Is Radeon REALLY better for old PCs? | Driver Overhead

[RDNAuser]

HUB already covered this but having a refreshed look at this is always good.

 

[oxrufiioxo]

I'd say yes and no.... While pairing an older CPU with a 6000/7000 series AMD card might give you a better experience in some games vs it's direct Nvidia competitor you likely will be bottlenecked regardless of card being used so a full system upgrade is likely the better option and just getting whatever GPU you want.


Bottom line I wouldn't buy an AMD card just for this reason it would have to tick all the other performance boxes in what I would be looking for in a gpu for me to want to buy it.

 

[Dr. Dro]

The situation is always evolving, but in short, yes, Radeon's design is friendlier towards machines with limited CPU power with the tradeoff of having less potential when ample CPU power is available. Like I mentioned on the Valley thread, AMD's DirectX 11 driver operates in an immediate context. This means that it issues commands through a limited amount of, and primarily one CPU thread. These commands are then processed by the GPU's hardware command processor itself. It is highly optimized, and thus quite fast, but there are situations where this approach doesn't work as well, notably, when you're talking about Bethesda's Creation Engine games.

Nvidia's driver, on the other hand, is capable of parsing an instance where an immediate context is created, generate driver command lists and defer these commands through multiple CPU threads. This results in an increased CPU load, but significantly higher potential, as it will be able to handle a much, much, *much* higher amount of draw calls than an immediate context would. While a little older now, this is a good technical explanation by Intel, I had been talking about this with another forum member shortly before you joined:

Game Development

Access tools, tutorials libraries, and code samples from Intel to optimize your games.

software.intel.com

Support for driver command lists is detected by GPU-Z and shown in the Advanced > DirectX 11 tab:

3DMark's old API overhead test is an useful tool to show the result of this behavior, you will find that NVIDIA GPUs have substantially higher performance in that benchmark, despite that not reflecting in real world applications. Thus, the UL guys dropped support, as it's a very technical tool which does not really serve to compare between different setups. The DirectX SDK samples will also contain examples where this will be quite apparent.

In DirectX 12 or Vulkan, this is irrelevant: they should be significantly closer apart, with NVIDIA's driver again using a little more CPU in exchange for slightly stronger performance if it can be helped, but nowhere near the drastic difference between vendors when it comes to DirectX 11. This is because they are low-level APIs and this abstraction is done in a completely different manner. Also, for the record: Intel does not support driver command lists in their Gen 12.2 architecture integrated graphics (such as UHD 770). Unsure if Arc does, but I would guess it does not either. Would appreciate if someone could clarify.

 

[RDNAuser]

The situation is always evolving, but in short, yes, Radeon's design is friendlier towards machines with limited CPU power with the tradeoff of having less potential when ample CPU power is available. Like I mentioned on the Valley thread, AMD's DirectX 11 driver operates in an immediate context. This means that it issues commands through a limited amount of, and primarily one CPU thread. These commands are then processed by the GPU's hardware command processor itself. It is highly optimized, and thus quite fast, but there are situations where this approach doesn't work as well, notably, when you're talking about Bethesda's Creation Engine games.

Nvidia's driver, on the other hand, is capable of parsing an instance where an immediate context is created, generate driver command lists and defer these commands through multiple CPU threads. This results in an increased CPU load, but significantly higher potential, as it will be able to handle a much, much, *much* higher amount of draw calls than an immediate context would. While a little older now, this is a good technical explanation by Intel, I had been talking about this with another forum member shortly before you joined:

Game Development

Access tools, tutorials libraries, and code samples from Intel to optimize your games.

software.intel.com

Support for driver command lists is detected by GPU-Z and shown in the Advanced > DirectX 11 tab:

View attachment 297658

3DMark's old API overhead test is an useful tool to show the result of this behavior, you will find that NVIDIA GPUs have substantially higher performance in that benchmark, despite that not reflecting in real world applications. Thus, the UL guys dropped support, as it's a very technical tool which does not really serve to compare between different setups. The DirectX SDK samples will also contain examples where this will be quite apparent.

In DirectX 12 or Vulkan, this is irrelevant: they should be significantly closer apart, with NVIDIA's driver again using a little more CPU in exchange for slightly stronger performance if it can be helped, but nowhere near the drastic difference between vendors when it comes to DirectX 11. This is because they are low-level APIs and this abstraction is done in a completely different manner. Also, for the record: Intel does not support driver command lists in their Gen 12.2 architecture integrated graphics (such as UHD 770). Unsure if Arc does, but I would guess it does not either. Would appreciate if someone could clarify.

Thank you for this information, very helpful!

 

[R-T-B]

Just don't pair an old PC with an old unsupported Radeon, as until like this year, both DX11 and OGL overhead was absurdly bad.

