Anyone else not getting the 64-bit reviews?

Unperson

[H]ard|Gawd
Joined
Sep 2, 2000
Messages
1,826
So far i've read 2 [H] reviews / tests regarding 64-bit gaming linked on the [H] main page and it just isn't making sense to me. It seems that the reviewer is maxing out the video card settings to the highest playable values before running the CPU tests. Does anyone else see a flaw in this logic? It has been my understanding in the past however many years that the best way to test CPU capability in games is to reduce the limiting factors of the video card to the point where the CPU becomes a limiting factor. Things like Anti-aliasing and anisotropic filtering are limited by the card's capability, not the CPU. Wouldn't a better test be to crank it way down to low 800x600 or 640x480 with 0 AA and 0 AF and see how many frames it does?

I could understand if the logic were simply to test in a "real world" environment, but I think what people are probably interested in is whether or not the 64-bit capability will make the difference when the CPU is the limit. For instance, should you buy a P4 or Athlon 64 if they are the same price?
 
But that's measuring what's playable!!!11

If you like the classic types of benchmarks, there are other sites. I didn't read the [H] 64-bit gaming review.
 
Pyromaneyakk said:
So far i've read 2 [H] reviews / tests regarding 64-bit gaming linked on the [H] main page and it just isn't making sense to me. It seems that the reviewer is maxing out the video card settings to the highest playable values before running the CPU tests. Does anyone else see a flaw in this logic? It has been my understanding in the past however many years that the best way to test CPU capability in games is to reduce the limiting factors of the video card to the point where the CPU becomes a limiting factor. Things like Anti-aliasing and anisotropic filtering are limited by the card's capability, not the CPU. Wouldn't a better test be to crank it way down to low 800x600 or 640x480 with 0 AA and 0 AF and see how many frames it does?

I could understand if the logic were simply to test in a "real world" environment, but I think what people are probably interested in is whether or not the 64-bit capability will make the difference when the CPU is the limit. For instance, should you buy a P4 or Athlon 64 if they are the same price?


but that would be logical. :D
 
He was testing logically.

If he maxed the gfx card's known performance, the only thing that'd increase performance would be the cpu..

~Adam
 
CleanSlate said:
He was testing logically.

If he maxed the gfx card's known performance, the only thing that'd increase performance would be the cpu..

~Adam

Actually, unless i'm wrong (maybe someone else can verify) I believe that is incorrect. If you are maxing out the video card, the video card becomes the limting factor. Therefore no matter how fast the rest of the parts in the machine are, the framerate will never increase. You could have a multi-threaded game with 8 dual-core opterons and 16 channels of DDR 4200 and it still would not increase.

Then again it could be a misunderstanding of words. By "max out" the video card, I mean put so much stress on the video card itself that it becomes the limiting factor.
 
when a game gets so old and the video card cranks out so much power compared to that video game. it will become cpu limted. and then you can test the cpu strength.

real people dont do that, they pair a highend chip with a highend card. and make sure there arent any flaws. or find a game that utilizes it.


of course throw all juck out the window, and play unreal tournament in software rendering mode :)
 
Pyromaneyakk said:
Actually, unless i'm wrong (maybe someone else can verify) I believe that is incorrect. If you are maxing out the video card, the video card becomes the limting factor. Therefore no matter how fast the rest of the parts in the machine are, the framerate will never increase. You could have a multi-threaded game with 8 dual-core opterons and 16 channels of DDR 4200 and it still would not increase.

Then again it could be a misunderstanding of words. By "max out" the video card, I mean put so much stress on the video card itself that it becomes the limiting factor.

when you test a CPU, you usually shut everything off, and turn the resolution down to 800x600, then run your tests. This takes the limiting factors off every other component and simply allows just the CPU to show its work and speed. This is what was done for the very large CPU chart present to show CPU influence in doom3.

That said, the reason why it would be meaningless to do this in the farcry review is because they had a goal to show a difference of the new visuals and how they impact performance under 64bit support. I agree they could of done both, but the simplest thing a gamer will want to know is what improves and what takes a hit. They make it clear. You gain visuals, some new neat things, and you dont lose any performance. I agree seeing a simple % increase under a common factor would of been interesting as well if it were possible.
 
Ocean said:
when a game gets so old and the video card cranks out so much power compared to that video game. it will become cpu limted. and then you can test the cpu strength.

real people dont do that, they pair a highend chip with a highend card. and make sure there arent any flaws. or find a game that utilizes it.

What ever happened to testing for the sake of testing? Maybe i'm too "old fashioned" but i've always thought that if you want to test how a piece of hardware or a configuration change stands up against another, the best way to do it is to eliminate all other factors as the cause of performance drop. I think it's kind of similar to something called the "scientific method" or a "controlled experiment".

To me that's like trying to do a review of a hard drive, meanwhile encoding videos and playing games in the background. Those tasks obviously will affect the transfer rate of the drive because they cause the CPU and RAM to become overloaded, but that doesn't necessarily reflect the true performance of the drive. If you shut down the rest of the programs and just test the drive, you will know how the drive itself performs. Adding those tasks in the background may be a good simulation of real world use, but in the end you get no real hard numbers on the drive.

