Another 9-series card. What's new this time, huh?
Trotting out our familiar table.Graphics cards | NVIDIA GeForce 9800 GX2 1024 | NVIDIA GeForce 9800 GTX 512 | NVIDIA GeForce 8800 GT 512 | NVIDIA GeForce 8800 GTS 512 | NVIDIA GeForce 8800 Ultra 768 | ATI Radeon HD 3870 X2 1024 | ATI Radeon HD 3870 512 |
---|---|---|---|---|---|---|---|
PCIe | PCIe 2.0 | PCIe 1.x | PCIe 2.0 | ||||
GPU clock | 600MHz | 675MHz | 600MHz | 650MHz | 612MHz | 825MHz | 775MHz |
Shader clock | 1,500MHz | 1,688MHz | 1,500MHz | 1,625MHz | 1,500MHz | 825MHz | 775MHz |
Memory clock (effective) | 2,000MHz | 2,200MHz | 1,800MHz | 1,940MHz | 2,160MHz | 1,802MHz | 2,250MHz |
Memory interface, size, and implementation | 2x 256-bit, 1,024MiB, GDDR3 | 256-bit, 512MiB, GDDR3 | 384-bit, 768MiB, GDDR3 | 2x 256-bit, 1,024MiB, GDDR3 | 256-bit, 512MiB, GDDR4 | ||
Memory bandwidth | 128GB/sec (card) | 70.40GB/sec | 57.60GB/sec | 62.1GB/sec | 103.68GB/sec | 115.328GB/sec (card) | 72.8GB/sec |
Manufacturing process | TSMC, 65nm | TSMC, 90nm | TSMC, 55nm | ||||
Transistor count | 1,508M | 754M | 681M | 1,300M | 666M | ||
Die size | 2x 296mm² | 296mm² | 484mm² | 2x 192mm² | 192mm² | ||
DirectX Shader Model | 4.0 | 4.1 | |||||
Vertex, fragment, geometry shading (shared) | 256 FP32 scalar ALUs, MADD dual-issue (unified) | 128 FP32 scalar ALUs, MADD dual-issue (unified) | 112 FP32 scalar ALUs, MADD dual-issue (unified) | 128 FP32 scalar ALUs, MADD dual-issue (unified) | 640 FP32 scalar ALUs, MADD dual-issue (unified) | 320 FP32 scalar ALUs, MADD dual-issue (unified) | |
Peak GFLOP/s | 768 | 432 | 336 | 416 | 384 | 1,056 | 496 |
Data sampling and filtering | 128ppc address and 128ppc bilinear INT8 filtering, max 16xAF | 64ppc address and 64ppc bilinear INT8 filtering, max 16xAF | 56ppc address and 56ppc bilinear INT8 filtering, max 16xAF | 64ppc address and 64ppc bilinear INT8 filtering, max 16xAF | 32ppc address and 64ppc bilinear INT8 filtering, max 16xAF | 16ppc address and 32ppc bilinear INT8 filtering, max 16xAF | |
Peak GTexel/s (bilinear) | 76.8 | 43.2 | 33.6 | 41.6 | 19.584 | 26.4 | 12.4 |
ROPs | 32 | 16 | 16 | 16 | 24 | 32 | 16 |
Peak TDP (claimed) | 196 | 156 | 105 | 140 | 175 | 196 | 110 |
Multi-GPU | SLI - two-board | SLI - three-board | SLI - two-board | SLI - three-board | CrossFire - two-board | CrossFire - four-board | |
Outputs | 2 x dual-link DVI w/HDCP, HDMI | 2 x dual-link DVI w/HDCP, HDMI (native, on GPU) | 2 x dual-link DVI w/HDCP, mini-DIN | 2 x dual-link DVI w/HDCP (discrete ASIC), mini-DIN | 2 x dual-link DVI (HDMI) w/HDCP, mini-DIN (VIVO) | ||
Hardware-assisted video-decoding engine | NVIDIA's PureVideo HD - full H.264 decode and partial VC-1 decode | AMD UVD - full H.264 and VC-1 decode | |||||
Reference cooler | dual-slot | dual-slot | single-slot | dual-slot | dual-slot | ||
Retail price (default-clocked model) | £389 | £220? | £139 | £179 | £399 | £249 | £129 |
9-series, really? Sure it's not a GeForce 8800 GTS 512 in disguise?
What's not so new, compared to GeForce 8800 GTS 512We know that the GeForce 9800 GTX is the formal successor to the long-in-the-tooth GeForce 8800 GTX - the GPU, launched in November 2006, that revolutionised the way that pixels are painted on to your screen.
