Table-comparison time
And here is the comparison table:
| Graphics cards | AMD Radeon HD 3870 512 | AMD Radeon HD 3850 256 | AMD Radeon HD 2900 Pro 512 | AMD Radeon HD 2900 XT 512 | NVIDIA GeForce 8800 GT 512 | NVIDIA GeForce 8800 GTS 320 | NVIDIA GeForce 8800 GTS 640 |
|---|---|---|---|---|---|---|---|
| PCIe | PCIe 2.0 | PCIe | PCIe 2.0 | PCIe | |||
| GPU clock | 775MHz | 670MHz | 600MHz | 743MHz | 600MHz | 500MHz | 500MHz |
| Shader clock | 775MHz | 670MHz | 600MHz | 743MHz | 1500MHz | 1200MHz | 1200MHz |
| Memory clock (effective) | 2250MHz | 1656MHz | 1584MHz | 1656MHz | 1800MHz | 1600MHz | 1600MHz |
| Memory interface, size, and implementation | 256-bit, 512MiB, GDDR4 | 256-bit, 512MiB, GDDR3 | 512-bit, 512MiB, GDDR3 | 256-bit, 512MiB, GDDR3 | 320-bit, 320MiB/640MiB, GDDR3 | ||
| Memory bandwidth | 72.8GB/sec | 53GB/sec | 101.376GB/sec | 105.984GB/sec | 57.60GB/sec | 64.00GB/sec | |
| Manufacturing process | TSMC, 55nm | TSMC, 80nm | TSMC, 65nm | TSMC, 90nm | |||
| Transistor count | 666M | 700M+ | 754M | 681M | |||
| Die size | 192mm² | 408mm² | 296mm² | 484mm² | |||
| DirectX Shader Model | 4.1 | 4.0 | |||||
| Vertex, fragment, geometry shading (shared) | 320 FP32 scalar ALUs, MADD dual-issue (unified) | 112 FP32 scalar ALUs, MADD+MUL dual-issue (unified) | 96 FP32 scalar ALUs, MADD dual-issue (unified) | ||||
| Peak GFLOP/s | 496 | 428 | 345.6 | 475.52 | 336 | 230.4 | |
| Data sampling and filtering | 16ppc address and 32ppc bilinear INT8 filtering, max 16xAF | 56ppc address and 56ppc bilinear INT8 filtering, max 16xAF | 24ppc address and 48ppc bilinear INT8 filtering, max 16xAF | ||||
| Peak GTexel/s (bilinear) | 12.4 | 10.72 | 9.6 | 11.888 | 16.8 | 12 | |
| ROPs | 16, 8Z or 8C samples/clk, 2clk FP16 blend 8xMSAA, 24xCFAA |
16, 8Z or 8C samples/clk, 2clk FP16 blend 8xMSAA, 16xCSAA |
20, 8Z or 8C samples/clk, 2clk FP16 blend 8xMSAA, 16xCSAA |
||||
| Outputs | 2 x dual-link DVI (HDMI) w/HDCP, mini-DIN (VIVO) | 2 x dual-link DVI w/HDCP, mini-DIN | 2 x dual-link DVI w/HDCP (discrete ASIC), mini-DIN | ||||
| Hardware-assisted video-decoding engine | AMD UVD - full H.264 and VC-1 decode | run via GPU's shaders | NVIDIA's VP2 - full H.264 decode and partial VC-1 decode | run via GPU's shaders | |||
| Reference cooler | dual-slot | single-slot | dual-slot | single-slot | dual-slot | ||
| Expected retail price | £140 | £110 | £165 | £220 | £165 | £170 | £220 |
We've included the stats of the Radeon 3850 for illustrative purposes only. Its expected retail price of £110 puts it in the same bracket as NVIDIA's GeForce 8600 GTS, really.
However, we've not include the fundamentally more expensive GeForce 8800 GTX (£300) or Ultra (£375).
We also note that the Radeon 2900 Pro was a limited-edition SKU that's not readily available and the Radeon HD 2900 XT is, for all intents and purposes, going to go end-of-line soon.
There's a fine balance between maintaining decent performance and cutting costs in the mid-range space. Both AMD and NVIDIA have chosen to equip their latest-and-greatest GPUs with optimised 256-bit memory buses, causing peak bandwidth to suffer as a result. The Radeon 3870's smaller manufacturing process gives it a considerable advantage over GeForce 8800 GT 512 in shading-power-to-die-size but NVIDIA hits back with a texturing fill-rate that's significantly higher.
We could discuss architectures all day long but it only constitutes one part of 3D performance. Remember that most games are overwhelmingly sponsored by NVIDIA's 'The Way It's Meant To Be Played' developer-relation initiative - and there's absolutely nothing wrong with that.
Thinking that through to the end, TWIMTBP-sponsored games code will take advantage of NVIDIA's texturing-to-shading balance more than AMD's, leading to higher frame rates, but that's something AMD will need to address.
While AMD is happily banging on the drum of reduced power-consumption figures for its Radeon HD 3870 - as compared to Radeon HD 2900 XT - the product does actually ship with a dual-slot cooler.
AMD argues that this is to allow the enthusiast to push the clock frequencies significantly higher, ignoring the fact that NVIDIA's GeForce 8800 GT - a no-less potent GPU - carries a single-slot cooler.
On paper, though, the Radeon HD 3870 512 and GeForce 8800 GT 512 seem to be reasonably evenly matched in specifications.
