If you’re one of those sadists who love to see 31 flavors of graphics cards slicing the market into $10 increments, then today is your lucky day. Not only do you have ATI’s new Radeon HD 4890 1 GB sliding into the $249 price point, but Nvidia is also launching its GeForce GTX 275 at the same price, set in between the GTX 285 and GTX 260 Core 216—decidedly closer to the GTX 285, as we found in our testing.
We won’t pretend that the simultaneous timing of these two unveilings is in any way coincidental. It’s certainly easy to understand the two companies’ line of thinking here, though.
On one hand, you have ATI coming off successful launches of its Radeon HD 4850, 4870, and 4870 X2 cards. The red team is out to show everyone that it still has the moves, and that its re-timed RV790 architecture is worth as much now as RV770 was 10 months ago.
On the other hand, you have Nvidia, which took a beating early on in the ATI RV770 GPU's life cycle—until it lowered prices on its own boards to compete a little more evenly. We have to imagine the green team is out to show that it can do battle based on performance and an attractive price tag right out of the gate this time around.
Filling In The Gaps
Nvidia has already tried the “let’s disable one thread processing cluster” angle with its GT200 architecture—that resulted in the GeForce GTX 260 Core 216. The GeForce GTX 260 Core 216 is about on par with the Radeon HD 4870 1 GB already. So, Nvidia needed something newer, something faster.
The only thing short of a GeForce GTX 280, which is being phased out in favor of the GTX 285, is a GeForce GTX 280 with the GTX 260’s back-end—the 28 ROPs and 896 MB of GDDR3 on a 448-bit memory bus. Incidentally, that’s the same GPU doubled up and slapped on Nvidia’s GeForce GTX 295. Now there’s a concept that works.
Nvidia is playing its price card very close to its proverbial chest. Set a target too early and ATI finds out, using price and performance data to re-orient the Radeon HD 4890. Loose lips sink ships and all of that. But by withholding the GeForce GTX 275’s price tag until the very last minute, comparing the card to its competition becomes a tricky matter.
With that said, early murmurs from Nvidia fall right around $249—right at ATI’s suggested retail price on the Radeon HD 4890 1 GB. Moreover, cards are expected to start trickling out shortly after launch in Europe and be widely available to the rest of the world by April 14th, so you very likely won’t be able to buy a card right away (in contrast, the Radeon HD 4890 should be available at launch). Finally, the drivers with which we’re testing are in beta, and will be posted to Nvidia’s download site as betas on April 2nd. Combine those three factors and we’re a little more comfortable calling this a preview. The hardware is final, but some of the other particulars could be subject to change between now and when the GTX 275 shows up for sale--especially once ATI and Nvidia find out what each other are charging and start jockeying for position.
- Introduction
- The GeForce GTX 275’s Inner Workings
- Still Waiting On A Killer PhysX App?
- New Features In GeForce 185
- Test Setup And Benchmarks
- Benchmark Results: 3DMark Vantage
- Benchmark Results: Far Cry 2
- Benchmark Results: Crysis
- Benchmark Results: Left 4 Dead
- Benchmark Results: Stalker: Clear Sky
- Benchmark Results: Grant Theft Auto IV
- Benchmark Results: World in Conflict
- Benchmark Results: Sum Of All Games
- Power Consumption
- Conclusion

kudos to nvidia for stepping up their game in the last second that had nothing to do with renaming cards. an attractive card at an attractive price.
it'd be smart for ati to not release the 4890x2. wouldn't make any sense, like my grammar skills. hopefully drivers can catch up and do some damage, id doubt they woudl do anything dramatic though.
No need to fuel the tin-foil brigade elsewhere on the Web re: hand-picked cards. When we can get our hands on these boards for our System Builder Marathons, straight from e-tail, then we'll give you the goods on overclocking with the same boards available to everyone else!
If the retail price for this card is turn to be over $300, they will never fool this monkey cause he will just get two HD 4770 and crossfire'd them, which surely beat a single 275
If the retail price for this card is turn to be over $300, they will never fool this monkey cause he will just get two HD 4770 and crossfire'd them, which surely beat a single 275
Not sure if you got a chance to read that page or not, but it says:
With the frame rates added together, we get a “big picture” number to compare the various cards, though a comparison like this one is going to shift its weight most heavily to high-frame rate games like Left 4 Dead and de-emphasize titles like Crysis.
I agree that the usefulness of such a metric is limited, yet we've been asked for them, so they're provided =)
Sorry I didn't read it. :-p But that doesn't change the fact, that any conclusion driven from it is highly distorted...
Who ever asked for this method of comparison probably didn't know what he wants. :-D
Would have been much more sound to take an average.
8800GT performance similar to the 8800GTX's, for lesser price.
GTX275 performance similar to the GTX285's, for lesser price.
Less than two years ago 8800 Ultra was priced $800-$900, just to make this monkey wondered, do they use gold in producing that ultra chip?
So then the GTX275 will be recycled 3 times because if it works just rename it!
But seriously G92 8800GT great release from nvidia that blew away ati in the price for performance
An example already exists:
EVGA has their GTX 280 1024MB in their "B-Stock" store still selling for $269.99. That's $20 more than their pre-order price set for the GTX 275 896MB. You really think lower clock and memory speeds with an additional 128MB memory is worth the difference? Even when their "B-Stock" items come with ZERO accessories?
Gimme a break nVIDIA. Your cards are solid, but it's about damn time you learned to name them properly.
Ford and GM for the win.
An average would still give more weight to the higher FPS games. The lower FPS games are more important as there is effectively no difference between 100FPS and 120FPS.
If you want an overall figure, you need to first normalize each game's benchmark so that they weigh evenly. To get a normalized score, you'd divide a cards FPS by the FPS of the highest performing card in the group. Whether you normalize across resolutions, game settings, or just games is up to you as long as you specify. Using this method you will often see a larger spread for low FPS games (though they should be more important anyways). However, this represents the true difference between the card (percentage wise) and should not be mitigated.
If you really wanted an overall apples to apples, you'd apply the above method, but only use settings for the games that each card being compared can play smoothly. It doesn't matter if you get 7FPS or 12FPS, you can't play it either way. You might want to divide the comparisons into classes (highend, mid-highend, etc) to get an effective overall comparison.