What is your opinion?

[mewi]

What is your opinion? On this kids opinion of passmark?

http://www.gamespot.com/forums/topic/28822061/need-help-with-gaming-rig.?msg_id=332545773#332545773

refer to swehunt's post

Can't you just look at the benchmarks, passmark has NO relevance because.
1) It's not a real benchamark.
2) It's NOT a gamebenchmark.
3) It's numbers are not a relative to the performance.
4) It's point's is nowhere relative to game performance.
5) IF you however want to know the performance in games, just click the link to anandtech, the have real games being tested on real CPU's.
6) Did I mention It's not a game benchmark?

Crysis, farcry 2, L4D, Fallout 3 ALL showed no gain from having a thuban over a X4. Clearly passmark is fooling with you.

[David (PassMark)]

Sounds like a rant to me.

But let's address it as if it was serious.

Point 1) & 3) are the same point and are just silly. Like saying a banana isn't piece of fruit.

Points above 2) 4) 5) and 6) are all just the same point repeated 4 times. Repeating it 4 times doesn't mean it is accurate however. The core of the criticism is that PerformanceTest doesn't reflect on game performance. This is also not true, with some qualifications below. The CPU and video charts larger reflect the type of gaming experience you are likely to have with any CPU / video card combination.

But what is true is that,
A) If you are only interested in playing Game X & Y, then in an ideal world you use these particular games as your benchmark. As they will best represent your usage of the machine.

B) You need to use your brain. So for example, looking at a disk benchmark doesn't make sense if you are running applications that don't use the disk. Same goes for looking at benchmarks that can run on 24+ cores if the application you are using is only single threaded.

C) PerformanceTest doesn't benefit greatly from SLI setups. Where as some games do in some configurations (particularly with very high resolutions). But the SLI performance boost in games is variable as well, depending on the game, the CPU in use & settings used.

Feel free to post this back to the Gamespot forum.

[mewi]

Thanks passmark for an insightful and clear post, it was difficult to reason with this person and I was running out of different scenarios to appease him lol.

Speaking of other websites like "anandtech.com" and "tomshardware" and "gaming benchmark websites" what is your opinion on them?

[David (PassMark)]

Anandtech and Tomshardware are both well known sites.
They do a good job in general. They do tend to be a little bit gaming focused. So for example if you were looking to buy a web server, database server or Xeon based server you aren't going to gain much insight from these web sites.

They also don't cover the entire market. For example, we have more than 1000 CPU types in our CPU charts. But Anandtech and Tomshardware combined might have covered only 200 CPUs combined. They focus on popular setups (often gaming) for the most part which is fine, if that is where your interest is.

PerformanceTest however lets you run your own benchmarks and compare to the community.

[mewi]

Anandtech and Tomshardware are both well known sites.
They do a good job in general. They do tend to be a little bit gaming focused. So for example if you were looking to buy a web server, database server or Xeon based server you aren't going to gain much insight from these web sites.

They also don't cover the entire market. For example, we have more than 1000 CPU types in our CPU charts. But Anandtech and Tomshardware combined might have covered only 200 CPUs combined. They focus on popular setups (often gaming) for the most part which is fine, if that is where your interest is.

PerformanceTest however lets you run your own benchmarks and compare to the community.

But under that technique wouldn't their benchmark tests be inaccurate since they only equate a single build or a couple builds per the hardware piece they are marking?

[David (PassMark)]

The Toms & Anandtech articles tend to look at a number of specific configurations. So for game X, on hardware Y, in configuration Z, I am sure the numbers they report are fairly accurate. And if you wanted to play game X, on hardware Y, in configuration Z, then their benchmark articles are definitely worth looking at. I read them myself from time to time.

If however on your PC you wanted to run a game or application that Anandtech didn't use as a benchmark, on a CPU that Anandtech didn't include in their article, then you would probably be better off using results from a general benchmark like PerformanceTest.

[mewi]

The Toms & Anandtech articles tend to look at a number of specific configurations. So for game X, on hardware Y, in configuration Z, I am sure the numbers they report are fairly accurate. And if you wanted to play game X, on hardware Y, in configuration Z, then their benchmark articles are definitely worth looking at. I read them myself from time to time.

If however on your PC you wanted to run a game or application that Anandtech didn't use as a benchmark, on a CPU that Anandtech didn't include in their article, then you would probably be better off using results from a general benchmark like PerformanceTest.

Personally I don't find myself trusting single result websites. Maybe if I was building an exact computer they were benchmarking, but I just can't see how it can be very accurate if I'm going to change something in it.

Amongst other things people read on one of these "gaming benchmark websites" that the i5-2500 is "twice as fast" as the "AMD PHenom II 1100T" perhaps you can enlighten me on that comment ... somehow? :S

[David (PassMark)]

i5-2500 is "twice as fast" as the AMD PHenom II 1100T

The i5-2500 is clearly a better CPU in pretty much every respect, except price.

How much better depends on the application you are running. If your application is single threaded there will be a big difference. If your application can use all 6 cores in the 1100T, the difference is more like 5% to 10%.