Something for Anandtech reviewers to consider

[yacoub]

Nevermind - some here would rather have nothing change and the majority of us not benefit from some more focused and beneficial reviews.

 

[CPlusPlusGeek]

I 100% agree, suggested this at guru3d.com

 

[crazydingo]

1280 x 1024 the 7800GTX etc wont pull ahead like its supposed to. It will show its power at higher resolutions under high settings.

 

[Lonyo]

It would be nice to see 9800/6600/6800GT/X800XL/7800/X1800 numbers in tests.
And also single core vs dual core setups to see how they work differently.

 

[yacoub]

Guess what? Nevermind!

 

[Megatomic]

Originally posted by: yacoub
Want to make it even better? Instead of using the FX-55 moster godbox you normally review on, make the test system with a 3200+ A64 Venice core, likely the most popular cpu right now.

This would help a lot as it gives an accurate idea how the fps would be for the majority of us (for once) because it uses a very common upper mid-range processor, allowing the GPUs to show their stuff on something resembling our actual systems.

Testing with a more mid-range clockspeed also means those of us with a 3000+ won't be seeing like half the fps of the review system. =P

Reviewers use high end CPUs so that graphics card being reviewed is not being held back. It allows the graphics card to show what it can really do better.

Perhaps they can run tests on a high end and a mid range system. But I wouldn't expect them to cut out the high end testing.

 

[yacoub]

Forget it...

 

[crazydingo]

Originally posted by: yacoub

Originally posted by: Megatomic

Originally posted by: yacoub
Want to make it even better? Instead of using the FX-55 moster godbox you normally review on, make the test system with a 3200+ A64 Venice core, likely the most popular cpu right now.

This would help a lot as it gives an accurate idea how the fps would be for the majority of us (for once) because it uses a very common upper mid-range processor, allowing the GPUs to show their stuff on something resembling our actual systems.

Testing with a more mid-range clockspeed also means those of us with a 3000+ won't be seeing like half the fps of the review system. =P

Reviewers use high end CPUs so that graphics card being reviewed is not being held back. It allows the graphics card to show what it can really do better.

Perhaps they can run tests on a high end and a mid range system. But I wouldn't expect them to cut out the high end testing.

The entire point is to make this accurate for those of us running such systems and to NOT make it the same irrelevant pure GPU test. We want a real world test, not a science lab experiment to see which GPU without any other restrictions is purely the fastest. Anandtech already does that. What we want is a review to see which GPU would best compliment a real world system based around a 3200+ Venice chip.

Actually that is a good point. I've seen articles on how games scale with different CPUs and one graphics card. They were very helpful.

 

[fbrdphreak]

I pretty much agree with what everyone said here

But I'm sure they will do a nice roundup when FEAR and Q4 come out, as they did with D3 & HL2

 

[lifeguard1999]

Haven't they already done it? ATI's X1000 Series: Extended Performance Testing The article shows everything from the X1300/6600GT up to the X1800XT/7800GTX with a myriad of resolutions and AA/AF settings.

 

[yacoub]

Why did I bother? Shouldn't have.

 

[Fern]

While I understand and agree with many of your points, seems to me in a "gfx card review" the point is to focus on the gfx card. So yeah, they wanna run it with a top-shelf CPU and at high rez's.

But that's where this forum can shine. It fills the void. Just make a post asking peeps with 3200 Venice and a X800 what FPS they are getting at certain rez's and you'll get plenty of answers from us "local reviewers" with common components.

Fern

 

[CP5670]

I agree that previous generation cards should be tested alongside the new ones, so people thinking of upgrading can get an idea of how much better the new card is. However, the 20x15 benchmarks should stay. There are very few sites that test stuff at that resolution and AT has recently become one of them, which is actually one of the major attractions of AT video card reviews for me these days.

The main thing I would like to see added in these articles is minimum framerate readings. I think these are much more important than the averages that most sites show.

 

[yacoub]

He doesn't get it either. Typical around here lately, it seems.

 

[Kogan]

I think I suggested something like this before.. Really though, not many review sites have time to benchmark 20 cards at 6 different resolutions in 10 different games.

It would be good if they had some sort of base system set up and then throughout a 1-3 year period, only use that system to benchmark video cards. Then whenever a new card comes out, they just add it's numbers to the chart.

But wait, tomshardware has already done something like that: http://graphics.tomshardware.com/graphic/vga_charts.html

Hopefully they expand it since that's about the only useful part of THG

 

[Fern]

The main thing I would like to see added in these articles is minimum framerate readings

Yup, I'm in big-time agreement with this.

I don't care that much for just the average either. Minimum is what is important.

 

[Jeff7181]

Originally posted by: yacoub

Originally posted by: Megatomic

Originally posted by: yacoub
Want to make it even better? Instead of using the FX-55 moster godbox you normally review on, make the test system with a 3200+ A64 Venice core, likely the most popular cpu right now.

This would help a lot as it gives an accurate idea how the fps would be for the majority of us (for once) because it uses a very common upper mid-range processor, allowing the GPUs to show their stuff on something resembling our actual systems.

Testing with a more mid-range clockspeed also means those of us with a 3000+ won't be seeing like half the fps of the review system. =P

Reviewers use high end CPUs so that graphics card being reviewed is not being held back. It allows the graphics card to show what it can really do better.

Perhaps they can run tests on a high end and a mid range system. But I wouldn't expect them to cut out the high end testing.

