Why is there a fairly large fps gap with certain cards between reviewers?

[dave343]

I've been looking at a few RX 580 reviews, and I noticed on guru3d the RX 580 MSI review has the 580 on par with the 1060 on GTA V. But over on Anand tech, the 580 trails far behind the 1060 on GTA V with what seem's like the same "very high" settings, but the 1060 is ahead by 15fps. If you're looking to purchase a card based on gtav, the 1060 would sell itself in this benchmark, but go read guru3d and you get a different picture.

 

[luke51087]

Would think the difference in the rest of the machine, cpu/ram. Also sometimes drivers. Also the very high and apples to apples testing is hard to say unless they list out ALL of the graphics options on both sites.

 

[dave343]

Would think the difference in the rest of the machine, cpu/ram. Also sometimes drivers. Also the very high and apples to apples testing is hard to say unless they list out ALL of the graphics options on both sites.

So Anandtech is actually using a 4960X @ 4.2ghz with DDR3 ram, and Guru used a 5960X @ 4.3ghz with DDR4 @2400, and yet Anandtech's gtx 1060 pulls ahead by 15fps, so I can't see this being in the CPU department, and I have high doubts seeing as how new the RX580 is it would be int he driver.

 

[Araxie]

there's no way one review or another can replicate results unless everything is tested under a canned benchmark, some review sites utilize full game experience, they just choose randomly where they want to perform their benchmarks and for how long, other's just use a pre-made ingame tool, those will give show a big difference in results, different part or scenarios from the games show always different results and some favor one or another vendor. also the hardware used show an appreciable differentiation in performance, CPU used, RAM, Storage but more important than all of this is "Time", some review just do about 30 secs - 1 minute of "gameplay" to collect data, others choose more extended periods 5, 10, 15 minutes which certainly have the biggest factor in a review as it cover more gaming itself, other reviewers just do a testing with a single card then they just add the data to older data collected.

about GTA, GTA is a pretty hard beast to dominate at the moment of making a benchmark, because the game always behave different. so the area the reviewer select to do the benchmark play the biggest role in the end result, this is one of the reason why you should never compare one review to another and never extrapolate results from one review to another even when they are made by the same reviewer.

 

[Armenius]

I've been looking at a few RX 580 reviews, and I noticed on guru3d the RX 580 MSI review has the 580 on par with the 1060 on GTA V. But over on Anand tech, the 580 trails far behind the 1060 on GTA V with what seem's like the same "very high" settings, but the 1060 is ahead by 15fps. If you're looking to purchase a card based on gtav, the 1060 would sell itself in this benchmark, but go read guru3d and you get a different picture.

View attachment 22417 View attachment 22418

Anandtech is using 4x MSAA, while Guru 3D is using 2x MSAA. That alone can explain the difference. Anandtech says they turn all in-game settings to the highest available, while Guru 3D only says Very High without any additional description.

 

[Parja]

/thread

 

[Archaea]

Also power save settings, if enabled, can make a big difference.

or this:

 

[bigdogchris]

I look at reviews as comparing percentage better/worse than other cards using the same test platform. Say if 10 sites said the card was 10% faster than <whatever>, but one site said it was 15% slower than <whatever>, then I would question the results/testing methodology.

 

[cageymaru]

The only way to review GTA V would be to find the areas that are most demanding on the system in the game and drive from area to area doing the same actions each run. Most reviews on the internet just run the in game benchmark and proclaim one card is better than another. Most video card manufacturers optimize for the benchmark to look good to reviewers.

That's why I hangout here on [H]ardocp. Everything here is real world game play. In Brent_Justice we trust.