The Speeds And Feeds
As you can see, Nvidia's reference design on the far right is the only board in this round-up based on that design. Everyone else used different coolers, core clock rates, and, in a few cases, memory clock rates.
| Header Cell - Column 0 | Asus | Gainward | Galaxy | Palit | Gigabyte | Zotac | Nvidia |
|---|---|---|---|---|---|---|---|
| CUDA Cores | 1344 | ||||||
| Texture Units | 112 | ||||||
| ROPs | 32 | ||||||
| GPU Clock | 1059 MHz | 1006 MHz | 1006 MHz | 1006 MHz | 980 MHz | 1098 MHz | 915 MHz |
| Boost Clock | 1137 MHz | 1085 MHz | 1085 MHz | 1085 MHz | 1059 MHz | 1167 MHz | 980 MHz |
| Pixel Fill Rate | 29.7 GPix/s | 28.2 GPix/s | 28.2 GPix/s | 28.2 GPix/s | 27.4 GPix/s | 30.7 GPix/s | 25.6 GPix/s |
| Texture Fill Rate | 118.6 GTex/s | 112.7 GTex/s | 112.7 GTex/s | 112.7 GTex/s | 109.8 GTex/s | 123 GTex/s | 102.5 GTex/s |
| Memory Clock | 1502 MHz | 1527 MHz | 1502 MHz | 1527 MHz | 1502 MHz | 1652 MHz | 1502 MHz |
| Memory Bus | 256-bit | ||||||
| Memory Bandwidth | 192.3 GB/s | 195.5 GB/s | 192.3 GB/s | 195.5 GB/s | 192.3 GB/s | 211.5 GB/s | 192.3 GB/s |
| Memory | 2 GB GDDR5 | ||||||
| Die Size | 294 mm² | ||||||
| Transistors | 3.54 Billion | ||||||
| Power Connectors | 6 + 6-pin | 6 + 6-pin | 8 + 6-pin | 6 + 6-pin | 8 + 6-pin | 6 + 6-pin | 6 + 6-pin |
| Street Price | $430 | Not Available | $440 | Not Available | $400 | $440 | $400 |
Pricing and Availability
It's difficult to keep a sharp bead on graphics card prices; they move every day, it seems. The best we could do here was check availability the day before this story went live in the U.S. and take a quick snapshot of the prices.
Nobody seems to have the Gainward or Palit cards in stock, meaning they're really only here for exhibition, since you can't get your hands on them even if you wanted to. One of the other models currently sells for the same $400 Nvidia charges for its reference design, while the other three cost notably more. We're just happy that the GeForce GTX 670 launch wasn't a predominantly paper-oriented affair like GeForce GTX 680, with miniscule quantities trickling out from Nvidia's board partners. Many of these models appear to be readily available, in contrast.
There's a budget GeForce GPU selling in China that not even Nvidia knew it made — RTX 4010 turns out to be a modified RTX A400 workstation GPU
US to patch loopholes that allow China to buy banned AI GPUs from other countries — new regulations include national quotas on GPU exports and a global licensing system
