More Back Pages stories
From this week's TechWatch
On evaluating graphics boards
By Jon Peddie
Almost anyone one reading this has at one time in his or her life tried to evaluate a graphics board. You ever wonder what goes into doing such an evaluation? Have you ever wondered why you should care?
Who needs a graphics board evaluated?
The most logical answer would be the prospective buyer of a graphics board. After all, he or she would like to know if they are going to get value for their money, or just a logo. With that assumption, the next question is who are these people who may be in the market for a graphics board and potentially interested in the board's performance? It's a mixed lot as you would imagine, with the game enthusiasts at the top of the list because of their frequent purchases of the high-priced boards. Then, there are the less affluent and/or less enthusiastic game players who get lumped into an amorphous category known as “casual.”
The other people who need to have a graphics board evaluated are application developers, and this is mostly game developers and to a lesser extent the professional graphics ISVs. The OS suppliers need to evaluate the graphics board primarily to determine if the board's driver is going to bring down the OS. And even though Leopard and Vista are supposed to be immunized from such occurrences, it doesn't do you much good if you can't get the screen to light up does it?
The other group who needs graphics boards evaluated, of course, is the graphics board manufacturers. Not that they don't know how the boards work, but they need (or think they need) an independent evaluation of them. That, in turn, creates another group of people who need graphic boards evaluated— the evaluators. It's the basis of some of their livelihood. This is what's known as a mutually dependent symbiosis.
This past week, we attempted to evaluate some graphics boards. I say attempted because the results of the attempt did not meet our original goals. We were thwarted by benchmarks, applications, operating systems, and hardware. However, we can report that the mice, keyboards, displays, MP3 player, and coffee pot performed as expected and satisfactorily.
What we concluded from our attempts to gain insight into the new AIBs we had to evaluate was that very few people can do it, and we don't necessarily count ourselves among them.
Who can evaluate a graphics board?
There is really only one group of people who can evaluate a graphics board and they are the GPU manufacturers. Anyone else will run into difficulties and will have to call the manufacturer for help. It doesn't matter if the difficulty is due to the application, the OS, the display, or even the CPU, only the GPU manufacturer has the knowledge and staff to sort out such problems. That's not exceptional praise, it's obvious, how else would they be able to sell their product if they didn't test it and in various conditions and on various platforms?
Some of the GPU manufactures work very closely with a few select benchmarking web sites, sites that have demonstrated competence, and often a favorable attitude toward the manufacturer. Those sites are usually the first to post results, or to leak results, and depending upon how favorably they are inclined toward a particular manufacturer, leak results about the competitor's AIB. This, of course, is pure heresy to suggest such a thing, but it seems to be common knowledge amongst the web sites and testers. These are the Faustian deals the advertising-selling web sites and magazines have with the GPU suppliers and their AIB partners, and although all claim “Chinese walls” those walls can be rather porous at times.
But that's business, it's not new, and it's not shocking. And, for the most part, the benchmarking sites and magazines do a good job of putting an AIB through its paces, often uncovering deficiencies in drivers and subsequent side effects.
Who can afford to run evaluations?
Testing a graphics board involves days of work and lots of equipment. First off you have to have the gear, the test platform, the OS, the apps, and the AIBs. If the vendors of each of those things don't loan or give them to the tester, then that's a pretty big outlay of cash. Then there are various benchmarks, resolutions, filters, dual and single board operation, different OS and different processor combinations. Do you pick just one each of these or several? And how do you decide? If you have a favorite GPU supplier, you pick items that show off their part the best, but it's still a lot of work and you have to get paid for it.
Who's evaluation can you trust?
So how then how does the average, and maybe the enthusiast gamer decide or figure out whose results to trust? The enthusiasts use forums, blogs, and email. Some of the average gamers do too, and most rely on the recommendations of their friends, who often enough are enthusiast gamers. And that closes the loop on the symbiotic dependency between the GPU supplier and the benchmarkers.
Why isn't this easier?
If there really was a need for board evaluations in order for a consumer to make an informed and confident purchasing decision, we would have a better, more accessible solution. And maybe here's a product idea for someone. The ideal situation would be to be able to type in the two or three applications you are or want to use and the budget you have. Then on a screen would appear a concise and simple list of boards and their relative benchmarks for those applications. And where is that screen you ask? Well for sure it should be your internet-connected PC. Secondarily, it should be in the retail shops, and third, on your phone.
Why don't we have such a thing now? Well I think it's because no one really cares about benchmarks except the GPU sellers and the web sites that sell them ads. Unless, of course, you're just a graphics nut and like to experiment with new boards, he said looking in the mirror.
