So, everyone is taking a guess as what the causes of this situation might be. Really, the things are simple if you look carefully at the specs. Quoting Hexus on this:
1. Titan X has 6,611 GFLOPS of SP compared to 8,602 for the Fury. So here is the first offset.
2. Titan X has, ready for it, 207 GFLOPS of DP compared to 537 for the Fury. But both of them are under what the 290X is capable (739GFLOPS DP). So here is the second offset.
3. The ROPS are for sure the key element and here I will quote the folks at techreport:
4. HBM memory is wider but slower
"In other respects, including peak triangle throughput for rasterization and pixel fill rates, Fiji is simply no more capable in theory than Hawaii. As a result, Fiji offers a very different mix of resources than its predecessor. There's tons more shader and computing power on tap, and the Fury X can access memory via its texturing units and HBM interfaces at much higher rates than the R9 290X.
In situations where a game's performance is limited primarily by shader effects processing, texturing, or memory bandwidth, the Fury X should easily outpace the 290X. On the other hand, if gaming performance is gated by any sort of ROP throughput—including raw pixel-pushing power, blending rates for multisampled anti-aliasing, or effects based on depth and stencil like shadowing—the Fury X has little to offer beyond the R9 290X. The same is true for geometry throughput"
So, they increased the efficiency by removing some of the DP hardware compared to Hawaii. Still, they didn't do as much of a cut-down as NVIDIA did with Maxwell. Maxwell is so efficient because it's releaved of the DP hardware. Basically, that is how nvidia got this efficiency jump from Kepler to Maxwell, so that's not magic. The second part is indeed related to the ROPS and I guess they could've taked a bit from the shaders (3584 instead of 4096) and raise the number of ROPS to 96. Well, they could've but they didn't.
Now, the HBM part is tricky. This must be done at the driver level, and we know that AMD is a master guru when it comes to optimizing them. Only time will tell.
In conclusion, nvidia got rid of most of the unimportant bits and used those savings (which are huge, btw) to increase graphics horsepower. So, kudos to nvidia for doing this, they always pursuit what's the best even if that means losing something else. Still, considering all these things, I think AMD has done a very good job.
And also, stop looking at the maximum power consumption of this card. If you read the description of this test you'll see that it's just a furmark test, which does not mean anything. The real power consumption you should follow is the peak or the average.
Cheers!
1. Titan X has 6,611 GFLOPS of SP compared to 8,602 for the Fury. So here is the first offset.
2. Titan X has, ready for it, 207 GFLOPS of DP compared to 537 for the Fury. But both of them are under what the 290X is capable (739GFLOPS DP). So here is the second offset.
3. The ROPS are for sure the key element and here I will quote the folks at techreport:
4. HBM memory is wider but slower
"In other respects, including peak triangle throughput for rasterization and pixel fill rates, Fiji is simply no more capable in theory than Hawaii. As a result, Fiji offers a very different mix of resources than its predecessor. There's tons more shader and computing power on tap, and the Fury X can access memory via its texturing units and HBM interfaces at much higher rates than the R9 290X.
In situations where a game's performance is limited primarily by shader effects processing, texturing, or memory bandwidth, the Fury X should easily outpace the 290X. On the other hand, if gaming performance is gated by any sort of ROP throughput—including raw pixel-pushing power, blending rates for multisampled anti-aliasing, or effects based on depth and stencil like shadowing—the Fury X has little to offer beyond the R9 290X. The same is true for geometry throughput"
So, they increased the efficiency by removing some of the DP hardware compared to Hawaii. Still, they didn't do as much of a cut-down as NVIDIA did with Maxwell. Maxwell is so efficient because it's releaved of the DP hardware. Basically, that is how nvidia got this efficiency jump from Kepler to Maxwell, so that's not magic. The second part is indeed related to the ROPS and I guess they could've taked a bit from the shaders (3584 instead of 4096) and raise the number of ROPS to 96. Well, they could've but they didn't.
Now, the HBM part is tricky. This must be done at the driver level, and we know that AMD is a master guru when it comes to optimizing them. Only time will tell.
In conclusion, nvidia got rid of most of the unimportant bits and used those savings (which are huge, btw) to increase graphics horsepower. So, kudos to nvidia for doing this, they always pursuit what's the best even if that means losing something else. Still, considering all these things, I think AMD has done a very good job.
And also, stop looking at the maximum power consumption of this card. If you read the description of this test you'll see that it's just a furmark test, which does not mean anything. The real power consumption you should follow is the peak or the average.
Cheers!
it's like Godwin's law for tech forums
