Exactly, Maxwell is entirely dependant on the driver. This also gives it a bit of flexibility in terms of driver optimizations.
Which is why NVIDIA will likely fine tune their DX12 driver for Ashes soon enough.
In other words: AMD GCN can adapt itself to application without the interference of driver.
Nvidia hardware MUST use driver to have proper scheduling, because without it - MegaThread Engine has no clue what to do with the application.
In other words: AMD GCN can adapt itself to application without the interference of driver.
Nvidia hardware MUST use driver to have proper scheduling, because without it - MegaThread Engine has no clue what to do with the application.
The Gigathread engine is not a processor. If it were, it would be called the "Gigathread Processor". It does not execute tasks, it waits for an available SMM to signal it for work.
Like I said, NVIDIAs driver re-orders grids (order of instructions in a warp). That's static scheduling. AMD, on the other hand, does this process in hardware.
It is a term created by nVidia. You are still make things up with your fanfiction.
BTW: It is called "GigaThread Engine". :\
http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review/3
This is not was nVidia is doing in the end. The compiler is just marking certain instructions which should not be issued before other. Scheduling of instructions as warps happens on the GPU. Otherwise something like Dynamic Paralellism wouldnt work...
https://www.nvidia.com/content/PDF/kepler/NVIDIA-Kepler-GK110-Architecture-Whitepaper.pdf
The end result is an interesting one, if only because by conventional standards it’s going in reverse. With GK104 NVIDIA is going*back*to static scheduling. Traditionally, processors have started with static scheduling and then moved to hardware scheduling as both software and hardware complexity has increased. Hardware instruction scheduling allows the processor to schedule instructions in the most efficient manner in real time as conditions permit, as opposed to strictly following the order of the code itself regardless of the code’s efficiency. This in turn improves the performance of the processor.
However based on their own internal research and simulations, in their search for efficiency NVIDIA found that hardware scheduling was consuming a fair bit of power and area for few benefits. In particular, since Kepler’s math pipeline has a fixed latency, hardware scheduling of the instruction inside of a warp was redundant since the compiler already knew the latency of each math instruction it issued. So NVIDIA has replaced Fermi’s complex scheduler with a far simpler scheduler that still uses scoreboarding and other methods for inter-warp scheduling, but moves the scheduling of instructions in a warp into NVIDIA’s compiler. In essence it’s a return to static scheduling.
Ultimately it remains to be seen just what the impact of this move will be. Hardware scheduling makes all the sense in the world for complex compute applications, which is a big reason why Fermi had hardware scheduling in the first place, and for that matter why AMD moved to hardware scheduling with GCN. At the same time however when it comes to graphics workloads even complex shader programs are simple relative to complex compute applications, so it’s not at all clear that this will have a significant impact on graphics performance, and indeed if it did have a significant impact on graphics performance we can’t imagine NVIDIA would go this way.
And if they had lied, we'd be seeing Fermi-like power usage@sontin
You used to argue vehemently that Maxwell could do Async Compute in DX12. You were wrong.
Perhaps you are also wrong on the software scheduler aspect of Maxwell too, of which the evidence is overwhelming. Has that thought crossed your mind?
Really, just read this, this is info that NV briefed to tech sites like AnandTech:
Someone with a good grasp on English would say there was a major change from Fermi's hardware scheduler, to a software scheduler in Kepler and later Maxwell. Unless NV lied to tech journalists, which then you could be correct.
Early on it this thread I posted my Benchmarks for my rigs below for both of my rigs below. I have now updated to Beta2 and updated my Radeon drivers with the latest hotpatch.
Ashes now supports multiple gpus so now both R9 290s are running in CF (previously a single R9 290 produced 36.4 in DX11 and 39.3 in DX12).
Now the scores for the 4790K @4.7Ghz and 2 R9 290s in CF is:
DX12 only as DX11 does not support CF 58.6fps overall while the single GTX980TI produces 45.4fps in DX12 and 46.4 fps in DX11.
Early on it this thread I posted my Benchmarks for my rigs below for both of my rigs below. I have now updated to Beta2 and updated my Radeon drivers with the latest hotpatch.
Ashes now supports multiple gpus so now both R9 290s are running in CF (previously a single R9 290 produced 36.4 in DX11 and 39.3 in DX12).
Now the scores for the 4790K @4.7Ghz and 2 R9 290s in CF is:
DX12 only as DX11 does not support CF 58.6fps overall while the single GTX980TI produces 45.4fps in DX12 and 46.4 fps in DX11.
Now you should do all 3 GPUs!
Someone did a test with 3 970s and it didn't scale above 2 970s.
This is only a joke. The real sh1tst0rm will begin with the Directx12 and the DX12 support of Maxwell cards. I can't tell you anything, but after april a lot of things will be disclosed.
Someone did a test with 3 970s and it didn't scale above 2 970s.
Saw it coming way back. And NV got away with lying about hardware specs, again, dominating marketshare.
By the time Pascal launches, they can simply say "With Pascal, we've improved on DX12 Async Compute support, on the hardware level over Maxwell", and folks will upgrade in droves.
IF ms is pushing DX12 hardcore, NV cannot win. way bigger wallet.well about that i have doubts, to me that seems like one of the reasons gameworks exist´s, according to their PR they only have NVLINK and mixed precision, mixed precision is something AMD already have pushed, and whit NVLINK you have to buy a new mobo, so to me pascal seems more like Maxwell the other half.
So Gameworks will keep games on dx11, but the consoles are pushing this, is a cuestion of money will NVIDIA have enough money to push industry back, while MICROSOSFT is trying to put the carrot on win10-dx12 and keeping the stock high.
my opinion is no, so they best bet is to spend money on sites and forums
Really? MS?IF ms is pushing DX12 hardcore, NV cannot win. way bigger wallet.
Saw it coming way back. And NV got away with lying about hardware specs, again, dominating marketshare.
By the time Pascal launches, they can simply say "With Pascal, we've improved on DX12 Async Compute support, on the hardware level over Maxwell", and folks will upgrade in droves.
Do your really believe this is how it will turn out?
Again, as asked several times. What happens when the DX12 reality isn't like the AMD sponsored AOTS that sold 20K copies?
Do your really believe this is how it will turn out?
Again, as asked several times. What happens when the DX12 reality isn't like the AMD sponsored AOTS that sold 20K copies?
I thought people knew better by now than trusting anything from the SA site.