Suggestions for Individual Video Card Reviews

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Mingon

Diamond Member
Apr 2, 2000
3,012
0
0
not all for this review but perhaps others

PCI cards + low profile
Tv out quality
Tv capture
Passive cards + heat
Full generation cards vs next gen e.g ti4X00 vs 5x00
Recommended settings which give a min of 60fps for a card
Noise
Onboard video shootout (nforce vs intel vs sis vs via vs ati)
 
Jan 31, 2002
40,819
2
0
Just don't be bribe-taking fanboys willing to bend over for the latest product sample like THG and I'm sure you'll do just fine. :p

- M4H
 
Jan 31, 2002
40,819
2
0
Originally posted by: Mingon
Onboard video shootout (nforce vs intel vs sis vs via vs ati)

You're joking, right?

Let me summarize that for you in a sentence.

nForce2 0wnz j00.

There, that's the entire article. If ATI ever gets their R9000-based chipset out the door, then we might need to add another sentence. :p

- M4H
 

spam

Member
Jul 3, 2003
141
0
0
I have some suggestions, perhaps you could add yours.
-Instead of using timedemos create custom demos for benchmarking purposes.
-Use more games for custom demos that are not prone to driver "optimizations"- that is, less well known ones.
-compare custom demo's to the standard time demo's and benchmarking utilities and look for irregularities.

think we need more of a moving target to stymy the cheats and exploits written into video drivers. That is why using different games presents something of an obstacle to would be "optimizations". In a review you could add a category called the "The Fudge Factor" You would still use your standard benchmarking games but additionally you would add other games that would serve as a control group to verify your results. How many hundreds of games are released every year? Surely some of them could be used to give a broader perspective on a video cards performance.
 

BoomAM

Diamond Member
Sep 25, 2001
4,546
0
0
Originally posted by: spam
I have some suggestions, perhaps you could add yours.
-Instead of using timedemos create custom demos for benchmarking purposes.
-Use more games for custom demos that are not prone to driver "optimizations"- that is, less well known ones.
-compare custom demo's to the standard time demo's and benchmarking utilities and look for irregularities.

think we need more of a moving target to stymy the cheats and exploits written into video drivers. That is why using different games presents something of an obstacle to would be "optimizations". In a review you could add a category called the "The Fudge Factor" You would still use your standard benchmarking games but additionally you would add other games that would serve as a control group to verify your results. How many hundreds of games are released every year? Surely some of them could be used to give a broader perspective on a video cards performance.
Excellent ideas.
But to prevent timedemos being optimised, they would have to come up with new ones every few months or so.
More games used to benchmark would be good. For example, i have seen some sites using more common/popular games, like Vice City for benchmarking (using fraps), which is good, cos it shows how the cards perform in more common games. Games such as Vice City, RalliSport Challenge, Splinter Cell, C&C Generals would be good games to benchmark them with.
 

jjyiz28

Platinum Member
Jan 11, 2003
2,901
0
0
nice suggestions so far. i agree, use custom timedemos or make your own.
and about image quality, comparing actual quality, not just what the driver says.

i doubt they will use custom timedemos though, giving better light to nvidia.
 

PLaYaHaTeD

Senior member
Oct 15, 2001
242
0
0
another vote for 1280x1024

another vote for reviewing the 9800 non pro and 5900 non ultra. most of us can't afford $500 for a video card. Personally, I wouldn't spend $500 on a video card even if i did have the money.


Also, hint us with some dirt on the FX Value and R420 :) We want to know SOOO Bad
 

Rogodin2

Banned
Jul 2, 2003
3,219
0
0
And some variation in games.

NOT just Q3, seriousam2, UT2003.

How about one flight sim (Forgotten Battles), MMPOG (Battlefield 1942 road to rome), Morrowind, and the latest VC release at least.

Also to a real IQ comarison. Run both cards and determine what settings are equivalent then bench using the IQ settings that are as close as possible.

All of Bensky's suggestions as well.

Also run highest levels of anisotropy and tell us if you see ANY variation in IQ between the cards.

I really hope that you will do an in depth review worth of the money spent.


rogo
 

BFG10K

Lifer
Aug 14, 2000
22,709
2,997
126
  • Custom timedemos that are available for download. They should have decent length and action to mimick real-world situations as closely as possible.
  • All demos should be run at equivalent image quality settings.
  • Screenshot comparisons, especially with all cards running at maximum detail levels.
  • Screenshots of in-game realtime framerate counters in places that are covered by the timedemos.
  • How about 2048 x 1536? Lets face it, most of these cards are now finding 1600 x 1200 too easy in a lot of games.
  • Some older games would be nice too. For example, GLQuake at 2048 x 1536 with 16x AF and 6x AA.
  • NO FLASH GRAPHS!
 

lopri

Elite Member
Jul 27, 2002
13,298
678
126
I want to see 1600x1200 on all the benches with no AA/AF. Being an RTS player I prefer higher resolution much more than textures/model quality.

 

ElFenix

Elite Member
Super Moderator
Mar 20, 2000
102,365
8,475
126
Originally posted by: BoomAM
But to prevent timedemos being optimised, they would have to come up with new ones every few months or so.
not really, you'd just have to keep some of the demos private. have some available for download so everyone can test and have some held private to ensure that any driver optimizations at least helps the whole game and not just a benchmark.
 

ElFenix

Elite Member
Super Moderator
Mar 20, 2000
102,365
8,475
126
and do get rid of the flash graphs. i don't have flash installed because advertising with it is just far too intrusive.
 

jjyiz28

Platinum Member
Jan 11, 2003
2,901
0
0
Originally posted by: ElFenix
Originally posted by: BoomAM
But to prevent timedemos being optimised, they would have to come up with new ones every few months or so.
not really, you'd just have to keep some of the demos private. have some available for download so everyone can test and have some held private to ensure that any driver optimizations at least helps the whole game and not just a benchmark.

elfenix, didn't even think about that. that would make the reviewers job more easier.
 

