What would you like to see in a video card review?

[ultimatebob]

I've been reading some video card reviews lately, and I've been noticing that they're kinda boring. They're filled with page after page of benchmarks, but aren't all that interesting to read.

I've been thinking of some things that I would like to see in a video card review. Let me know what you think:

1) Sound samples of what the video card sounds like when it's running. I've been burned on this before... trying to build a "quiet" PC only to have a sound like a jet engine when it's running a game.

2) Power usage tests under load... I'd like to know if I REALLY need one of those 1000W "l33t gam3r" power supplies to run an SLI rig like they recommend.

3) How many times the system crashes due to driver issues... they always seem to overlook driver stability in video card reviews.

4) How well the video card holds up to a close range shotgun blast... for those particularly BAD video card reviews

 

[MustISO]

1 and 2 are already done in the reviews I read. System stability would be an interesting factor but how much of it is the video cards fault versus the system it's running on. I just don't think it would be a reliable gauge of quality.

 

[Schmide]

Stock goop vs replacing goop temps and the obligatory naked card.

 

[TC91]

I'd like to see benches of older games too and not just only the newest games out there but it's unlikely due to the time that would be required...

 

[ShawnD1]

Originally posted by: TC91
I'd like to see benches of older games too and not just only the newest games out there but it's unlikely due to the time that would be required...

They use new games to test the card's ability to do newer graphical features like OpenGL 3.0 or DirectX 10 or whatever the latest shader model is. Newer games also use more shaders and larger textures, so they are better for spotting where a bottleneck starts to occur. A classic example of this is in the Geforce 8 era where there was a 320mb version of the 8800GTS. The card did great at low resolution, then at a certain resolution it would drop like a rock because it ran out of memory. Testing the card with an older game will not show this bottleneck, but testing it with a newer game that uses large textures will. I bet a lot of people who looked at the low res benchmarks are pissed that they can't run a modern game on that thing while a 512mb or 768mb card still works.

Running a newer game at several resolutions also gives you a pretty good idea of how the different cards will scale in the future (depending on how long your personal upgrade cycles are). COD 5. As you can see in that picture, the Nvidia cards start to win as the resolution gets higher and higher. That's information I want to know. Another one is Fallout 3; this set of graphs show the Radeon 4850 and 4870 512mb hit a wall when the resolution goes from 1920x1200 to the next resolution higher. Benchmarking with an older game will not show this.


What I want to see is Anand writing more articles. They always have a certain entertainment quality to them.
http://www.anandtech.com/video/showdoc.aspx?i=3539&p=8

Our goal was simple: we wanted to know if GPU accelerated PhysX effects in these titles was useful. And if it was, would it be enough to make us pick a NVIDIA GPU over an ATI one if the ATI GPU was faster.

To accomplish this I had to bring in an outsider. Someone who hadn?t been subjected to the same NVIDIA marketing that Derek and I had. I wanted someone impartial.

Meet Ben. I met Ben in middle school and we?ve been friends ever since. He?s a gamer of the truest form. He generally just wants to come over to my office and game while I work. The relationship is rarely harmful; I have access to lots of hardware (both PC and console) and games, and he likes to play them. He plays while I work and isn't very distracting (except when he's hungry).

 

[Modelworks]

I would like to see exactly what a video card uses in amps, measured off the power cable connector on the video card, not the wall outlet.

 

[MTDEW]

My #1 want in video card reviews is rather simple.
Along with every benchmark/chart, how about a downloadable link to the EXACT timedemo or whatever was used to get the numbers.

Too many reviews nowadays are using custom timedemos instead of built in game benchmarks due to drivers being optimized for the in-game benchmarks, which is just fine.

But it would be nice to have access to the custom benches that we can run on our own systems to see how or current card stacks up or better yet if our new card is running properly.

Im not sure driver stability tests would really help many users, since bench systems are usually a clean install (image) which eliminates 99% of problems.
Which is why i always take a few hrs and do a clean install/ acronis image when purchasing a new card, cuz ya know thats how Nvidia/AMD test their new cards and drivers anyway.

 

[reallyscrued]

Originally posted by: Schmide
Stock goop vs replacing goop temps and the obligatory naked card.

Yes. Skin is good.

 

[ochadd]

SETI, Folding, and GPUGrid ppd. Other forms of pure computational performance and performance per watt.

 

[Stumps]

Lots of 3dmark scores...that's all that matters these days

 

[Wreckage]

Newer games instead of old games and no crappy console ports like GRID.
Folding@home performance
Video transcoding performance
Noise, power and heat is always good

 

[kreacher]

Test with stock clock speeds across the board, specially for new launches (overclock cards should have separate reviews that compare them with stock cards).

 

[theprodigalrebel]

I like the performance-over-time graphs that HardOCP and PC Perspective do. A line graph that shows performance over the length of the test is far more informative and revealing than a pair of min/avg numbers. It isn't enough knowing that a game throws 45fps avg/18fps min, if that dip to 18fps barely lasted a second and the game stayed above 30fps 95% of the time.

I haven't played many PC games recently but I liked how the original FEAR's built-in test gave a summary of performance. I think it was something like:

x% below 25 fps
y% between 25 and 40 fps
z% above 40fps

Generating reports like those (from the FRAPS date) for a range of GPUs shouldn't be too hard to do in Excel. The reviewer would just need to determine what fps-ranges can be considered unplayable/decent/smooth.

