Nvidia's New Flagship: The Geforce GTX 280 Review

Thumb it up! Save this page

Introduction
Almost two years after the release of the 8800 GTX, here’s the newest and latest flagship from Nvidia, the Geforce GTX 280. We finally have a succesor, after the highly successful Geforce 8000 series and the not so popular 9000 series. Here it is, in pictures:

At 1.4 billion transitors and with a size of 576mm2, it is huge! It’s the biggest GPU ever created. Produced with the 65nm process, Nvidia can fit up to 94 of their 65nm GTX 280 dies on a 300mm wafer. In comparaison, Intel can fit nearly 2500 45nm Atom processors on the same wafer. To give you an idea of just how “huge” it is, here’s a picture of a Nvidia GTX 200 die compared to the newest Penryn processor die from Intel.

Yes, this is up to scale. You could fit almost 6 Penryn dies in the space that the GTX 280 takes. So, why is so big you ask? Well, first of all, it’s produced using the aging 65nm process, compared to 55nm for ATI and 45nm for Intel. The reason for that is probably because Nvidia did not want to switch to a smaller process yet, to avoid delays, which makes sense considering that the GTX 280 is already 6 months late. What do we get from this huge die? Let’s do a comparaison with the 9800 GTX

GTX 280 vs 9800 GTX
Stream Processors: 240 vs 128; Advantage GTX 280 by 87.5%
Texture Address/Filtering: 80 / 80 vs 64/64; Advantage GTX 280 by 25%
ROPs:  32 vs 16; Advantage GTX 280 by 100%
Core clock:  602 Mhz vs 675 Mhz; Advantage 9800 GTX by 12.2%
Shader clock:  1296 Mhz vs 1690 Mhz; Advantage 9800 GTX by 30.4%
Memory Clock:  1107 Mhz vs 1100 Mhz; Advantage GTX 280 by 0.7%
Memory Bus Width:  512 Bit vs 256 Bit; Advantage GTX 280 by 100%
Memory Size:  1GB vs 512MB; Advantage GTX 280 by 100%
Transitor Count: 1600 Millions vs 754 Millions; “Advantage” GTX 280 by 112%
Manufacturing Process: TSMC 65nm vs TSMC 65nm; Equal
Retail Suggested Price: 650$ vs 300$; Advantage 9800GTX by 116%

So, what can we tell from this comparaison?
Nvidia is using all that space to pack a whole lot more of stream processors, texture adress and filtering modules and ROPs. The bus width is also finally back to 512 bit, which will lift the bandwidth limitation plaguing the 9000 series. The other thing I can tell is that Nvidia had to be reasonable with clock speeds, as 1.4 billion transitors, manufactured using the same 65nm process, produces a lot more heat compared to “only” 754 millions on its 9800 GTX counter-part. All of this together makes for a much more powerful chip, but a very expensive one for Nvidia to produce as well, which is reflected in the retail suggested price.

It is likely that Nvidia will make a shrinked version of this chip for mainstream and low-end version eventually , in the GT 200 family and possibly a new series if the competition from ATI is strong enough, in a similar way that Nvidia did with the G80(90nm, 8800 GTX/Ultra) and the G92 following later on (65nm, 8800 GT, new 8800GTS and 9000 series).

What about the performances?
Nvidia GTX 280 outpaces the 9800 GTX easily, in all games by quite a margin. However, it cannot beat Nvidia’s own 9800GX2 in the majority of games and benchmarks, which is quite a deception as you can buy the GX2 for 150$ less compared to the GTX 280. Sure, the GTX 280 is now the most powerful one chip card, but I don’t think that a lot of people care for that. Most people wants to know which card offers the best performance/price ratio and the new GTX 280 is currently a very bad option, considering it’s starting price tag of 650$ USD. Of course, it is expensive to produce and it is currently the most powerful single chip card and Nvidia is taking advantage of ATI’s delay for releasing the new Radeon HD 4850 and HD 4870, to sell their new products for more money. Nvidia also know that some people just don’t care about the price and will pay the premium asked for this new product.

My advise?
Wait for ATI to release their new generation cards, see what kind of performance they offer and for what price. What from I read, the incoming Radeon HD 4870 offers a vastly superior performance than the Geforce 9800GX2, which is great if it’s true, as the GTX 280 cannot beat the 9800GX2 in most cases. It’s also supposed to be offered for much less than 650$, Nvidia’s RSP for the GTX 280. Even though those are rumors, I wouldn’t be surprised, as the new Radeon HD 4870 is equipped with GDDR5, which is way faster than the GDDR3 on the GTX 280. In any case, even if ATI’s products don’t beat the GTX 280, it will force Nvidia to cut their prices, as ATI’s new products are likely to create a dangerous competition for Nvidia. We’ll see by the end of the month, when ATI’s releases their Radeon HD 4870.

Until that happens, you are better off with a 9800GX2 or even better, a SLI of 8800GTs. In any case, I recommend that you wait to see what ATI will bring to the market and to see Nvidia reaction afterwards. I’m sure we’ll see really good prices during the back to school promotions, in August

On a side note, I think that the name chosen (GTX 280) is really lame. Couldn’t the marketing from Nvidia come up with a more appealing name than that?!

What do you think of this new GTX 280 from Nvidia? Are you going to buy it?

Cheers, Mathieu