BoomAM

Diamond Member
Sep 25, 2001
4,546
0
0
Good idea.
The standard custom demos could be kept by reviews, and only used by reviews, and the non stardard custom demos, could be released by the reviews, for users to test out their systems with.
 

Webthug

Member
Jun 29, 2003
98
0
0
Would it be possible to benchmark cards on an AGP 4x Mobo. This is because most people who run Intel systems, still only have AGP 4 avaiable through the 845 or 850 chipsets. I feel that it is currently implied that the performance at AGP 8x will be the same at AGP 4x. Whilst this might be the case, this can misleading. Also for the upcoming benchmarks a suggestion would be to benchmark it with a 4600Ti and perhaps a 9700 pro and a softmodded 9500 Pro (how many people don't mod a 9500 Pro)
 

BoomAM

Diamond Member
Sep 25, 2001
4,546
0
0
Originally posted by: Webthug
....and a softmodded 9500 Pro (how many people don't mod a 9500 Pro)
Prehaps cos a 9500pro cant be modded to a 9700. Its only initial batch of 9500s that could.
 

Blastman

Golden Member
Oct 21, 1999
1,758
0
76
To compare IQ you might do what gamespy did, a side by side comparison ? ?

gamespy ?the output from the GeForce, especially with AA and AF on, has a certain fuzzy quality to it. I probably wouldn't have noticed it as much if I didn't have the two cards side by side, but the difference is fairly obvious when you're able to compare and contrast in real-time. Both cards produced vibrant colors and detailed texturing, but things like rendered text appeared slightly diffused on the GeForce cards. By this, I mean things like a player's name over their head in an FPS or menu options just seem less clear.

HardOCP also picked this issue up ?

hardocp ? (5900U) ... The second shot above shows this even more clearly. Look closely at the shot from Serious Sam 2 where I have it circled in white. The 4X AA shot looks slightly blurred compared to the 2X AA shot.


 

oldfart

Lifer
Dec 2, 1999
10,207
0
0
I think the overwhelming theme here is to spend more time with IQ than raw FPS. Lets face it, any of the top cards will have more than a high enough framerate to play current games smoothly. 220 Vs 230 FPS is not important. I think we should have matured past making a purchasing decision on a cards raw FPS score by now.

Still, many (maybe most) sites still have not much more than a bunch or bar graphs showing which card is fastest, and that is how the "winner" is picked. 3D chipset makers know this, and that how this whole driver cheating mess got started.

Again, Image quality, driver quality, game compatibility, 2D clarity, and a quiet cooling design should get AT LEAST as much attention as an FPS score.
 

Pete

Diamond Member
Oct 10, 1999
4,953
0
0
Yes, the point is a more well-thought-out review, not straight benchmark number production and simplistic conclusion straight from those numbers. (I'm not accusing AT of doing that, though I think a little more work on the IQ side of things is warranted.)
 

Jeff7181

Lifer
Aug 21, 2002
18,368
11
81
Well first, just wanna say, I like the old graphs you use to use vs. the new ones. Switching back to those would be a plus in my opinion.

As far as the reviews... I'd like to see AA and AF tests compared on the same page, on the same chart, to non-AA/AF tests.

I'd also like to see screen shots of each test performed, so we can get an idea if 4X AA and 16X AF is really much better than 2X AA and 4X AF.

Another thing that would be nice that I haven't seen from you in a long time is overclocking results.

And finally... maybe throw in some oddball tests that you don't normally do... most of the benchmarks done are done by everybody and their brother, so the card makers can optimize specifically for those. Maybe throw in some tests of Soldier of Fortune 2 or Ghost Recon... or go old school and benchmark Counter-Strike or just plain ol' Half Life. Or hey... go to PCPitstop.com and run their tests... their video tests are very primative, but they're still tests, and since nVidia and ATI aren't likely to optimize for those, it might show some differences you wouldn't normally see.
 

DAPUNISHER

Super Moderator CPU Forum Mod and Elite Member
Super Moderator
Aug 22, 2001
29,800
25,353
146
I'd like to see you review the eVGA G4 440MX 8X AGP Personal Cinema. While it obviously isn't going to match up to most AIW versions in gaming because it's just a 440MX I've found the hardware/connection setup and software user interface to be more user friendly in comparison to ATi's way of doing it even ATI's MMC 8.5. The remote is almost identical and at 127$ shipped from Newegg for the retail version including Intervideo WinDVR, Ulead DVD moviefactor and videostudio, NVDVD, Ghost Recon, and America's Army, it is a superior buy compared to the 7500 AIW OEM for 123$ even with gaming consider IMO.

Anywho, you never hear about this card but it's a very good value that it would be helpful to see a major site like Anandtech review so many people who have no knowledge of it will become aware that ATI isn't the only game in town for inexpensive HTPC all-in-one graphics card solutions.
 

Errors

Junior Member
Jul 16, 2003
18
0
0
Originally posted by: PLaYaHaTeD
another vote for 1280x1024

another vote for reviewing the 9800 non pro and 5900 non ultra. most of us can't afford $500 for a video card. Personally, I wouldn't spend $500 on a video card even if i did have the money.


Also, hint us with some dirt on the FX Value and R420 :) We want to know SOOO Bad



And yet another vote for a review/comparison of the 9800 non pro here... this looks like a very interesting card price/performance wise ,it would be nice to see the facts on it