 

[QuietBob]

I had been talking about this with another forum member shortly before you joined

For the context, here's our original conversation

 

[Avro Arrow]

Adrenalin drivers do have less CPU overhead but not all Radeons are good for older PCs. For example, the RX 6500 XT has a PCI-Express 4.0 x4 interface which means that motherboards with PCI-Express v3.0 or older will choke the card's performance. If you're comparing a GeForce card with a Radeon and they both have x16 interfaces, then yes, the Radeon is absolutely the better choice.

 

[gt362gamer]

The situation is always evolving, but in short, yes, Radeon's design is friendlier towards machines with limited CPU power with the tradeoff of having less potential when ample CPU power is available. Like I mentioned on the Valley thread, AMD's DirectX 11 driver operates in an immediate context. This means that it issues commands through a limited amount of, and primarily one CPU thread. These commands are then processed by the GPU's hardware command processor itself. It is highly optimized, and thus quite fast, but there are situations where this approach doesn't work as well, notably, when you're talking about Bethesda's Creation Engine games.

Nvidia's driver, on the other hand, is capable of parsing an instance where an immediate context is created, generate driver command lists and defer these commands through multiple CPU threads. This results in an increased CPU load, but significantly higher potential, as it will be able to handle a much, much, *much* higher amount of draw calls than an immediate context would. While a little older now, this is a good technical explanation by Intel, I had been talking about this with another forum member shortly before you joined:

Game Development

Access tools, tutorials libraries, and code samples from Intel to optimize your games.

software.intel.com

Support for driver command lists is detected by GPU-Z and shown in the Advanced > DirectX 11 tab:

View attachment 297658

3DMark's old API overhead test is an useful tool to show the result of this behavior, you will find that NVIDIA GPUs have substantially higher performance in that benchmark, despite that not reflecting in real world applications. Thus, the UL guys dropped support, as it's a very technical tool which does not really serve to compare between different setups. The DirectX SDK samples will also contain examples where this will be quite apparent.

In DirectX 12 or Vulkan, this is irrelevant: they should be significantly closer apart, with NVIDIA's driver again using a little more CPU in exchange for slightly stronger performance if it can be helped, but nowhere near the drastic difference between vendors when it comes to DirectX 11. This is because they are low-level APIs and this abstraction is done in a completely different manner. Also, for the record: Intel does not support driver command lists in their Gen 12.2 architecture integrated graphics (such as UHD 770). Unsure if Arc does, but I would guess it does not either. Would appreciate if someone could clarify.

You seem to be quite right! I upgraded my GPU from a 4G Nvidia GTX 960 to a 12G AMD 6700 XT, having an i5 4690k at 4.3 Ghz (4 cores without hyperthreading, so 4 threads total), 2400 Mhz (Or should I say 2400MT/s, anyway...) Cl11 2x8 RAM sticks and a 960 evo nvme drive, and the AMD driver seem to significantly reduce stuttering on Assetto Corsa Competizione (probably the extra gpu power and video memory helps too, as it usually consumes about 7G in the game, with mixed settings and 1440p internal resolution [133% resolution scale], with the GTX 960 I was using similar if not exact settings at internal 1080p resolution, and the video ram usage is not much lower like this, still way above 4G). My target is 60 fps as I'm aware of the CPU/system RAM limitations and my monitor (tv really) is 60hz. A few months back I upgraded from 2x8 1600cl10 as one of the sticks started failing, and that upgrade was noticeable, and this also was, probably more.

Unfortunately it seems the best would be to have a toggle in driver settings to have either Nvidia approach for older not very well threaded (and hopefully not very CPU demanding) games, or the less CPU consuming AMD approach, but I don't see that happening anytime soon. It may be placebo but I feel like I lost a bit of performance for instance in a very old Directx 9 game, NFS: MW from 2005, probably because here the Nvidia driver is capable of squeezing more draw calls out if it.

The performance loss in the lower end of the CPU spectrum due to the Nvidia drivers in more or less recent demanding games can be very noticeable. Look at this comparison in Battlefield V, the frametime graph... :

PD: It's clearly not ideal to pair the 6700 xt with an i5 4690k, but I couldn't afford a whole PC upgrade, still upgrading was noticeable.

 

[Space Lynx]

this is interesting, never even heard of any of this until I just stumbled on this thread.

i doubt it matters all that much, but it is nice to know. i like my all AMD setup, everything seems very fluid so far regardless of what I throw at it

 

[phanbuey]

Low end Radeon gaming boxes are very strong 6700xt rigs that I built did feel snappy as hell.