The entire point is to make this accurate for those of us running such systems and to NOT make it the same irrelevant pure GPU test. We want a real world test, not a science lab experiment to see which GPU without any other restrictions is purely the fastest. Anandtech already does that. What we want is a review to see which GPU would best compliment a real world system based around a 3200+ Venice chip.

That's why there's a forum here for people to share their results with different parts. When one of the crew at AnandTech does a video card review, they're not reviewing processors, they're not reviewing RAM, they're not reviewing hard drives, they're not reviewing network controllers, they're not reviewing sound cards... they're reviewing the video card only. To do that, you need to eliminate as many other variables as possible. If they ran 1024x768 with 512 MB of RAM with an A64 3200+ a 7800GTX might perform the same as a 6800 Ultra. As we all should know, the 7800GTX is most definately faster than the 6800 Ultra. But if you make the test CPU dependant, or limited by the amount of RAM, you get a false report on the video card.

*EDIT* Also, since most of these reviews are done using samples from manufacturers that sometimes must be returned, it makes it impossible to keep testing old cards. Not to mention a lot of extra work to test 6800 Ultra's and 6600GT's and 9800 XT's and FX5950's. You can't just transfer the results from a previous benchmark to a new one either because if the test was done with a sample motherboard for example, that motherboard may have been returned before the new video card review, so the results would be slightly different with a different motherboard.

 

[Sunrise089]

100% agree with OP. Even though cards like the X800XL and 9800pro have been reviewed before, for someone purchasing a card NOW, how the card ran in Doom 3 or HL2 is less important than how it runs FEAR or COD:2, so adding these cards to new benchmarks would be really appreciated.

 

[yacoub]

Hey look, another one!

 

[Keysplayr]

Originally posted by: yacoub

Originally posted by: Jeff7181
That's why there's a forum here for people to share their results with different parts. When one of the crew at AnandTech does a video card review, they're not reviewing processors, they're not reviewing RAM, they're not reviewing hard drives, they're not reviewing network controllers, they're not reviewing sound cards... they're reviewing the video card only. To do that, you need to eliminate as many other variables as possible. If they ran 1024x768 with 512 MB of RAM with an A64 3200+ a 7800GTX might perform the same as a 6800 Ultra. As we all should know, the 7800GTX is most definately faster than the 6800 Ultra. But if you make the test CPU dependant, or limited by the amount of RAM, you get a false report on the video card.

*EDIT* Also, since most of these reviews are done using samples from manufacturers that sometimes must be returned, it makes it impossible to keep testing old cards. Not to mention a lot of extra work to test 6800 Ultra's and 6600GT's and 9800 XT's and FX5950's. You can't just transfer the results from a previous benchmark to a new one either because if the test was done with a sample motherboard for example, that motherboard may have been returned before the new video card review, so the results would be slightly different with a different motherboard.

You make this too easy when you don't even bother comprehending what's already been written. I can copy and paste most of my answer for you.

You're misinterpreting my request. I'm not asking for them to REPLACE current reviews with this review. I'm asking for this review as an additional review. You're right it's not a typical video card review - it's not supposed to be!

The point of this additional review is to help the reader identify which GPU will run the latest games smoothly at their resolutions (most commonly 1280x and 1600x) most cost-effectively for their 3000+ to 3500+ Venice based system.

Pretty simple concept for a new review.

The entire point is to make this one review worthwhile and accurate for those of us running such systems and to NOT make it the same irrelevant pure GPU test. We want a real world combination test, not a science lab experiment to see which GPU without any other restrictions is purely the fastest. We already have that! Anandtech already does that. What we want is a review to see which mainstream (figure $200-$350) GPUs best compliment a real world system based around a 3200+ Venice chip playing the latest games on a real world budget. Not a 7800GTX, not an FX-55.

The odds of all of us having the same "mid range" test system that "AT" would use is slim and none. Say that they did do a test with a 3200+ 1GB mem and a 6800nu in FEAR.
How does that help me with my P4 3.0E?

Do you realize how extensive that testing would be? The combinations are incredibly high and would take quite an enormous amount of time. I'm talking weeks to months.

No, I say AT and other sites should keep on showing us what they have been, and we can rely on each other in here for more well rounded answers on how "My System Here" performs with a "My potential video card here" in Quake4/FEAR/CoD2 etc.etc.

P.S. No misinterpretation here. I got your point and its an enormous cumbersome task that will not cater to everyone's needs exactly.

 

[nRollo]

That equipment all suxorz- who'd want to see reviews of that old junk?! 3200? Pfft.


















JK

I'm guessing ATs policy is to review cards when they come out, not when they're a year old. Other sites do this though, you can find this info at places like Xbit or Firing Squad.

You're also right in that most people don't buy the "latest and greatest" but I imagine AT is just reviewing the stuff as they receive and has made a decision not to spend as much time benching the older stuff.

 

[tuteja1986]

force call of duty 2 to run on dx 7 and you can get 60frame on outdated graphic card but you will be giving up on shader 3.0 , hdr and dx 9 and all the other high end making the game loook good technology.

 

[Stretchman]

FX-55 moster godbox

Me likes that description.

And thumbs up for your suggestions. Real world machines for real world reviews.

 

[yacoub]

Blah blah blah

 

[Ackmed]

I would like them to consider WS resolutions. 1680x1050 and 1920x1200. WS is the way of the future, and there are a LOT of people gaming with these resolutions. 1600x1200 is probably close enough for 1680x1050, but my tests show 1680x1050 to be more demanding, at least in HL2. 1920x1200 is obviously more stressful than 1600x1200, and is whats needed to really stress these new high end cards.