The guts of GTX 465
So how does it stack up?| Graphics cards | NVIDIA GeForce GTX 480 1,536MB | NVIDIA GeForce GTX 470 1,280MB | NVIDIA GeForce GTX 465 1,024MB | NVIDIA GeForce GTX 295 1,792MB | NVIDIA GeForce GTX 285 1,024MB | ATI Radeon HD 5970 2,048MB | ATI Radeon HD 5870 1,024MB | ATI Radeon HD 5850 1,024MB |
|---|---|---|---|---|---|---|---|---|
| General clock | 700MHz | 607MHz | 607MHz | 576MHz | 648MHz | 725MHz | 850MHz | 725MHz |
| Shader clock | 1,401MHz | 1,215MHz | 1,215MHz | 1,242MHz | 1,476MHz | 725MHz | 850MHz | 725MHz |
| Memory clock (effective) | 3,696MHz | 3,348MHz | 3,206MHz | 1,998MHz | 2,484MHz | 4,000MHz | 4,800MHz | 4,000MHz |
| Memory interface and size | 384-bit, 1,536MB GDDR5 | 320-bit, 1,280MB GDDR5 | 256-bit, 1,024MB, GDDR5 | 896-bit (2 x 448-bit), 1,792MB, GDDR3 | 512-bit, 1,024MB, GDDR3 | 512-bit (2 x 256-bit), 2,048MB | 256-bit, 1,024MB, GDDR5 | 256-bit, 1,024MB, GDDR5 |
| Memory bandwidth | 177.4GB/s | 133.9GB/s | 102.6GB/s | 2 x 111.9GB/s | 159GB/s | 2 x 128GB/s | 153.6GB/s | 128GB/s |
| Manufacturing process | TSMC, 40nm | TSMC, 40nm | TSMC, 40nm | TSMC, 55nm | TSMC, 55nm | TSMC, 40nm | TSMC, 40nm | TSMC, 40nm |
| DirectX/ Shader Model | DX11, 5.0 | DX11, 5.0 | DX11, 5.0 | DX10, 4.0 | DX10, 4.0 | DX11, 5.0 | DX11, 5.0 | DX11, 5.0 |
| Vertex, fragment, geometry shading (shared) | 480 FP32 scalar ALUs, MADD + MUL | 448 FP32 scalar ALUs, MADD + MUL | 352 FP32 scalar ALUs, MADD + MUL | 480 FP32 scalar ALUs, MADD + MUL | 240 FP32 scalar ALUs, MADD + MUL | 3,200 FP32 scalar ALUs, MADD + MUL | 1,600 FP32 scalar ALUs, MADD + MUL | 1,440 FP32 scalar ALUs, MADD + MUL |
| GFLOPS throughput | 1,345 | 1,088 | 855 | 1,192 |
708 |
4,176 | 2,720 |
2,088 |
| Texturing | 60ppc
bilinear 30ppc FP16 15ppc FP32 |
56ppc
bilinear 28ppc FP16 14ppc FP32 |
44ppc
bilinear 22ppc FP16 11ppc FP32 |
160ppc
bilinear 80ppc FP16 40ppc FP32 |
80ppc
bilinear 40ppc FP16 20ppc FP32 |
160ppc
bilinear 80ppc FP16 40ppc FP32 |
80ppc
bilinear 40ppc FP16 20ppc FP32 |
72ppc
bilinear 36ppc FP16 18ppc FP32 |
| ROPs | 48 |
40 |
32 | 56 |
32 |
64 | 32 |
32 |
| GPixels/s throughput | 33.6 |
24.28 |
19.424 | 32.26 |
20.74 |
46.4 |
27.2 | 23.2 |
| GTexel/s bilinear | 42 |
33.99 |
26.7 | 92.2 |
51.84 |
116 |
68 |
52.2 |
| Board power (max) | 250W | 215W | 195W? | 289W | 183W | 294W | 188W | 170W |
| Multi-GPU | Four-way SLI | Three-way SLI | Three-way SLI | Two-way SLI | Three-way SLI | Two-way XFire | Four-way XFire | Four-way XFire |
| Board length | 10.5in | 9.5in | 9.5in | 10.5in | 10.5in | 12in | 11in | 9.5in |
| Connectors (native) | 2x
dual-link DVI Mini-HDMI |
2x
dual-link DVI Mini-HDMI |
2x
dual-link DVI Mini-HDMI |
2 x dual-link DVI, HDTV-out,HDMI | 2x
dual-link DVI HDTV-out |
2x
dual-link DVI Mini-DisplayPort |
2x
dual-link DVI HDMI, DisplayPort, |
2x
dual-link DVI HDMI, DisplayPort, |
| Etail price | £399 | £299 |
£230+ | £399 |
£285 |
£550 | £325 | £235 |
The easiest and most apt comparison GPU to the GTX 465 is the GTX 470, shown directly to the left.
Cut to the core(s)
NVIDIA has to cheapen the cost of the GTX 465 GPU by, we imagine, using silicon that doesn't quite make the GTX 470 grade - both are hewn from the same 3bn-plus-transistor die. This is why the newer GPU uses three fewer Streaming Multiprocessors - 11, down from 14, with each being composed of 32 CUDA-capable cores - that bring shader/CUDA-count down from 448 to 352.
Four SMs form what is known as a Graphics Processing Cluster (GPC), and NVIDIA drops the GTX 465's complement from four to three - enough for 12 working SMs. GeForce GTX 465, however, keeps the clocks at a GTX470-matching 607MHz/1,215MHz. The end result is computational performance which is over 20 per cent lower than GTX 470's.
Memory-bandwidth snip
Another obvious snip comes in the form of lower memory bandwidth. NVIDIA accomplishes this in two ways. Firstly, the effective memory speed is reduced by four per cent or so, down from 3,348MHz to 3,206MHz. This reduction is coupled with the use of a narrower memory-bus, down from 320-bits to 256-bits - or, pragmatically, a removal of a 64-bit memory partition. Combine the two together and the GTX 465's potential memory-bandwidth is reduced from 133.9GB/s to 102.6GB/s.
ROP reduction
A narrower memory-bus means fewer raster backends (ROPs), dropping from 40 on the GTX 470 to 32 on the new GPU. ROP performance predicates the gigapixel/s throughput and, more importantly, the general effectiveness of running anisotropic filtering and antialiasing. Ceteris paribus, the lower the ROPs/memory bandwidth, the lower the IQ performance, especially at high resolutions.
Probable performance
The numbers suggest that GeForce GTX 465 will provide around 80 per cent of the performance exhibited by the GTX 470. Our previous numbers show that GTX 470 offers some 10 per cent extra performance over ATI's HD 5850 GPU when evaluated over five games, so the GTX 465 vs. Radeon HD 5850 fight will be interesting, particularly as both are priced at similar levels.
Summary
NVIDIA's harvested the big GTX 400-series die for all it's worth with the GTX 465. Snipping performance-enhancing features across the board, NVIDIA is enabling partners to produce retail cards that will compete against the Radeon HD 5850 on an equal price footing.