Another analogy would be a car. Maybe you want to improve your performance. Someone tells you "a turbocharger, exhaust system, nitrous system, and modified comptuer will add 200 horsepower". But that doesn't help much unless you have the money to buy them all at once, because you do'nt know what each individual component does.

of course throw all juck out the window, and play unreal tournament in software rendering mode :)

If only Unreal Tournament had a 64-bit version, that would be the perfect test (then again, as i'm sure will be the case for a long time to come, the difference between 32-bit and 64-bit executables will depend on how well the developer implements the 64-bit support)
:eek: :D
 
FanATIc said:
when you test a CPU, you usually shut everything off, and turn the resolution down to 800x600, then run your tests. This takes the limiting factors off every other component and simply allows just the CPU to show its work and speed. This is what was done for the very large CPU chart present to show CPU influence in doom3.

That said, the reason why it would be meaningless to do this in the farcry review is because they had a goal to show a difference of the new visuals and how they impact performance under 64bit support. I agree they could of done both, but the simplest thing a gamer will want to know is what improves and what takes a hit. They make it clear. You gain visuals, some new neat things, and you dont lose any performance. I agree seeing a simple % increase under a common factor would of been interesting as well if it were possible.

Maybe i can do the test for myself. I personally do not have a 64-bit rig but one of my friends just put one togehter (A64 3500+, Radeon X800 Pro) and I believe he will be getting Win XP 64 any day now.
 
i only glanced over the review, but wasnt the puupose to check overall system performance, a system with a cpu running 32 bit with a 32 bit os with 32 bit drivers and 32 bit games vs. a system with a cpu running 64 bit with a 64 bit os with 64 bit drivers and 32 bit games, but nonetheless i do agree there should have been benchmarks done with video settings on the bare min.
 
The focus of those articles was on gameplay experience.

The question is does WinXP x64 with an Athlon64 improve the gameplay experience?

The articles answered that.
 
But it isn't all that useful when you install Quake 3 and play it at the minimum settings at 640x480 and you're measuring the fps in "hundreds of". Especially as the numbers will probably be wildly different. Just for a laugh, I did just this with my rig (see sig) which isn't exactly uber, and the demo ran so quickly it was silly.

What is needed is (as someone says) a UT (or any other no-AI 3d game really) with software rendering to eliminate the card but then that doesn't reflect on the real world (no one will play a game on software rendering nowadays), so really you're stuck. There is no catch-all answer.

A happy medium needs to be found, and in my opinion [H] hasn't got it quite right. I'd like to see less emphasis (but not zero) on the card in cpu reviews.
 
Brent_Justice said:
This and This were not a CPU review.

Brent do some serios 64bit vs 32bit farcry benchies. This artical was a letdown once i saw you weren't actually showing CPU benchmarks, even tho you state you were looking for performance differences in your article.

Can we get some 640x480 low graphics benchies?
 
Nobody (in their sane mind at least) plays in 640x480. That's just stupid if you ask me (not to mention a waste of time and energy). You want those types of reviews? Go somewhere else where you can certainly find that type of garbage. It's reviews like these that make [H] stand out from the rest. :)
 
1c3d0g said:
Nobody (in their sane mind at least) plays in 640x480. That's just stupid if you ask me (not to mention a waste of time and energy). You want those types of reviews? Go somewhere else where you can certainly find that type of garbage. It's reviews like these that make [H] stand out from the rest. :)

My point is not about anyone playing at those resoultions, it's whether or not the CPU will be faster in 32-bit VS 64-bit . The "real-world" tests show you that the CPU is at least as good as in 32-bit when it is not a limiting factor of system performance, but they do not prove whether or not 64-bit is faster.

In other words, it proves that 64-bit is not slower than 32-bit, but does not do anything to prove that it's faster.

If someone told you that a new high-performance sports car was not slower than a honda civic, wouldn't you say "well how fast is it?"

If resolution in a CPU test concerns you so much, the other way to do it would be to massively under-clock the CPU. That way you could play at your 2048x1536 and still have the CPU as a limiting factor (not sure A64 clocks that slow though :-P)

How many people do you see testing a Geforce 6800 Ultra card, whether it's a review of the card, or new drivers, or something else, on a Sempron 2400+, or running Prime 95 torture test in the background with normal priority? None. That's because they are putting the card in an environment where the change in the driver / card / whatever is the limiting factor. Again, it's a controlled experiment. I don't think I would be wrong if I said a lot of people would be interested to see not necessarily what real-world performance is, but if the 64-bit makes a difference when the CPU becomes a limiting factor.
 
In reply to 18. Thats very true, any seriosly who in their right mind was worried that code optimized for 64bit was going to run slower??? Thats crazy thinking.

The article as i see it was to show what the graphical benifit there was in the 64bit version, not because its "64bit" but because CryTek added some things.. Fair enuf. But on the most interesting aspect of the test, the benchmarks fail to show anything atall except that running cards 1600 will result in them being a bottleneck.

