Another GF110 Part

The new NVIDIA GeForce GTX 570 1.25GB graphics card brings very similar specifications to the GTX 480 with slightly improved clocks and a lower price that should please gamers looking for a holiday upgrade. But how does it compare to the best from AMD’s single GPU offerings or even the high-end GeForce GTX 580?
For the latter half of 2009, the review schedule at PC Perspective was dominated by graphics cards from AMD as we saw the entire HD 5000 series come to fruition.  Since then, NVIDIA has slowly but surely caught up with numerous releases like the GTX 480, 470, 460 and so on, and more recently, the GeForce GTX 580 based on the newer GF110 GPU. 

Today we are looking at another new card from NVIDIA in the GTX 500-series, the GeForce GTX 570.  As you might expect from the name, the GTX 570 is a slight step below the GTX 580 in terms of performance but turns out to be a noticeable step lower in price: $349 versus $529. 

The GeForce GTX 570 is based on the same GF110 GPU that the GTX 580 is, a slightly revised version of the original Fermi-architecture. 

The GF110 is still a 3.0 billion transistor GPU (despite rumors to the contrary) and has not dropped features like ECC to reach these performance levels.  The secret to NVIDIA’s improved performance and yield comes from a monitoring and rearranging of the transistor paths that caused the most leakage on GF100 in real-world environments.  NVIDIA’s engineers then rearranged many segments of the die to better improve efficiency and power consumption finally allowing them to enable the full power of the architecture as GF110.

There have been some other architectural enhancements made on the GF110 GPU as well including a full speed FP16 texture filtering addition and improved Z-cull efficiency.  These clock-for-clock improvements in the GPU design offer performance improvements of anywhere from 5% in Metro 2033 to 13% in Dirt 2 based on NVIDIA’s own testing.

Remember this diagram that we have shown you so many times before?

NVIDIA GeForce GTX 570 1.25GB Graphics Card Review - Graphics Cards 77

This is the full, 512 CUDA core version of the Fermi architecture seen with the GTX 580.  The new GTX 570 that we are reviewing today will ship with 480 CUDA cores, the original count of the GTX 480 released in March. This means that a single SM is disabled on the core and also results in 60 texture units on board compared to the 64 on the GTX 580. 

The ROP count on the GTX 570 is 40 resulting in a 320-bit memory bus and slightly lower total memory throughput than the GTX 580. 

NVIDIA GeForce GTX 570 1.25GB Graphics Card Review - Graphics Cards 78

The spec sheet above gives us the rest of the speeds and feeds for this new graphics card including the 732 MHz core clock and 1464 MHz shader clock which is just below the 772/1544 MHz speeds on the GTX 580.  The memory frequency is set at 950 MHz, just a bit lower than the 1002 MHz speed on the GTX 580 though this card gets 1.25GB of memory rather than 1.5GB.   

It might be just as interesting to compare the new GTX 570 to the original GeForce GTX 480 card that shares similar specs. 
  • GeForce GTX 570 versus GTX 480
  • 480 vs 480 CUDA cores
  • 60 vs 60 texture units
  • 40 vs 48 ROPs
  • 732 MHz vs 700 MHz core frequency
  • 1.25GB vs 1.5GB memory size
  • 950 MHz vs 924 MHz memory frequency
So while the shader performance should be higher on the GTX 570 the memory performance and AA performance might be slightly faster on the original GTX 480.  I should note though that the GTX 480 will be selling out very soon, what is in the market is all that will be released, so the GTX 570 is effectively replacing it with a lower price.

« PreviousNext »