Slashdot Mirror
Hands On With Nvidia's New GTX 280 Card
notdagreatbrain writes "Maximum PC magazine has early benchmarks on Nvidia's newest GPU architecture — the GTX 200 series. Benchmarks on the smokin' fast processor reveal a graphics card that can finally tame Crysis at 1900x1200.
'The GTX 280 delivered real-world benchmark numbers nearly 50 percent faster than a single GeForce 9800 GTX running on Windows XP, and it was 23 percent faster than that card running on Vista. In fact, it looks as though a single GTX 280 will be comparable to — and in some cases beat — two 9800 GTX cards running in SLI, a fact that explains why Nvidia expects the 9800 GX2 to fade from the scene rather quickly.'"
In most reviews, the 9800GX2 is faster, and it's also $200 cheaper. As a multi-GPU card it has some problems with scaling, and micro-stutter makes it very jumpy like all existing SLI setups.
I'm not well versed in the cause of micro-stutter, but the results are that frames aren't spaced evenly from each other. In a 30 fps situation, a single card will give you a frame at 0 ms, 33 ms, 67 ms, 100 ms, etc. Add a new SLI card and let's say you have 100% scaling, which is overly optimistic. Frames now render at 0 ms, 8 ms, 33 ms, 41 ms, 67 ms, 75 ms, 100ms, and 108ms. You get twice the frames per second, but they're not evenly spaced. In this case, which uses realistic numbers, you're getting 60 fps might say that the output looks about the same as 40 fps, since the delay between every other frame is 25 ms.
It would probably look a bit better than 40 fps, since between each 25 ms delay you get an 8 ms delay, but beyond the reduced effective fps there are other complications as well. For instance, the jitter is very distracting to some people. Also, most LCD monitors, even those rated at 2-5 ms response times, will have issues showing the 33 ms frame completely free of ghosting from the 8 ms frame before the 41 ms frame shows up.
Most people only look at fps, though, which makes the 9800 GX2 a very attractive choice. Because I'm aware of micro-stutter, I won't buy a multi-GPU card or SLI setup unless it's more than 50% faster than a single-GPU card, and that's still ignoring price. That said, I'm sort of surprised to find myself now looking mostly to AMD's 4870 release next week instead of going to Newegg for a GTX280, since the 280 results, while not bad, weren't quite what I was hoping for in a $650 card.
"I zero-index my hamsters" - Willtor (147206)
Because graphics operations are embarrassingly parallel whereas regular programs arn't.
Full Tilt
You may be wondering, with a chip this large, about power consumptionâ"as in: Will the lights flicker when I fire up Call of Duty 4? The chip's max thermal design power, or TDP, is 236W, which is considerable. However, Nvidia claims idle power draw for the GT200 of only 25W, down from 64W in the G80. They even say GT200's idle power draw is similar to AMD's righteously frugal RV670 GPU. We shall see about that, but how did they accomplish such a thing? GeForce GPUs have many clock domains, as evidenced by the fact that the GPU core and shader clock speeds diverge. Tamasi said Nvidia implemented dynamic power and frequency scaling throughout the chip, with multiple units able to scale independently. He characterized G80 as an "on or off" affair, whereas GT200's power use scales more linearly with demand. Even in a 3D game or application, he hinted, the GT200 might use much less power than its TDP maximum. Much like a CPU, GT200 has multiple power states with algorithmic determination of the proper state, and those P-states include a new, presumably relatively low-power state for video decoding and playback. Also, GT200-based cards will be compatible with Nvidia's HybridPower scheme, so they can be deactivated entirely in favor of a chipset-based GPU when they're not needed.
This is how graphics cards used to work. You would plug a VGA cable from your standard 2D graphics card to your, for example, Voodoo II card, and the Voodoo II card would go out to the monitor. You could just have the 3D card working in passthrough mode when not doing 3D stuff. Something like this could work on a single board though. There's no reason you couldn't power down entire sections of the graphics card that you aren't using. Most video cards support changing the clock speed on the card. I'm wondering if this is a problem at all, with any real effects, or whether it's just speculation based on the poster assuming what might happen. Anybody have any real numbers for wattage drained based on idle/full workload for these large cards?
Anthropic principle: We see the universe the way it is because if it were different we would not be here to see it.
The 9800GX2 may be cheaper but it most certainly is not faster, even considering your links. From Anandtech, the charts show a significant speed increase with the new hardware.
In fact, from the article:
The GTX 280 delivered real-world benchmark numbers nearly 50 percent faster than a single GeForce 9800 GTX running on Windows XP, and it was 23-percent faster than that card running on Vista. In fact, it looks as though a single GTX 280 will be comparable to--and in some cases beat--two 9800 GTX cards running in SLI, a fact that explains why Nvidia expects the 9800 GX2 to fade from the scene rather quickly.
Which leads me to the question, are you trolling?
When run under Vista, it features tons of additional effects. Those are the reasons why the speed improvement in Crysis aren't that much impressive under Vista.
PS: And for the record, Radeon HD3870X2 uses the exact same GDDR3, not GDDR4 as TFA's review says. ATI choose to go for GDDR3 to cut the costs of the dual GPU setup. (Only a few non standard boards by 3rd party manufacturer use GDDR4 and a PCI-express 2.0 bridge).
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Pet peeve: Profane people propagating perfunctory pedantry.
Yes, they're comparing it to two 9800GTXs, which is what a 9800GX2 is: Two 9800GTXs on one board. RTFS indeed. It seems like a case of give and take. The GTX280 is more expensive, but is a single GPU solution, which tends to be more stable. The 9800GX2 is cheaper, runs about as fast, but is a dual GPU unit, so you might have a few more "issues" to deal with in your game playing adventures.
Precisely. This is something that can be solved by simply throwing more transistors in. Their biggest challenge is probably power and heat, not architecture.
Not to mention that "programs" on GPUs are ridiculously simple compared to something on a general purpose CPU. Next time you write a shader, try branching (i.e. if, else), your shader will slow to a relative crawl.