 

[Boobs McGee]

I would like to see comparisons with cards from older generations. There are a few reasons this would be a plus. It gives folks looking to upgrade a good feel for what real gains they will be seeing. I would think that ati and nvidia would like that as well to show how much their new architecture improves and would drive sales. The reviewer could just use data from previous reviews or run new benchmarks as necessary. Also if new benches are ran on the older cards that would be a good way to see how well drivers have improved gameplay performance throughout a products lifecycle. I know it is already time consuming, but it is something I would like to see. Maybe Anand will add some gpu stuff to his "Bench" that could apply here.

The other thing I would like to see is more roundup style reviews. Anybody remember all the guides and roundup reviews that used to be constantly popping up on AT? I miss that style of review.

 

[TurtleCrusher]

How cool it makes you on your block.

 

[gevorg]

No "zoomed in" scales in benchmark charts. Like when one card gets 150 points, and the other gets 140 points, the difference in their chart bar lengths is is halfway! This is so misleading and annoying. The scale should be always linear and complete.

 

[Scali]

Originally posted by: Scholzpdx
How cool it makes you on your block.

Haha, I can just imagine a Cool Wall for videocards, like in Top Gear.
Ofcourse with Kristin Scott-Thomas as measure for how cool a card is
"This card is French, therefore it cannot possibly be cool".

 

[nemesismk2]

I want to see a couple of older video cards (such as the Radeon 3870 or GeForce 8800 GT) used in some of the latest reviews so we can then decide if it's worth upgrading.

 

[ochadd]

Originally posted by: nemesismk2
I want to see a couple of older video cards (such as the Radeon 3870 or GeForce 8800 GT) used in some of the latest reviews so we can then decide if it's worth upgrading.

Agree 100%. There are so many people out there with 8800GT, 3870, 3850, and those in SLI/Crossfire that it would be great to see them included.

 

[RobbieD]

Originally posted by: TC91
I'd like to see benches of older games too and not just only the newest games out there but it's unlikely due to the time that would be required...

I understand that a focus on newer games makes sense because that's where the bottlenecks tend to show up.

Unfortunately, we tend to see the chip companies focusing on these benchmarks and neglecting the others. For example, ATI/AMD owners have been having a LOT of trouble with World of Warcraft. This is an interesting situation... 12m players pay $15/mo to play this game, but because it's not a typically used benchmark, ATI doesn't pay much attention.

There has been a lot of complaining about these cards, and I myself moved to a 4870 Turbo 1G and have had regular locking issues, crashes, stuttering and lag. I just don't think that big green or big red are going to focus on problems like this as long as the reviewers don't hold their toes to the fire.

Here's a link to the forum thread.

I really wish I'd have the benefit of some WoW reviews before buying this card. It had terrific benchmark results, and won "Editor's Choice"/etc awards, so seemed a safe bet.

Clearly the proof is in the pudding, not in a review. I understand that it must be hard to run a WoW benchmark, given you can never have the same test run over and over. I honestly don't know the answer to this, but I certainly see it as a problem.

 

[LittleNemoNES]

It would be nice if they added more interesting AA settings to their Reviews. I don't care if Battlefield 2 runs @ 300 FPS w/ 4x AA.

I want to see (as I already play myself) 16xs AA + 16xAF PNGs and report if it is playable. (BTW, this setting is perfectly playable BF2)

Also links to HD video of the card running.

 

[tcsenter]

Originally posted by: Modelworks
I would like to see exactly what a video card uses in amps, measured off the power cable connector on the video card, not the wall outlet.

+1, but I would include power delivered through the interface, like Xbit Labs does (or used to) using a modified motherboard with testing shunts to measure the voltage drop. These 'total system power under load v. idle' are fine for, well, assessing total system power.

 

[ShawnD1]

world of warcraft is not tested because it has relatively low graphical requirements and is crippled by poor multithreading. When I upgraded from a GeForce 6600GT to a 7950GT, I saw literally no frame rate improvement at all. The game was still using 100% CPU. Then I got a Core 2 E6600 and the performance still sucked because it only used 1 core and that core was 100% all the time. Only way to improve performance is to lower the draw distance. Draw distance on an E6600 is the difference between 90fps and 20fps in Ironforge when it had hundreds of people in it.

Also, that thread clearly shows that it's not an ATI or Nvidia specific problem. The first guy has a Radeon 4870, second guy has a GeForce 8800GTS. Lower down, a guy says he has an i7 and GTX 285 and gets the same performance. An i7 doesn't help with that game because it's poorly threaded.


I don't want to jump to conclusions but I'm going to guess that the people bitching have no idea what they're doing. Remember Crysis? I can run that game just fine on a 3 year old computer with a GeForce 7 video card. People were posting on this forum and many others that they couldn't get it to work on a GeForce 8. People are idiots.

 

[Jumpem]

Multiple CPU speeds.

 

[nitromullet]

Originally posted by: MTDEW
My #1 want in video card reviews is rather simple.
Along with every benchmark/chart, how about a downloadable link to the EXACT timedemo or whatever was used to get the numbers.

I agree with this. In addition to this I'd like to have more info available for download. Stuff like:

- Actual FRAPS spread sheets or graphs that show the entire benchmark run as well as the interval between frames (especially for multi-gpu tests).

- Detailed settings used. Not just "resolution, everything on high, 4xAA/16xAF". I want to know what kind of AA, whether or not you forced settings via the driver or used the application settings, and if you used the default image quality settings in the driver.

I get that not everyone wants this level of detail, but make it available for download for the people that do.

Edit: Maybe for lesser graphically intensive and older games, stability testing might be a better idea instead of performance data. Maybe just a random selection. Like they dusted off a few old games, and ran them on a GTX 295 and 4870 X2 to see how smoothly the experience went. I very much think this type of testing should be done for very popular games (like the Sims and WoW), regardless of the fact that they aren't necessarily graphically intensive apps.