Most accurate benchmark to compare relative CPU performance?

FunkTrains

Reputable
Jul 21, 2015
9
0
4,510
I spent a great deal of time recently researching a platform upgrade from AMD to Intel. In my quest, I researched gaming benchmarks from various sites, and meta-benchmarks like PassMark and UserBenchmark. I came to the conclusion that none of these metrics on their own seem to paint a picture that makes sense.

For instance in gaming benchmarks you sometimes have Skylake blowing out Haswell, and other times there's almost no difference. This occurs even when the test setups are very similar across different sites.

With PassMark, you have the FX-8350 doing as well as an i7 2700k, which most people regard as still being a viable option.

UserBenchmark shows that the i5 6600k has better single threaded performance than the 4790k, even though we know that can't be possible given the clock speed advantage Haswell has.

All of this research has left me confused as to whether or not any of these benchmarks are even useful in making an informed decision.

So how do you wade through all of this contradictory data? What benchmarks do you trust?


 
Solution
Personally I go with benchmarks per application. Passmark isn't of much use since performance varies from one program to another. Games are all different pieces of software and it would be hard to say cpu xyz is great for gtaV because it's good in photoshop. If I'm interested in playing gtaV I'd look for gtaV benchmarks, not passmark.

As you said no individual benchmark paints a clear picture. Myself, I try to compare several even for the same game. Obviously the setup test rig's will be a little different here and there but results should be similar all else equal. If two sites give similar fps for a particular game and a 3rd site's results are way off I start questioning why they get such unusual anomalies.

If you're interested in...
Personally I go with benchmarks per application. Passmark isn't of much use since performance varies from one program to another. Games are all different pieces of software and it would be hard to say cpu xyz is great for gtaV because it's good in photoshop. If I'm interested in playing gtaV I'd look for gtaV benchmarks, not passmark.

As you said no individual benchmark paints a clear picture. Myself, I try to compare several even for the same game. Obviously the setup test rig's will be a little different here and there but results should be similar all else equal. If two sites give similar fps for a particular game and a 3rd site's results are way off I start questioning why they get such unusual anomalies.

If you're interested in gaming specific performance I would look for gaming benchmarks. Try and find scenarios similar to yours, if you're gaming at 1080p look for 1080p benchmarks. If gaming at 1440p try and look for those. Synthetic benchmarks wouldn't give you the most accurate data for comparison. Raw performance doesn't always translate to higher fps.

For instance an overclocked cpu vs stock may perform better in video editing or photoshop but the differences between a stock and oc'd cpu in a game may only be 2-3fps. Technically the overclocked/faster cpu IS performing better and will show in some situations. It doesn't mean it will show in gaming performance necessarily.
 
Solution

FunkTrains

Reputable
Jul 21, 2015
9
0
4,510


And that's basically what I was trying to do. I was trying to get a large cross-section and make a decision based on all the available data and my individual use case. It just made me wonder about the integrity of many of these sites and benchmarks. I understand Passmark, but UserBench seemed pretty good until Skylake came out. In fact I've seen lots of weird Skylake results and websites avoiding a direct 4790k to 6700k comparison. Oh well, I guess that's just part of the fun of being a computer nerd. Thanks for the reply.