Yes..m12ii will be loud when under load. I used it (750w) for my GTX970 before, only 350w load also consider loud. Finally changed to Enermax Revolution87 750w, now ok already.

Funny how NVIDIA actually losses performance when going for DX12 API at Ashes of Singularity benchmark.. maybe they optimized their DX11 drivers too well?

Support async or not, what's important is the actual FPS of the game.. if anyone is quoting Ashes of Singularity benchmark saying AMD has better implementation, check again.. R9 390X is also performing close to R9 Fury too. (source: http://www.pcgameshardware.de/Ashes-of-the...tX-11-1167997/) It is just that AMD DX11 implementation is so bad that it makes DX12 looks very good.

One thing we know for sure, currently NVIDIA has the market share (82%!). Who knows what will happen to future DX12 games, especially those GameWorks titles.

Huh? You mean the GTX970 has physically only 3.5GB GDDR5? Ok...thanks for info

This post has been edited by Demonic Wrath: Sep 1 2015, 11:22 PM

Since both all graphics hardware vendor doesn't have a card that can fully support DX12, just wait till 2016 or 2017 to get a full DX12 feature support card.. but then again, maybe that time got DX12.1 feature?

Edit: For those who really needs full DX12 support, just wait till DX12 games are released first then only see benchmarks..

This post has been edited by Demonic Wrath: Sep 2 2015, 09:35 AM

From AoTS benchmark, it seems that NVIDIA's DX11 without async support performs faster/similarly to AMD's DX12 with async compute. (source) edit: see frame rate by batch type.

Secondly, Anandtech (source) previously show that NVIDIA does benefit from DX12 implementation too. StarSwarm is also developed by Oxide.

This post has been edited by Demonic Wrath: Sep 2 2015, 12:35 PM

Apparently from latest update, the supposedly "Asyncompute.exe" program doesn't send compute workload to the compute queue on NVIDIA cards..whereas PhsyX games do async compute? source

So latest update on the Async situation:-

source

Just state your budget and I'm sure someone can help you to select the best performance/price card.. specifying all game apps is very vague (you want to run highest setting at 4K? or lowest setting at 720p? or PS4 like performance?)

I completed Batman AK on PC long time ago. Very smooth gameplay..

Installed on my SSD to play (pagefile also need to be on SSD, or else will stutter too). Heavy write on SSD due to writing on pagefile >.<

This post has been edited by Demonic Wrath: Sep 12 2015, 01:44 PM

Extract new driver from GFE? Currently GFE downloads the driver exe same as what you download from Geforce.com website.

You can see all the drivers downloaded from GFE here: C:\ProgramData\NVIDIA Corporation\NetService

Like Najmod said, you can probably get the driver from other reliable 3rd party source such as Techpowerup or Guru3D since they have been hosting the driver for quite a while now..

Actually what kills the performance is Shadow Distance only... Putting it to Ultra makes my FPS drop to 30-40s even when the GPU usage is only ~60%..

Why not use HBAO+?

Also, no SLI?

Actually there's a SLI profile for Rainbow Six Siege.. It's not working because the new R6siege exe (rainbowsix.exe) shares the same profile as old Rainbow six game.

Need to fix using NVIDIA inspector.

The whitepaper details on the GPU to GPU connection (p2p) only. The connection to CPU is still through PCI-E.

NVLink for CPU-GPU is being implemented by IBM for supercomputers only.

IMHO, if NVIDIA want to implement NVLink succesfully, then it can be through:-
i) akin to SLI bridge
ii) Dual GPU cards

It seems unlikely that NVIDIA will introduce 2 variants of GPU i.e. 1 for PCI-e interface, 1 for "NVLink" interface..

This post has been edited by Demonic Wrath: Dec 10 2015, 04:08 PM

Last time my MSI GTX460 also need RMA, I contacted NVIDIA straight and they forwarded my request to MSI for RMA. A lot faster..

Oh...my 460 was still under warranty that time..

Yea buy a GTX970 and physically take out 512mb.

Lol....price for GPU really crazy now...last year bought GTX970 MSI gaming is RM 1400+.. now 2nd hand is RM 1500.. new is RM 1800

who says computer parts is not investment ;p

On the technical side,
1) Chip design nowadays are getting more and more complicated (billions transistor count) and there are limitations in fabrication process and also heat issue.
2) Debugging these chips will take longer time. If they fast track it and release a buggy chip, they stand to lose more from RMA and bad rep.
3) It is right to salvage chip defect and reuse as lower end parts. Less wastage, recover R&D cost and fabrication cost.
4) Architecture performance improvement and/or power saving is normally achieved by fine tuning the architecture, better power gating or new ways to perform the task. Over the years, technology company already fine tuned a lot of aspect in their architecture, leaving less and less things to fine tune. Intel's CPU core improvement is generally only ~10% from architecture to architecture (instead of having more cores, it's really hard to increase performance for each core given the same thermal envelope).

From the above points, you can see design process is getting longer and longer, this also translate to R&D cost getting higher and higher. This is why some company choose to rebrand their products. IMO, Nvidia is already doing a lot of enhancement on both hardware architecture designs and software side (CUDA/Gameworks) in order to optimize the performance possible. (Anyone may argue open source is better, but let's face it, almost no investor and hardware vendor will be willing to invest too much time/effort on open source things. Open source = broad compatibility = hard & longer to optimize.)

TLDR, long product cycle for GPU is good since this forces the developer to use smarter technique and optimize to achieve better results instead of just brute-forcing it. This is why console graphics are generally on par/or better with PC graphics, even at a (much) lower performance capability.

Hmm...in the video it mentioned it is not being gimped. Why would you think otherwise.