However, take a close, close look at the vital statistics for the GeForce 9800 GTX and 8800 GTS 512. Both are built on a 65nm manufacturing process; harness 754M transistors; are endowed with 128 stream processors; interface with external memory - 512MiB, typically - via a 256-bit memory bus; and can sample and address the same number of pixels per clock cycle, etc.
[advert]The point here is that the 9800 GTX is, essentially, a faster-clocked version of the 8800 GTS 512 that was launched at the end of November. The core frequency sees a meagre ~4 per cent increase, shaders run at 12.5 per cent faster, and GDDR3 memory is 22 per cent faster.
NVIDIA states that the GPU has integrated HDMI support - just like the GeForce 8800 GT and GTS 512 - offering audio pass-through via the DVI port and through a certified DVI-to-HDMI adaptor, but the user still needs to connect a S/PDIF cable to the card, which isn't as elegant as ATI's approach on its Radeon HD 3000-series cards.
Our previous testing has showed that the 8800 GTS 512 was, on balance, just a little slower than the now-discontinued Ultra. GeForce 9800 GTX should become NVIDIA's fastest single-GPU board and should benchmark around 10 per cent higher than a stock-clocked 'GTS 512.
Initial pricing estimates indicate that stock-clocked cards will sell for $299-$349. Knowing that default-clocked 8800 GTS 512s currently e-tail for around $250 in the US and £180 in the UK, it's safe to assume UK-based pricing of around £230 today, launch day, dropping to £215 in a week or so's time. Incidentally, that's the same price as the current GeForce 8800 GTX 768MiB.
What's kind of new but seen before on other NVIDIA GPUs?
Now, one of the erstwhile 8800 GTX - and, indeed Ultra's - features not available on the G9x-class of GPUs was the ability to run three cards in tandem - three-way SLI - through the provision of two SLI connectors on the cards. The recently-released GeForce 9800 GX2, we noted, shoehorned in two GPUs per board and could run a four-GPU combination by adding another card, but the GeForce 9800 GTX fills the vacuum left by two-connector 8800 GTX well enough.
Also note that, like the GeForce 9800 GX2, the 9800 GTX supports NVIDIA's HybridPower, where, on a compatible chipset that features an IGP, the discrete, wattage-hungry card can be turned off when not in heavy 3D use, saving power and routing video through the IGP instead.
Released with the ForceWare 174.xx drivers and currently software-upgraded on the 9-series cards for now, NVIDIA introduced dynamic contrast and colour enhancements for better-looking images, it says. Further, the driver supports dual-stream decode, which hardware-accelerates the decoding of two video streams running concurrently. There are only few instances where we can see this being useful - picture-in-picture commentary, for example. NVIDIA will debut these incremental increases in PureVideo HD's functionality on 8-series cards at a later date.
Other bits, and summary
NVIDIA recommends a 450W PSU with at least 24A on the 12V rail if running a single card. That recommendation spirals to 750W for two-way SLI and 1,000W for three-way SLI. The card also works in conjunction with NVIDIA ESA (Enthusiast System Architecture) for thermal monitoring and user-defined fan-speed modulation.The GeForce 9800 GTX 512 is significantly cheaper to produce than the 768MiB-equipped GeForce 8800 GTX that it replaces. The 9800's frame-buffer, then, is smaller, and memory bandwidth suffers when compared to the 384-bit setup on the G80 GTX, but that's somewhat masked by its superior texturing ability. It will lose some bandwidth-intensive gaming benchmarks against the former GTX but win others that rely on furious texturing ability.
With all that in mind, the nomenclature hides the fact that, really, it's just a faster-clocked version of the current GeForce 8800 GTS 512, albeit it with three-way SLI and integrated HDMI support.
Apart from faster speeds and the possibility of three-way SLI, the GeForce 9800 GTX brings nothing really new to bear that we haven't seen before. By rights, then, it should be called the GeForce 8900 GTX, because we're not sure how NVIDIA's pulled the 9-series moniker out of the PR bag. All it will do, really, is potentially confuse buyers into thinking it's a whole generation ahead of the GeForce 8800 GTS 512, which it clearly isn't. We reckon the fact that it supports HybridPower and three-way SLI isn't enough for a generation's gap in nomenclature.
Retail GeForce 9800 GTX pricing is such that it's relatively close to ATI's dual-GPU Radeon HD 3870 X2, which can now be picked up for around £249. Let's now see it if it's faster....