GeForce GTX 560 Ti: Old Suffixes Mean New Cards
Because Nvidia enables the complete GF114 GPU, the GeForce GTX 560 Ti’s specs match the graphics processor’s full implementation. The result is a GPU with ample shader power, thanks to 384 CUDA cores. Its wider SMs facilitate 64 texture units across those eight modular elements. That’s the same number as a GF110, though it takes 16 SMs to enable similar functionality on the flagship GeForce GTX 580. Nvidia is able to run the GPU’s fixed function circuitry at 822 MHz, while the CUDA cores get clocked at 1644 MHz (the 1:2 ratio we’re used to seeing).
As with GeForce GTX 460 1 GB, GeForce GTX 560 arrives with all four of its ROP partitions enabled, outputting up to 32 pixels per clock. Four 64-bit memory pathways aggregate to 256 bits. Armed with 1 GB of 1002 MHz GDDR5 memory, that’s a maximum of 128.3 GB/s of bandwidth.
| Header Cell - Column 0 | GeForce GTX 560 Ti | GeForce GTX 570 | GeForce GTX 460 1 GB | GeForce GTX 470 |
|---|---|---|---|---|
| Graphics Processing Clusters (GPCs) | 2 | 4 | 2 | 4 |
| Streaming Multiprocessors (SMs) | 8 | 15 | 7 | 14 |
| CUDA Cores | 384 | 480 | 336 | 448 |
| Texture Units | 64 | 60 | 56 | 56 |
| ROP Units | 32 | 40 | 32 | 40 |
| Graphics Clock | 822 MHz | 732 MHz | 675 MHz | 607 MHz |
| Shader Clock | 1644 MHz | 1464 MHz | 1350 MHz | 1215 MHz |
| Memory Clock (Data Rate) | 1002 MHz (4008 MT/s) | 950 MHz (3800 MT/s) | 900 MHz (3600 MT/s) | 837 MHz (3348 MT/s) |
| Memory Capacity | 1 GB GDDR5 | 1.25 GB GDDR5 | 1 GB GDDR5 | 1.25 GB GDDR5 |
| Memory Interface | 256-bit | 320-bit | 256-bit | 320-bit |
| Memory Bandwidth | 128.3 GB/s | 152 GB/s | 115.2 GB/s | 133.9 GB/s |
| Fillrate | 52.6 GTexels/s | 43.9 GTexels/s | 37.8 GTexels/s | 34.0 GTexels/s |
| Manufacturing Process | 40 nm TSMC | 40 nm TSMC | 40 nm TSMC | 40 nm TSMC |
| Form Factor | Dual-slot | Dual-slot | Dual-slot | Dual-slot |
| Display Outputs | 2 x DL-DVI, 1 x mini-HDMI | 2 x DL-DVI, 1 x mini-HDMI | 2 x DL-DVI, 1 x mini-HDMI | 2 x DL-DVI, 1 x mini-HDMI |
| Thermal Design Power | 170 W | 219 W | 160 W | 215 W |
On the outside, GeForce GTX 560 Ti looks a lot like GeForce GTX 460. Both cards employ axial cooling fans that blow over an aluminum array of heatsink fins on a copper base. They’re both dual-slot cards with the same array of display outputs, including two dual-link DVI connectors and one mini-HDMI port. Moreover, the two cards require a pair of six-pin auxiliary power inputs.
There are actually quite a few differences between the boards, though. For one, the GeForce GTX 560 Ti is nine inches long, while the GTX 460 is an 8.5” card. Under the plastic shroud, Nvidia arms the GTX 560 with one additional heatpipe (it now sports three) to dissipate thermal energy more quickly. There’s also a metal baseplate sitting on the board’s memory ICs and power circuitry. Previously, those components were simply left exposed under the shroud.
Granted, I’m only talking about the reference design here. We’ve already pulled several GTX 560s in-house that don’t abide Nvidia’s version of this card. The Gigabyte card I cover later in this piece, for instance, is 9.5” long, employs two fans, four heatpipes, 6+1 power phases (versus 3+1 on the stock version), and a completely different PCB.
Nvidia also arms its reference design with the same power-monitoring circuitry first introduced with GeForce GTX 580 to prevent overloading the voltage regulation circuit. The company says it’s up to each board partner to decide whether to include it on their GeForce GTX 560 Ti offerings, though.
So Yeah, About That Ti…
Depending on your age, the Ti suffix attached to Nvidia’s GeForce GTX 560 Ti may or may not make sense. The short story is that, back in 2002, Nvidia capped its “performance-oriented” cards with the Ti designator and its “budget-friendly” cards with MX. There were a number of different models in the GeForce4 Ti lineup, mostly based on the NV25 GPU, but I distinctly remember the GeForce4 Ti 4200 being the value leader. If you’re good with Google, you can even find a couple of GeForce4 Ti 4200 reviews I wrote back in 2002. Great, now I feel really old.
Anyway, Nvidia is bringing its Ti suffix back. When the company asked me what I thought about this card’s name, I rolled my eyes at first. “Cute, guys.” But the name actually makes some sense if you peer into the crystal ball. GF114 is an ASIC that could conceivably drive a number of different graphics cards. And rather than start hacking into its number scheme (GTX 555, GTX 550, and so on), I imagine we might see a GeForce GTX SE if Nvidia introduces a less-capable iteration of GF114.
I hope Nvidia will tread carefully there, though. Differentiating with a suffix isn’t as concise as using the already-accepted number system. And if you remember back to the GeForce4 days, the less-expensive MX boards got slammed for trailing the slowest GeForce3s in performance. These days, performance isn’t the end-all—we have a lot more base functionality to take into account. Even still, the potential exists to confuse the less-savvy folks who pay more attention to games than the hardware they buy.
Fortunately, I don’t think that’s what Nvidia has in mind. I’m speculating here, but the company does have GeForce GTX 460 and GF104 fresh in its memory. As capable as it was back in July, partner boards started trickling out with increasingly more aggressive clocks (and higher prices). By the time AMD launched its Barts-based GPU, Nvidia had a hard time talking anyone into testing with specifically-picked higher-end partner SKUs to compete against AMD’s reference boards. Perhaps by segmenting the GeForce GTX 560 more granularly, it’ll get around a repeat of that situation, which I know was internally frustrating.