 

[Dr. Dro]

You seem to be quite right! I upgraded my GPU from a 4G Nvidia GTX 960 to a 12G AMD 6700 XT, having an i5 4690k at 4.3 Ghz (4 cores without hyperthreading, so 4 threads total), 2400 Mhz (Or should I say 2400MT/s, anyway...) Cl11 2x8 RAM sticks and a 960 evo nvme drive, and the AMD driver seem to significantly reduce stuttering on Assetto Corsa Competizione (probably the extra gpu power and video memory helps too, as it usually consumes about 7G in the game, with mixed settings and 1440p internal resolution [133% resolution scale], with the GTX 960 I was using similar if not exact settings at internal 1080p resolution, and the video ram usage is not much lower like this, still way above 4G). My target is 60 fps as I'm aware of the CPU/system RAM limitations and my monitor (tv really) is 60hz. A few months back I upgraded from 2x8 1600cl10 as one of the sticks started failing, and that upgrade was noticeable, and this also was, probably more.

Unfortunately it seems the best would be to have a toggle in driver settings to have either Nvidia approach for older not very well threaded (and hopefully not very CPU demanding) games, or the less CPU consuming AMD approach, but I don't see that happening anytime soon. It may be placebo but I feel like I lost a bit of performance for instance in a very old Directx 9 game, NFS: MW from 2005, probably because here the Nvidia driver is capable of squeezing more draw calls out if it.

The performance loss in the lower end of the CPU spectrum due to the Nvidia drivers can be very noticeable. This is an example of this happening in Battlefield V, the frametime graph... :

An option to disable threaded optimization in the Nvidia drivers seems to exist, but I've never tried it.

 

[Space Lynx]

An option to disable threaded optimization in the Nvidia drivers seems to exist, but I've never tried it.

interesting, this is an option in sid meir's civilization beyond earth, i was playing it the other day and noticed it said "threaded optimization" in settings, and i was like wth is that lol

 

[delshay]

My current pairing of old PC is Phenom 1100T/Vega 56 Nano.

Because I'm a hardware modder, I can extract more performance than most users when it comes to old hardware. It's built to be extraordinary snappy.

 

[Assimilator]

You seem to be quite right! I upgraded my GPU from a 4G Nvidia GTX 960 to a 12G AMD 6700 XT, having an i5 4690k at 4.3 Ghz (4 cores without hyperthreading, so 4 threads total), 2400 Mhz (Or should I say 2400MT/s, anyway...) Cl11 2x8 RAM sticks and a 960 evo nvme drive, and the AMD driver seem to significantly reduce stuttering on Assetto Corsa Competizione (probably the extra gpu power and video memory helps too, as it usually consumes about 7G in the game, with mixed settings and 1440p internal resolution [133% resolution scale], with the GTX 960 I was using similar if not exact settings at internal 1080p resolution, and the video ram usage is not much lower like this, still way above 4G). My target is 60 fps as I'm aware of the CPU/system RAM limitations and my monitor (tv really) is 60hz. A few months back I upgraded from 2x8 1600cl10 as one of the sticks started failing, and that upgrade was noticeable, and this also was, probably more.

Unfortunately it seems the best would be to have a toggle in driver settings to have either Nvidia approach for older not very well threaded (and hopefully not very CPU demanding) games, or the less CPU consuming AMD approach, but I don't see that happening anytime soon. It may be placebo but I feel like I lost a bit of performance for instance in a very old Directx 9 game, NFS: MW from 2005, probably because here the Nvidia driver is capable of squeezing more draw calls out if it.

The performance loss in the lower end of the CPU spectrum due to the Nvidia drivers can be very noticeable. This is an example of this happening in Battlefield V, the frametime graph... :

Why on God's green Earth would you pair an ancient, underpowered CPU like that with a far faster GPU? That 6700 XT is probably spending 50% or more of its time idle, waiting for your CPU to catch up.

My current pairing of old PC is Phenom 1100T/Vega 56 Nano.

Because I'm a hardware modder, I can extract more performance than most users when it comes to old hardware. It's built to be extraordinary snappy.

Cool story bro!

 

[Vya Domus]

This is wild because it used to be that AMD had way bigger driver overhead I wonder how Nvidia screwed this up.

 

[Shou Miko]

Depending because in my Windows XP system I am using a GTX 750 Ti MSI TwinFrozr version (Only wish it had a DP port) with the latest supported driver for XP and it's basicly not doing much in all the games I game on my XP machine but it gives a good experience in both older and older older titles that I run compared to what AMD can have of issues from the time.

 

[ARF]

This is wild because it used to be that AMD had way bigger driver overhead I wonder how Nvidia screwed this up.

It's weird when at 4K the CPU sits almost idle at ~10% utilisation. Just waste of resources. In this sense the bigger the driver overhead, the better.

