Recent content by Tridam

Please note that Nvidia has never been crystal clear about where those DP units sit. I did my best to represent what they were willing to share. And some details are missing to keep those diagrams readable, each 32-wide SP unit should actually be 2 16-wide units.

There was an unfortunate error in my excel sheet for the match at 185W. While the individual results were all corrects, the averages were not processed correctly which should have been obvious when looking at those graphs but as usual there was some rush involved to get that review out hehe...

I actually tried that in my usual testing (1h heavy gaming-like load inside a case). I just pushed the power limit to the max (+50%) without overclocking but that wouldn't make any difference as I hit the power limit (~280W instead of 185W and average clock of 907 MHz instead of 849 MHz). GPU...

There are many points of view about what the base and turbo/max clocks are. Engineers validating those chips, marketers selling them and reviewers looking at them of course have a different opinion. I think the main problem is this notion of turbo which doesn't really make sense when we talk...

I'm wondering if AMD has a hard time optimizing some parts of its drivers or took the decision not to spend resources to try to match Nvidia's lower D3D11 overhead / better multithreading. Development resources are always limited and right now maybe it's best to direct most of those resources...

The problem with many reviews, IMHO, is that they put those customs R9 Fury up against reference blower designs for other GPUs. It's kind of a nice gesture towards AMD hehe but it doesn't paint the whole picture. There can be a huge gap between reference designs and custom ones with the GTX...

I just noticed a new picture when checking AMD's HBM page : http://www.amd.com/en-us/innovations/software-technologies/hbm Nice piece to speculate about so I guess it was worth sharing here :) It would be compatible with a 50x50mm packaging, 32x26mm interposer, 4 HBM modules and some...

165W is some cool magic power number :cool: The actual TDP is 180W. Then it's a bit tricky. The power % reported by Afterburner or similar tools is the biggest of those 4 values : - total 12V power / TDP - 12Vbus power / 12Vbus power limit (maybe with a bias to take into account 3.3Vbus...

You can run into this kind of extra power consumption in many games actually as Gigabyte boards keep their highest boost clock in every case (even in Furmark :o). They increased significantly the TDP value in the bios, which is also the GPU Boost power target. GTX 970 ref TDP : 150W GTX 980...

http://www.hardware.fr/articles/926-24/tonga-vs-tahiti.html I've just added some extra numbers : Tonga - 28 CU @ 918 MHz - 163.9 Gio/s (256-bit @ 1375 MHz) VS Tahiti - 28 CU @ 918 MHz - 163.9 Gio/s (384-bit @ 917 MHz) Of course nothing is perfect and 384-bit @ 917 MHz is not exactly the same...

This block diagram is just a rough estimate of what Tonga should look like, to help readers understand the main differences between GPUs. It's based on current knowledge. As written in the story, Tonga as at least 28 CUs but maybe a bit more which is why some CUs are lighter. Same with the L2 as...

Based on measurements, then confirmed by Nvidia. However they wouldn't say what exact part of the process is the limitation (vertex fetch ?). Note that this high throughput can only be achieved when a single vertex has to be fetched per primitive. If you need to fetch 2 vertices, rate is one...

Looks like I answered to this second part at the same time as you were posting hehe I agree that everything points to the FP32 blending throughput being limited to 1/16. It's actually what I wrote previously.

Thanks, I fixed that! Of course the GM107 has 16 ROP as well but benefits from the bigger rasterizer and 4-5 SMM able to deliver 16-20 4 bytes pixels per clock. And you're correct that FP32 throughput is not limited by memory bandwidth, I just double checked that. I'm using my own test. It...

I won't claim obviously that all details in those diagrams are correct. There are still some details Nvidia is not willing to share that are difficult to extract from testing :/ However, compared to the official material, I think they get much closer to the reality :) 1. Nvidia confirmed to me...