is hardware Accelerated GPU Scheduling really worth?

Hi there
While the "feature" might make no discernable difference to the quality of the video you see -- anything that offloads work from the main processor allows the main processor to run more tasks concurrently. Hi level graphics is usually quite a process intensive bound function so removing that away from the main CPU allows the CPU to do a lot more other things.

For most people though - especially on anything but the very latest machines , the biggest improvement you can make on a home computer for everyday use is to improve the Disk system -- except for large volumes of relatively static storage (e.g Music, Video films etc) and external backup storage all HDD's should be replaced via SSD's or even NVMe type SSD's if poss.

Over 30 years of using Windows - by far the biggest performance issues I've seen have all been to do with hideously slow consumer grade HDD's (5400 RPM , tiny or non existent cache, incredibly slow access times etc ).

Slow disks would kill even the fastest Cray supercomputer !!!!

Cheers
jimbo

jimbo45 said:

Hi there
While the "feature" might make no discernable difference to the quality of the video you see -- anything that offloads work from the main processor allows the main processor to run more tasks concurrently. Hi level graphics is usually quite a process intensive bound function so removing that away from the main CPU allows the CPU to do a lot more other things.

For most people though - especially on anything but the very latest machines , the biggest improvement you can make on a home computer for everyday use is to improve the Disk system -- except for large volumes of relatively static storage (e.g Music, Video films etc) and external backup storage all HDD's should be replaced via SSD's or even NVMe type SSD's if poss.

Over 30 years of using Windows - by far the biggest performance issues I've seen have all been to do with hideously slow consumer grade HDD's (5400 RPM , tiny or non existent cache, incredibly slow access times etc ).

Slow disks would kill even the fastest Cray supercomputer !!!!

Cheers
jimbo

Finally someone who gets it. I've been "playing"(tweaking) windows for performance since windows xp was in beta. I've tried every guide, tweak, reg edit, forum post out there. Nothing competes with switching from a spinner to a ssd. I have always said the single handed biggest performance jump in consumer computing in the past 20 years was the release of the ssd.

Mumbai said:

I have a Turing GPU,Intel 10th gen CPU, and Windows 10 1803 version. I am wondering is it worth to upgrade windows to latest version, and turn on hardware Accelerated GPU Scheduling? Will it improve my fps of games? if so of which games? new or old games? if so by how much? has anyone tested gaming performance after turning on hardware Accelerated GPU Scheduling fairly recently?

Faster Gaming for FREE - Hardware GPU Scheduling Explained - YouTube

So i know this is an old thread but I wanted to post something that just happened to me, and why I believe this setting may be more important than people may realize.

I just built a new custom PC that has IMHO fairly high spec's - 5900x, 3080, x570, 32 GB RAM, and tons of high speed storage - you'd think a rig like this could run Cyberpunk quite well, especially only at 1440p. Well, I finally got around to installing it yesterday, and i was immediately surprised to see some weird screen artifacts - specifically white horizontal lines that looked somewhat like static periodically appearing on my display. At first I thought it was part of the "cyberpunk" look of the game, kinda like TV scan lines or film grain, but then I realized it was not just in the game and was also showing on my windows desktop, but only when the game was running in a window. Well I spent a ton of time following different instructions on uninstalling old display drivers, customizing the graphics / display settings in game, all to no avail.

Eventually I thought to look in my windows display settings, because I kind of had a feeling there was some setting conflict between the OS, the Nvidia control panel, and the game, and I come across the "hardware accelerated gpu scheduling" toggle, which was set to off. Turning that on seems to have fixed my problem when nothing else did, and now the game runs BEAUTIFULLY. So, I'm not sure what other people may have found, but in my opinion this setting is a pretty big deal and should always be tried if you experience any weird screen issues while gaming.

I feel its testing has been done in the wrong way and with the wrong games.

The description indicates it will thrive in environments where CPU is the limiting factor, but not when GPU is limiting factor or if is no limiting factor.

So logically the best way to test HAGS is to run old DX9 games on a low end dual core processor with low clocked speeds.

What do the reviewers do? They get out their 9900k testing rig and pick games from their DX11/12 test suite lol. Like they cant think out of the box. They also ran games at uncapped framerates which loads up the GPU.

I have a high end CPU but tested this on known problematic games in the PC space.

Lightning returns - DX9 game, single CPU thread.
Tales of Zestiria, micro stutter hell, DX9 game. Single CPU thread.
FF13-2, DX9 game stutters when loading assets. Two CPU threads, but first one taking the brunt.

I also tested FF7 remake, DX11 and DX12 game, has known stutter problems which are much worse in DX12 mode.

On Zestiria , micro stutters gone, killed, *poof* whatever way you want to describe it, can flip it on and off to see it consistently, Lightning Returns it removes some stutters but no effect on others, no regressions anywhere.
FF13-2, similar to Lightning Returns, helped in some areas, no effect elsewhere.

FF7 remake, no impact, its issues are caused I think by dynamically compiled shaders and texture thrashing so HAGS isnt expected to help much here I suppose.

I expect improvements for me would have been more pronounced on a weaker CPU so I might restest with most cores disabled and clocks capped to 2ghz to emulate a laptop.

Tomshardware did put at end of his article he regretted not testing on a low end CPU but also said he couldnt be bothered to retest so didnt.