Slashdot Mirror
NVIDIA's 55nm GeForce GTX 285 Launched
Visceralini writes "NVIDIA is launching yet another high-end 3D graphics offering, an optimized version of their top shelf GeForce GTX 280 single GPU card, dubbed the GeForce GTX 285.
This new GeForce is a
55nm die-shrunk version of the legacy GTX 280 with lower power consumption characteristics that don't require an 8-pin PCI Express connector, rather just a pair of more standard 6-pin plugs.
Performance
metrics are shown here in a number of the latest game titles including
Fallout 3, Left 4 Dead, Far Cry 2 and Mirror's Edge. The new GTX 285 is
about on par or slightly faster than a GTX 280 but with
less power draw and some room for overclocking over the reference design."
The new GTX 285 is about on par or slightly faster than a GTX 280 but with less power draw and some room for overclocking over the reference design.
40W less while idle (vs. 280), @ $0.12 kWh, means if I can pick one up for $400 (I can dream, can't I?), it will have paid for itself - through power savings - in less than 10 years!! I know what I'm spending my tax refund on!!
Now that I think about it, I'm pretty sure everything I just said is completely wrong.
- The current best performing single card is the GeForce GTX 295
- The best performance setup was (before this card) a tossup between dual GeForce GTX 295s (quad SLI) and three GeForce GTX 280s (three-way SLI).
- The overclocking potential of the GeForce GTX 285 & reduced power consumption might make a three-way 285 setup preferrable to a dual 295 setup (for enthusiasts)
...who lack unlimited funds, the best buy at the moment are the ATi HD 48x0 series cards, which have ridiculously good price/performance and will run any current or near-future game easily at high detail.
Read Pynchon.
Barely.
You just need to write an X86 emulator as a fragment shader. Part of me actually wants to see somebody attempt this, and get a PC OS running on a video card. (Larrabee doesn''t count, using x86 to start with spoils the fun.)
Unless you live way up north or play games only in the winter, dealing with 840 Watts of heat is going to be problematic for a dual GTX295 setup. Summer is worse in that you now have to pump out that heat through the AC system.
People often will bitch about their cable/DSL bill, but have they ever tried to calculate the monthly cost of electricity their gaming rig racks up alone?
Some of us don't care. :D
"Two things are infinite: the universe and human stupidity; and I'm not sure about the universe." Albert Einstein
the parent post misses the fact that for dual and more 280 setups, you need a huge case and a kickass power supply. ati 4870s are much easier to fit in a single standard case, and give similar or better performance.
Read radical news here
I did say quite a bit that it was more of a driver parity issue than anything else.
The 9600 is faster than 8600. I meant that the price point of $85 was for the 'higher quality' 8600s. $85 was also a happy price point for decent 8600GTs over the holiday. Pre-overclocked, better fans, other things that would supposedly warrant a few dollars extra, compared to the $45-60 'standard' 8600GTs.
The 3870 has better specs, obviously. Specs don't mean much if the driver doesn't take advantage. On Nvidia drivers, some driver versions have been better for some hardware than others, newer hardware can take a few versions for both Nvidia and ATI to exploit the hardware better.
Rivatuner can present the GPU load statistics (as represented by the internal performance counters on the ATI GPU) along with framerate and temeprature via its statistics monitoring, with an overlay similar to FRAPS so you can see it while you play. OpenGL tended to squarely hit into 30% or less range unless blatantly simple. Various games, including GTA4, rarely moved above 50-55%, just so happened to get same or slightly better framerate on the 8600 (usually 2-4 times faster on the 9600). Things that only used basic shading or none at all (including emulators) tended to represent the highest percentage of load on the 3870 (and highest actual performance).
The games where GPU load was low, also happened to be where atrocious unplayable performance was overwhelmingly likely to occur.
The 8600GT didn't beat the 3870 in everything. Most obviously, in games that took better advantage of 512MB, the 256MB on the 8600GT limited it. There was a rather distinct trend across a wide variety of games made since 2004 or so (I play around with a lot of things) that the 8600GT would usually get the same framerate, in some cases better. Quality-wise, the 3870 would usually have minimum framerates that sagged lower and usually there'd be a significant lag in games, particularly with vsync+triple buffer enabled (a necessity with LCD).
Half Life 2 (heck, the 8600GT got ~180 frames on Ep2 with max settings, including AA 4x, 3870 got ~40), GTA 4, Far Cry 2, Crysis, Crysis Warhead, Brothers in Arms: Hell's Highway, Fallout 3, San Andreas, Mercenaries 2, No One Lives Forever, Oblivion, UT2004, World of Warcraft, Tabula Rasa, among others.
People say 'specs specs specs', but considering a single driver revision can change performance for a particular game for 15% or more, drivers matter a great deal. You could have a really fast car with high horsepower and a novice driver...who crashes it, or has to drive slower to keep it on the road, compared to a Ford Focus, which they drive more easily and confidently, and rack up a higher average speed.
The press tends to use static demos and benchmarks. A benchmark in a 'real game' doesn't much matter if it uses a prescripted static environment which doesn't change and is specifically designed to provide higher framerates than is at all representative of the actual game proper. Very few sites who publish reviews do so in actual Real World game conditions, where they're actually playing in a meaningful circuit and record detailed framerate statistics per second to file. The 3870 scores pretty well on nearly every benchmark, including OpenGL ones. It's only a 'few hundred' points shy of the 9600GT replacement on 3DMark, despite there being no similarity in any game I've yet found. Lies, damned lies, and benchmarks.
The performance counters only specify what percentage of GPU resource is being used, not how exactly they're defined, and don't relate how different drivers affect it. Hardware counts, how well the drivers can actually spread the load around would generally make the load percentage (and performance) go up. Presumably, the counters count based on how many shaders/render/texture pipelines are being utilized, or some subset thereof, though certainly shaders are being counted.
If you're getting 15 laggy frames per second on a game that should be pushing as hard as the graphic
"A Goddess rarely smiles for she is forced by others to be an island unto herself." - Zephiris
sry but those benchmarks are not even close to right. since when does the 280 score 2000 points more than the 4870 1 gig. sry not a chance. when you guys post benchmarks, at least post correct ones. and 1920x1200 at 4x aa and 16x af is not a valid benchmark because those are not 3d marks default (and online comparable for that matter) settings. this is not a nvidia fan boy cry, i run nvidia myself lol but i know what the 4870's can really do.