs. e.g. Server vs Desktop vs Mobile vs Netbook CPUs. For example how does a Intel Core Solo U1400 compare to an VIA Nano U3100 to an AMD E-300 APU? Just because you (and me for that matter) aren't interested in these particular CPUs doesn't mean everyone is disinterested. Not everyone is a gamer.

There is a beta release of PerformanceTest V8 available here,
http://www.passmark.com/forum/showthread.php?3838-PerformanceTest-8-0-Beta-Release

It would be great to get some (constructive) feedback.

Again, in no way do we expect it to replace the benchmark / review sites mentioned above. It does, we hope, fill a hole and is reasonable accurate. Also keep in mind that different benchmarks return different results on different CPUs. The fact that benchmark A's results don't exactly match benchmark B's results doesn't make either of them invalid.

Disclaimer: In case it wasn't obvious, I am the author of the original PerformanceTest program.

 

[amuffin]

How come you list results such as these?

The FX-6200 outperforms the Phenom II X6 1100T?

 

[PassMark]

How come you list results such as these?

The FX-6200 outperforms the Phenom II X6 1100T?

At the moment the scores are
AMD FX-6200 is 6,304
AMD Phenom II X6 1100T 6,194
Which is a 1.7% difference. So more or less the same. They are too close to claim there is a meaningful difference.

The specs of the CPUs are,
AMD FX-6200 3.8Ghz, 4.1Ghz Turbo, 6 cores, 6MB L2 cache, SSE4
AMD Phenom II X6 1100T, 3.2GHz, 3.6GHz Turbo, 6 cores, 3MB L2 Cache, SSE3

So on paper the FX-6200 looks like a clear winner. But results from other benchmarks vary dramatically.

Some tests show the AMD FX-6200 is significantly faster (Sandra 2010 CPU Multimedia Integer test, Valve Map Compilation Test).
Other tests show the AMD FX-6200 it slightly slower (3DMark Vantage, Skyrim)
Still others show them being about on par. (Cinebench 10 - Single thread, PassMark)

There is also the issue of the Windows patches for thread scheduling inefficiencies when dealing with the Bulldozer architecture. Having these patches installed, or not, can move the results around a bit.

 

[tagadvance]

Best answer selected by tagadvance.