In reply to 17, link me a correct 64bit vs 32bit FarCry bench and i'll happily go read it?
 
Wow.

He wanted to show if there was any difference or reason to update to the 64 bit version. I honestly doubt your going to be playing farcry with your fancy hardware at 800x600. So what if the CPU at the resolution cranked up the FPS by 20. If it doesnt translate to higher performance at higher resolutions then there's no reason.

Not hard.
 
Nemon said:
In reply to 18. Thats very true, any seriosly who in their right mind was worried that code optimized for 64bit was going to run slower??? Thats crazy thinking.

The article as i see it was to show what the graphical benifit there was in the 64bit version, not because its "64bit" but because CryTek added some things.. Fair enuf. But on the most interesting aspect of the test, the benchmarks fail to show anything atall except that running cards 1600 will result in them being a bottleneck.

In reply to 17, link me a correct 64bit vs 32bit FarCry bench and i'll happily go read it?

They are 32bit vs 64bit reviews - the 32bit and 64bit sections are just not condensed into one article. The point of these articles, as other posters have said, was not to test the ability of the CPU. The ability of the CPU has already been tested by the relevant 32bit benchmarks. The recent 64bit benchmarks only aimed to test the new features added to Crytek's 64bit engine while looking for performance drops or errors. They don't need to do a full review because they're esentially just "making sure' Crytek didn't royally mess something up while compiliing their 64bit executables.

64bit also entails increased memory usage so that could cause a performance drop on ill-equipped machines. You don't get that extra floating ponit accuracy for free...
 
I'll accept that. But what benchmark actually benches a real world game in 32bit and 64bit modes? I've not seen one and admitedly this article was the first one that caught my eye.

There is a similar FarCry64 review on AT that runs at 10x7 HD and shows a 7FPS improvment. Reviewer states that 10x7 HD on an X850 and its CPU bound, so im kinda happy with that.
 
Thanks, good artical, many variations.

Looks like nVidia drivers arnt up to par with ATIs atm.
 
Nemon said:
Brent do some serios 64bit vs 32bit farcry benchies. This artical was a letdown once i saw you weren't actually showing CPU benchmarks, even tho you state you were looking for performance differences in your article.

Can we get some 640x480 low graphics benchies?

It wasn't a CPU review and was never meant to be. It was from a gamers perspective. I'm sorry you don't understand what relating the gaming experience means.

Thanks for the feedback though.
 
Brent_Justice said:
It wasn't a CPU review and was never meant to be. I'm sorry you don't understand what relating the gaming experience means.

Thanks for the feedback though.

Can you explain that for me or were you just trying to apear condescending?
 
In those articles we aren't reviewing the CPU, we aren't reviewing the video cards.

What we are reviewing is the gameplay of the games tested in 32-bit Windows and then 64-bit Windows from a gamers perspective relating performance and image quality from that angle. Our goal was to find out if there were any real world tangible benefits or downsides from upgrading to 64-bit Windows. We looked at installation, compatibility and performance. The FarCry article's goal was to see if there were any real world tangible differences from a gamers perspective running the 32-bit version vs. the 64-bit version.

We used our method of highest playable gameplay quality settings that we find is playable to determine the best combination of resolution, AA and AF with the best playable performance in the game. We've been doing this for a very long time now and it shows what gamers want to know, are there any real world tangible benefits between video cards in games, and thus in these articles are there any real world benefits with the 64-bit version of FarCry or any differences in 64-bit Windows with 32-bit games.
 
Thank your for explaining that to me.

I guess im looking at this from a complete other angle and this article wasnt intended to have the information i was after. I mean 64bit vs 32bit at 16x12 high settings, i wouldnt expect to see much if any change atall, which kinda renders the bench part of the article a bit of a "duh". Perhaps time was a factor, but i'd look forward to reading future articles more if they had more variety to the benchmarks. Just a bit of friendly feedback.
 
Nemon said:
Thank your for explaining that to me.

I guess im looking at this from a complete other angle and this article wasnt intended to have the information i was after. I mean 64bit vs 32bit at 16x12 high settings, i wouldnt expect to see much if any change atall, which kinda renders the bench part of the article a bit of a "duh". Perhaps time was a factor, but i'd look forward to reading future articles more if they had more variety to the benchmarks. Just a bit of friendly feedback.

I come to [H]ardOCP because of the way they benchmark.
Wich I find more informative for me, than other test that show 732923 FPS in Quake 3...
If they change their benchmark style, I will leave too...

Terracide - You play games for FPS...or the IQ? ;)
 
Thanks for the feedback

Nemon said:
I guess im looking at this from a complete other angle and this article wasnt intended to have the information i was after.

I can respect that.
 
Terra said:
Terracide - You play games for FPS...or the IQ? ;)

I play games for fun.

FPS or IQ, that would depend on the game. I wouldnt accept anything under 60fps and if it was fast paced probably 80.

I dont want them to change something that as Brent already said is what people like. I would've like them to supplement atleast this artical with with a little more bench info. I have no idea if benches are conducted the same in all reviews, i only check reviews while i'm shopping.
 
Back
Top