Review: Breathing down GTX 280’s neck
GT200 GPU core is the basis for two of Nvidia’s latest graphics offerings. The full version of GT200 is used on GTX 280, whereas the crippled version of GT200 serves as a backbone to GTX 260 cards. GTX 280 packs all the hardware treats, whereas the GTX 260 has one ROP channel and two Shader clusters less. Each shader cluster features three shading multiprocessors, each packing 8 stream processors (SP), totaling to 24 SPs per cluster. That tells us that GTX 260 is left with 192 stream processors, but it’s still more than G80 8800 GTX (128 SP) or G92 9800 GTX (128 SP) cards have to offer. Of course, all this means less performance, but the price is also set accordingly and is significantly cheaper than GTX 280.
Our today’s guest is an overclocked EVGA GTX 260 card running at 666MHz. Three sixes are here to win, and the card bears the name FTW – “For The Win”. You’ll see that EVGA GTX 260 FTW card is very fast, thanks to its overclocked core, and it’s closer to GTX 280 results than reference GTX 260 running at reference 576MHz. The shaders on EVGA’s card run at 1404MHz and the memory at 1107MHz.
Nvidia has built quite a large chip, and GT200 features 1400 million transistors in a 65nm process. Most of the new transistors are used to improve GPU’s computing capabilities, but some were used for, among other tasks, CUDA (Compute Unified Device Architecture). Compared to the last generation GT200 brings us more threads, stream processors, better shading and texturing and more memory and memory bandwidth. Video engine is left unchanged – VP2 is integrated in the core, just like we’ve seen on G92. That ensures some nice HD video processing, and we can get HDMI through any of the two DVIs with HDCP.
GTX 260’s clock speed is almost identical to 8800 GTX’s. We often compare it to 8800 GTX, in terms of 8800 GTX (G80) revolutionizing the way we look at graphics card performance. However GT200 is a chip with double the number of transistors and a juiced up setup we’ve seen on G80 – so we’re expecting some serious improvement. Still, the performance increase is not 100%, although it will be in certain cases. While 8800 GTX runs at 575MHz, GTX 260 runs only 1MHz faster. Stream processors on the old 8800 GTX run faster, but GTX 260 has 192 SP running at 1242MHz.
The next photo shows 600mm2 large GT200 dominating EVGA's GTX 260 card
Talking about memory, we once again see a strange configuration and GTX 260 packs 896MB of GDDR3 memory. Memory interface is 448bit compared to 512bit on GTX 280. This number is a direct result of GTX 260 having one ROP partition less. Each of the GTX 280’s eight ROP partitions is connected to 64-bit memory controller, so we get 512bit memory interface. Knowing that GTX 260 features 7 ROP units, it totals to 448bit memory interface. Each memory controller houses two memory chips, meaning GTX 260 needs 14 chips for its 896MB of memory.
896MB of memory is simply done by leaving out two memory chips, one on the back and one on the front of the PCB. The card is nicely wrapped and we had to carefully take the back cover off in order to show you the PCB. The cooler is dual slot and identical to those found on GTX 280 cards. It’s made of plastics, metal, aluminum and copper alloy. The back cover is made of metal and it leans on the memory, cooling it in the process. Although the core temperature is cooler on this card (78°C) than on GTX 280 (80°C), we could easily hear the fan running when the card was under a workload, but it wasn’t too loud. In idle mode it ran silent.
Both the front and the back of the card are painted in EVGA’s colors. This angle shows the power connectors; next to them is the audio in, whereas the other side hides SLI connectors.
Powering the card is done through two 6pin power connectors, whereas GTX 280 requires one 6pin and one 8pin. The card’s consumption will max out at 182W. However, this is the worst case scenario, because (with an appropriate chipset) HybridPower will power down the card and leave less-demanding tasks to be handled by the IGP. Furthermore, GT200 can turn specific parts of the chip on and off when they’re not needed and can dynamically regulate voltages and frequencies.
For those who want more than just one GTX 260 cards – this card is Tri SLI capable. For that, we have two SLI connectors that you can leave hidden until it’s their time to shine.
Nvidia’s marketing is selling GTX 200 as a multifunctional architecture locked and loaded for more than just gaming. We see them increasingly use the phrase “beyond gaming”, and it means we can use this card for other tasks besides gaming. That’s mostly thanks to Compute Unified Device Architecture (CUDA). Using CUDA, we communicate with the GPU that feeds back the processed data. One of the functions we haven’t yet tested is this card’s PhysX functionality, but we’ll get into that when we see some games with nicely implemented physics on the GPU.
EVGA GTX 260 FTW comes in the box that clearly states how this is the champ of GTX 260 cards. This time it appears they’re right, because For The Win GTX 260 is, as far as we’ve seen, the fastest GTX 260 on the market.