No More Rumors: the Official Radeon 300 Series Specs Are Here
AMD has been leaking information on their “currently under NDA” 300 series graphics cards like a sieve. Never mind the press conference held earlier this week or demonstrations of their new GPUs at E3; it’s been a well-known fact for months now that the 300 series will be rebranded versions of existing GPUs, with some minor clock speed increases to help improve things in some cases. We’re not fans of renaming hardware and sending it out for another spin on the GPU merry-go-round, but both AMD and Nvidia have done it enough that we’re used to it by now. The fact that the major fabrication partner for high performance GPUs, TSMC, has been stuck at 28nm for three years certainly hasn’t helped matters, but regardless, we now have five “new” Radeon graphics card options in AMD’s 300 series. Let’s start with the specs table:
| AMD 300 Series Specifications | |||||
| Card | R9 390X | R9 390 | R9 380 | R7 370 | R7 360 |
| GPU | Hawaii (Grenada) |
Hawaii (Grenada) |
Tonga (Antigua) |
Pitcairn (Trinidad) |
Bonaire (Tobago) |
| GCN Version | 1.1 | 1.1 | 1.2 | 1.0 | 1.1 |
| Lithography | 28nm | 28nm | 28nm | 28nm | 28nm |
| Transistor Count (Billions) | 6.2 | 6.2 | 5 | 2.8 | 2.1 |
| Die Size (mm2) | 438 | 438 | 359 | 212 | 160 |
| Compute Units | 44 | 40 | 28 | 16 | 12 |
| Shaders | 2816 | 2560 | 1792 | 1024 | 768 |
| Texture Units | 176 | 160 | 112 | 64 | 48 |
| ROPs | 64 | 64 | 32 | 32 | 16 |
| Core Clock (MHz) | 1050 | 1000 | 970 | 975 | 1050 |
| Memory Capacity | 8GB | 8GB | 2GB/4GB | 2GB/4GB | 2GB |
| Memory Clock (MHz) | 1500 | 1500 | 1425 | 1400 | 1625 |
| Bus Width (bits) | 512 | 512 | 256 | 256 | 128 |
| Memory Bandwidth (GB/s) | 384 | 384 | 182.4 | 179.2 | 104 |
| TDP (Watts) | 275 | 275 | 190 | 110 | 100 |
| Price | $429 | $329 | $199 | $149 | $109 |
One of the interesting aspects of the 300 series is that AMD actually has new codenames for the graphics cards, i.e., Hawaii is now Grenada. But don’t let that codename fool you; just like Curacao was the same core hardware as Pitcairn last year, we’re still looking at GPUs that have already seen the light of day. There might have been a respin of the die to tweak a few minor details, but fundamentally performance at the same clocks shouldn’t change. While that might be disappointing for the high-performance enthusiasts, rest assured that AMD has also given plenty of information on their upcoming Fury X graphics card that will launch next week. Basically, the 300 series launch is the calm before the storm.
There are a few items worth calling out, even with these recycled GPUs. For one, the R9 390X and R9 390 are only shipping in 8GB configurations. That’s good for those times when you’re pushing so many textures at such a high resolution that 4GB VRAM proved insufficient, but AMD already quietly launched an 8GB R9 290X last year. Perhaps even stranger, the 390X MSRP is currently $50 higher than the least expensive 8GB 290X, though once supplies dry up that may change—or the street prices on 390/390X may end up lower than the MSRP. The other item worth noting is that both the 390X and the 390 are now sporting official TDPs of 275W—25W higher than previously. As far as clock speeds go, nearly all of the 300 series GPUs are seeing a 50MHz increase to core clocks speeds (give or take a few MHz), and that’s about it. The one major exception is that the R9 390X/390 now clock the GDDR5 at 1500MHz (effectively 6000MHz), a rather impressive 20 percent increase over the stock 1250MHz clock of the 290X/290.
Moving down to the R7 series, the 370 now holds the distinction of being the oldest GPU core in AMD’s lineup, with the Trinidad/Pitcairn core dating all the way back to March 2012 with the HD 7850—and strangely, we’re missing the full 1280 core version, so we might even see a 370X in the future. These are all basically baby steps, and while there will be 2GB and 4GB models of the 370, this time we’re not as worried about having additional VRAM—the GPU isn’t going to be fast enough to benefit from having 4GB in most games. Then there’s the R7 360, which uses the Bonaire core that’s only two years old.
If you’re wondering, the main advantage of GCN 1.1 is that it added TrueAudio and an improved version of AMD’s PowerTune technology. The bigger jump came with GCN 1.2, which made some architectural tweaks and improved efficiency, including lossless delta color compression, improved tessellation, a higher quality video scaler, and a new multimedia engine. To date, GCN 1.2 is only found in the Tonga GPUs, but the Fury X will likely use the same features. Finally, it’s worth noting that the R9 300 cards all support bridgeless CrossFireX now, routing data over the PCIe bus via AMD’s XDMA instead of using a bridge connector; the two R7 GPUs however still use a bridge connector.
AMD didn’t sample any GPUs to the press for this launch, leaving it up to their board partners. We were able to get the Sapphire Tri-X R9 390X, but since it’s not a stock configuration we’re covering that in a separate review. The takeaway here is that the 300 series refresh is mostly about making the model numbers look new and current. Yes, performance is going to be a bit higher, and the added memory will certainly prove beneficial on certain models, but the Fury X is what everyone really wants to see. And for that, we have to wait another week….
Follow Jarred on Twitter
Radeon R9 390X Reference Card
Radeon R9 380 Reference Card
Radeon R7 370 Reference Card
From maximumpc
from http://bit.ly/1L2p8OP
