Alright, let's get this puppy started up, and get your freak on! Today NVIDIA will release two high-end class DirectX 11 capable products. First off, the new graphics adapters are of course DirectX 11 ready. With Windows 7 and Vista also being DX11 ready all we need are some games to take advantage of DirectCompute, multi-threading, Hardware Tessellation and new shader 5.0 extensions. DX11 is going to be good and once tessellation kicks into games, much better looking.
Most of the rumors you heard over the past seven days or so have been either correct or at least, close, these are the bold numbers:
Some of you will say "huh?" as everybody expected 512 shader processors on the top of the line model, 272 cores more than the GT200 (GeForce GTX 280/285). We were as much surprised as you guys when we learned that it turned out to be 480 shader processors for the GeForce GTX 480. See, the Fermi architecture, the GF100 graphics processor has sixteen shader clusters embedded in it (called SMs). For the GeForce GTX 480 one such a cluster is disabled, and on the GeForce GTX 470 (though expected) two are disabled.
As to why one SM is disabled on the GeForce GTX 480, we can only speculate. But we have reason to believe that NVIDIA did this to secure better yields and ramp volume production up. Also, much like the GeForce GTX 260 SP216, for the future to be able to introduce a slightly faster SKU with the full 512 SPs. The reality remains that it will not matter 'heaps' in terms of performance as you lose only one-sixteenth of shader performance.
The GeForce GTX 400 cards have the Fermi family chip embedded on them. That chip is called GF100. The GF100 GPU is placed onto an 8-layer PCB and tied to gDDR5 memory. On the topic of graphics memory; NVIDIA made their memory controller GDDR5 compatible, which was not the case on GT200 based GeForce GTX 260/275/285/295, hence their GDDR3 memory.
Memory wise expect large, exemplary large numbers, as we are passing 1 GB. Each utilized memory controller on the respective GPU will get 256MB of memory tied to it.
The GTX 470 has five memory controllers (5x256MB) = 1280 MB of GDDR5 memory
The GTX 480 has six memory controllers (6x256MB) = 1536 MB of GDDR5 memory
As you can understand, the massive memory partitions, bus-width and combination of GDDR5 memory (quad data rate) allows the GPU to work with a very high framebuffer bandwidth. So we had a chat about the GF100 GPU and memory. It's now time to break down the GPU clocks:
GeForce 9800 GTX
GeForce GTX 285
GeForce GTX 295
GeForce GTX 470
GeForce GTX 480
Stream (Shader) Processors
Core Clock (MHz)
Shader Clock (MHz)
Memory Clock (effective MHz)
Two Dual link DVI
So the GeForce GTX 480 has a big fat memory partition alright, 1,54 GB of GDDR5 memory which is tied to a 384-bit memory bus that binds to six memory controllers inside the GPU; six memory controllers x 64-bit = 384-bit. This memory is clocked at 924 MHz (= 3696 MHz effective).
For the GeForce GTX 470 that would be five memory controllers x 64-bit = 320-bit. The GTX 470 memory will be clocked at 837 MHz (= 3348 MHz effective).
Core frequencies then. The GeForce GTX 480 will have its core clocked at 700 MHz and its shader processors at 1400 MHz. The GeForce GTX 470 on its end will be clocked lower at 607 MHz on the core and 1215 MHz on the shader processors.
Let's take it to the next step now and look at more in-depth specifications.
GeForce GTX 470
GeForce GTX 480
Streaming Multiprocessors (SM)
Graphics Clock (Core)
Shader Processor Clock
Memory Clock / Data rate
837 MHz / 3348 MHz
924 MHz / 3696 MHz
Texture Filtering rate (Bilinear)
1x6-pin PEG, 1x8-pin PEG
Max board power (TDP)
Recommended Power supply
GPU Thermal Threshold
105 degrees C
105 degrees C
So we talked about the core clocks, specifications and memory partitions. Obviously there's a lot more to talk through. Again, for an in-depth explanation of the GF100 architecture please visit this article we wrote earlier this year.
Now, at the end of the pipeline we run into an improved ROP (Raster Operation) engine, and the GTX 480 has 48 units, the 470 has 40 for features like pixel blending and AA. There's a total of 60 texture filtering units available for the GTX 480 and with one SM less for the GTX 470; 56 texture units.
The math is simple here, each SM has four texture units tied to it.
GeForce GTX 470 has 14 SMs X 4 Texture units = 56
GeForce GTX 480 has 15 SMs X 4 Texture units = 60
A fictional GeForce GTX 485 would have 16 SMs x4 TUs = 64
What's also interesting about the GTX 400 release, but was a challenge, is the move towards a smaller die and fabrication process (40nm) which often brings several advantages. One advantage is that you can insert more transistors into the silicon; as such the GF100 GPU comes with 3 billion transistors embedded into this GPU.
Despite a smaller die however, TDPs are relatively high, the GTX 480 will consume at maximum 250 Watts whereas the GTX 470 is using as much as the last generation product, roughly 225W. Still, that's definitely not as bad as the first rumors suggested.
TDP = Thermal Design Power - roughly translated, when you stress everything on the graphics card a 100% your maximum power consumption is the TDP.
The GeForce GTX 480 comes with both a 6-pin and 8-pin power connector to get enough current and a little spare for overclocking. This boils down as: 8-pin PEG = 150W + 6-pin PEG = 75W + PCIe slot = 75W is 300W available.
The GeForce GTX 470 comes with two 6-pin PEG connectors each delivering 75W and another 75W over the PCIe slot = 225W. The GeForce GTX 480 targets a price point around $449-499, while the GeForce GTX 470 is expected to be priced $299-$349.
I want to leave it at that for the physical tech side of the GPU and now move forwards to the photo shoot and other things like DirectX 11 and, quite specifically, Tessellation. First up the photo's, then our standard snippet on GPU and shader technology to help you understand how a graphics card actually works.
Nvidia GeForce GTX 1070 review In this review we test the GeForce GTX 1070 (Nvidia Founders Edition). The 8 GB graphics card is the somewhat limited little brother of the GTX 1080, this little demon on the Pascal architecture and 1...
Nvidia GeForce GTX 1080 review We review the all new Nvidia GeForce GTX 1080 (founders edition). The new 8GB beast based on the Pascal architecture and 16nm FinFET has arrived. It's cool, it's silent and it rocks hard when it com...