Slashdot Mirror
Complete Nvidia GTX280 Scores Posted
Groo Wanderer writes "Since boredom is a dangerous thing on the weekends, I decided to alleviate mine by running 233 benchmarks on the new GT280. This includes 28 gaming related tests across up to nine resolutions, and 9800GTX numbers thrown in for good measure. Since there were no NDAs involved in getting you these numbers, we are not bound by the pesky NDA that lifts tomorrow. You can read all of the numbers here; enjoy."
Second, the TV set has a crappy resolution for its size; 1920×1080? If I wanted that resolution I'd get a 24" monitor (or two) which is at least going to have a sensible DPI for, well, a monitor. here's the kicker, the difference in response time?
And what makes you think games were my primary motivation for it? It's a computer, I do work on it, and that resolution gives me space to run a browser side by side with editors, file browsers, half a dozen terminals, TV guides and media players. With the addition of one of my old 20"'s, it's like having 3 monitors.
Have you ever tried sitting on a desk in front of a 46" Monitor? I wouldnt be able to work (or game) on anything larger than 28". The sides of the monitor are almost outside of my peripheral vision. A monitor is not really workable if you have to move your head side to side to see the whole thing (this of course wouldn't matter if you werent sitting at a desk, but I would assume most PC users do).
Also at 2560x1600 resolution, CPU would not even factor in modern PC games. At that resolution, graphics cards would max out first due to memory and bandwidth limitations (benchmarks for games like Crysis and SupCom show this happening).
Especially when it comes to ATI, I wouldn't count the chicks before they actually hatch. They have a knack, really a talent, a _gift_, for screwing up half the time.
I also wouldn't take any theoretical numbers as gospel. For either of the two. They both love to talk about theoretical gigatexels per second and GB/s of memory bandwidth as if the pipelines will always be fully used, every cycle, and memory was a continuous read that never had any RAS and CAS cycles in there.
But especially for ATI. They have that talent again for screwing up, but with good maths to back up the idea that it should have worked.
Starting from the original Radeon, which _should_ have had exactly the number of texturing units to fully utilize the memory bandwidth, and no more. (I.e., without wasting silicon on units which would just stall waiting for data anyway.) It should have run circles around NVidia's card of the same generation, right? In practice, it fell a bit short.
Or look at the more recent HD2xxx and HD3xxx cards. In theory the HD2600's unified architecture should have done miracles, but in practice it never quite worked that well. In theory, the HD38xx should have fixed that, except it barely made it competitive with an 8800 GT. In some games. At some resolutions. On a good day.
Don't get me wrong, I'm not against either of the two. If anything, I used to dislike Nvidia a bit, which made me favour the ATI's. But, if I'm to learn anything from history, I'm a bit wary of making such bold pronouncements as "This will crush the GT280 in just about every conceivable benchmark". There are already plenty of cases where such prophecies were made, and just ended up with the prophecised Nvidia-killer quickly repackaged as "see, we don't want to compete at the top end. We'll just make it a mid-end card, ok? It's not that we can't compete at the top, mind you. We just suddenly find it obscene to charge $500 for a graphics card."
Or to put it otherwise, too often I've heard one or the other (but, again, ATI a bit more) crying "Wolf" and they barely managed to produce a dachshund. I'll wait until I actually see the big bad wolf this time, before joining in the chorus marveling at what a big, strong and fast wolf it is.
A polar bear is a cartesian bear after a coordinate transform.