 

[gt362gamer]

Why on God's green Earth would you pair an ancient, underpowered CPU like that with a far faster GPU? That 6700 XT is probably spending 50% or more of its time idle, waiting for your CPU to catch up.

I couldn't afford a whole PC upgrade, and having seen this driver performance gap when CPU limited plus the fact that I wasn't that far from reaching fixed 60 fps on the most CPU demanding game I was playing made me just buy it second hand with warranty, and the performance difference is noticeable. In top of that, now I should have enough GPU power for a good amount of time, so I'm glad I buy it.

One thing that's also noticeable is that, when I tested uncapping the fps, the game didn't suddenly turn into an unplayable stutter mess, something that I would expect with the Nvidia driver and the GTX 960, it runs to about 75-90 fps when I'm alone or not much cars around me, and about 60-70 fps when I'm surrounded by several other cars. It's not a very even experience but the fact it wasn't straight up unplayable was a pleasant surprise. IIRC I tested this uncap setting with the time progression in the game at 1x (real time progression) and Zandvoort with 24 AI cars in a dry race, which isn't the most demanding scenario that can happen in the game, but it still was surprising.

And I tried to disable the thread optimization with the GTX 960 too, @Dr. Dro . It seemed to me the performance was a little bit improved, but I wasn't sure, maybe it was placebo effect. Still it's a setting Nvidia users should try if they're CPU limited in a certain scenario. I could have tested it further but I guess I wanted to upgrade anyway.

 

[ViperXZ]

Generally:

Radeon has less driver overhead because DX12 / Vulkan (low level APIs) are directly calculated in the hardware, more so than with Nvidia who have a more software approach to it.

This means weaker CPUs run DX12 / VK games better on Radeon. But generally DX12 / VK will work better for weaker CPUs, as those APIs are designed to reduce driver overhead or eliminate it altogether.

 

[A Computer Guy]

This is wild because it used to be that AMD had way bigger driver overhead I wonder how Nvidia screwed this up.

Perhaps it's by design since you can throw a little more power at the Intel CPU and it won't matter?

 

[gt362gamer]

Perhaps it's by design since you can throw a little more power at the Intel CPU and it won't matter?

The thing is, it seems that this performance delta isn't only showing in low end CPUs being maxed out (or very close to) on all threads, as I've came across a video in which an RX 480 is seen beating the RTX 3060 in average and min* framerate in Hogwarts Legacy and in The Witcher 3 remaster (in the latter case also in the 1% lows) with a Ryzen 5 3600x:

*PD: I still find weird that min figure > 1% low, maybe low average, sustained min or something like that would be a better tag to define what it measures, because when I read a minimum figure I guess I expect the absolute lowest frametime translated into a second in frames.

 

[Ruru]

Damn, I totally forgot this fact. I have a XP rig with Phenom 9750 and GTX 660 3GB, I have a spare HD 7970 which would be a better choice I guess.

 

[R-T-B]

This is wild because it used to be that AMD had way bigger driver overhead I wonder how Nvidia screwed this up.

They didn't. AMD just really got it's act together only recently, but they did so with such advancements that it left nvidia behind.

Nvidia is still as good as its always been, overhead wise. But AMD is now slightly better.

 

[ViperXZ]

They didn't. AMD just really got it's act together only recently, but they did so with such advancements that it left nvidia behind.

Nvidia is still as good as its always been, overhead wise. But AMD is now slightly better.

with dx11 etc they did, with low level apis they already had. They basically introduced that stuff with mantle. dx11 etc is much more driver dependant so it took them more time.

 

[Dr. Dro]

This is wild because it used to be that AMD had way bigger driver overhead I wonder how Nvidia screwed this up.

It isn't a screw-up, it's a tradeoff that Nvidia opted for, (wisely) deducing that in the future, CPUs with an ample amount of threads and high clock frequencies would be available. This means that DirectX 11 games are all but guaranteed to scale better on Nvidia hardware as long as there is CPU power for the driver to batch and defer commands to the GPU.

On AMD, the driver interfaces directly with the GPU, which has a dedicated command scheduler. The new DirectX and OpenGL on PAL drivers that were released are highly optimized, but the same limitations apply: single-thread immediate context only; maximum amount of draw-calls before the frame rate tanks highly dependent on the GPU's command processor, etc.

with dx11 etc they did, with low level apis they already had. They basically introduced that stuff with mantle. dx11 etc is much more driver dependant so it took them more time.

AMD still doesn't and likely will never support the features I have mentioned in my post above. Check the github link on my reply to Vya, you'll be able to grasp a little about what is making AMD's newer driver releases tick. This is what's important, anyway: