Far Cry 6 GPU performance not bad at all but is severely bottlenecked by CPU

[mmj_uk]

Can anyone post screenshots comparing standard textures to the texture pack? interested to see how big the difference is.

 

[Ravenger]

So is this one game where i can say im glad i have a 3090 with 24gb vram? The only time its really relevant in gaming lol.

I'm glad I got a 3090FE too. Way too expensive, but better to have the extra RAM, as it's not upgradable. I can add more memory to my PC easily and relatively cheaply, but that's not possible on a graphics card.

 

[LtMatt]

Here's a new one, this 3080 user is getting warnings about low video memory usage in the built in benchmark from what looks like Ubisoft overlay. Not seen that before.


He's running 4K, max settings but motion blur off and he's using FSR Ultra Quality.

Not sure why his minimum FPS is so low, 38 min and 63 average.

6900 XT gets a higher minimum FPS (66) at the same settings than his average FPS.

Spoiler

It's almost like he's running low on video memory, according to the message.

 

[Nexus18]

Here's a new one, this 3080 user is getting warnings about low video memory usage in the built in benchmark from what looks like Ubisoft overlay. Not seen that before.

He's running 4K, max settings but motion blur off and he's using FSR Ultra Quality.

Not sure why his minimum FPS is so low, 38 min and 63 average.

6900 XT gets a higher minimum FPS (66) at the same settings than his average.

Spoiler

It's almost like he's running out of video memory.

Yet:

- his vram is not going above 9GB vram
- cpu usage is very low
- gpu usage is all over the place
- doesn't seem to match 5/6+ other sites performance nor other users experience so far

EDIT:

Also, further back in the thread, weren't we saying not to use the in game benchmark because it doesn't reflect actual in game experience?

 

[Bill Turnip]

Maybe the vRAM will be an issue, but there are updates to come for this yet I would have thought. Even if after updates and everything it is absolutely shown to be the vRAM which is the limitation, and that is the worst that happens, it's not exactly a disaster rendering the game unplayable.

I'm of the belief that 10GB will be enough, but not so much that I'm going to sit on a hill and hold a position forever - If it's proven not to be "enough" but that video is what happens? Fine.

I would ask why in the settings screen at the bottom, it suggests that it will need only 6.82gb of vRAM though, unless I'm missing something obvious. If it thinks it will need 7GB, and in actuality try to use over 10GB, it does seem that there are significant improvements to be made down the line.

 

[mmj_uk]

The only time the frame-rate dips down below 50fps is at the end of the benchmark before the results screen.

 

[LtMatt]

Can anyone post screenshots comparing standard textures to the texture pack? interested to see how big the difference is.

There are in depth reviews of the texture pack image quality and its affect on various GPUs here.

The only time the frame-rate dips down below 50fps is at the end of the benchmark before the results screen.

Yes, right at the end of the benchmark the FPS does drop to be fair, even I saw that as my minumum FPS was in the 70s until the end.

Maybe the vRAM will be an issue, but there are updates to come for this yet I would have thought. Even if after updates and everything it is absolutely shown to be the vRAM which is the limitation, and that is the worst that happens, it's not exactly a disaster rendering the game unplayable.

I'm of the belief that 10GB will be enough, but not so much that I'm going to sit on a hill and hold a position forever - If it's proven not to be "enough" but that video is what happens? Fine.

I would ask why in the settings screen at the bottom, it suggests that it will need only 6.82gb of vRAM though, unless I'm missing something obvious. If it thinks it will need 7GB, and in actuality try to use over 10GB, it does seem that there are significant improvements to be made down the line.

Fair and agreed, but some folks can't accept that I am not really sure why.

It is enough for most games, there's only a few where it's an issue.

The other games can be debated since in depth testing has not been done, but there's no getting away from it in Far Cry 6.

Spoiler

That's been confirmed by Computerbase, PCGamesHardware and DSO now. It really comes down to the settings used and the areas tested. Sure you can reference a 30 second benchmark sequence from HUB or techpowerup and say there's no issue, but that's not really giving the full picture.

 

[mmj_uk]

Yes, right at the end of the benchmark the FPS does drop to be fair, even I saw that as my minumum FPS was in the 70s until the end.

but why would a transition screen affect the benchmark and how is it that the result graph shows that the frame-rate never dips below 55fps and yet it says 38fps minimum?

 

[LtMatt]

but why would a transition screen affect the benchmark and how is it that the result graph shows that the frame-rate never dips below 55fps and yet it says 38fps minimum?

No idea, the benchmark sequence is pretty useless other than for quickly judging raw graphics performance.

It's something like 30 seconds from start to finish so not much effort went into it by the looks of it.

Most of the tech sites will do testing over several areas of actual gameplay for periods of time longer than 30 seconds, see the links provided above.

 

[stooeh]

Is that from your testing? I see you have 5800x and a 3080.

Have you seen any fps drops to 5fps as per 2 other sites benchmark? and how long have you played for? And if it does happen, does taking a screenshot bring the fps back up?

I didn't but I was swapping between FSR on/off. I did quickly start a new game and while legging it through the intro I did find a spot where the FPS tanked with per app VRAM showing at 9431MB!. At the ladders that take you back out of the sewers if anyone wants to check the spot out

 

[shankly1985]

Can anyone post screenshots comparing standard textures to the texture pack? interested to see how big the difference is.

Image won't always show much difference for texture. For ultra high texture the higher the resolution you running the game will have the greater impact on quality.

 

[Bill Turnip]

It is enough for most games, there's only a few where it's an issue.

Which other games is it an issue? Just catching myself up on this. And also, why does the settings screen only estimate 7gb of vRAM usage when that is apparently not the case?

 

[dualsense1673]

ltmatt 5600x seems to be huge bottleneck for 3080 in that video. min. fps is also probably affected by cpu tbh. cpu bottlenecked games usually run at horrendous frametimes. best framepacing is usually at when you have perfect %99 gpu usage across all times

6900xt high %1 lows are with 5950x

its known that 5950x have way more cache than 5600x and it is also known that that large cache increases single thread cpu bound performance in games greatly (there are some odd benchmarks i've seen before. fc6 probably likes the extra cache 5950x provides)

game is clearly single thread bound, 5600x is no slouch i know, but 5950x cache is massive. any singular core can tap into that large cache and do not even touch dram by large amounts of time.and this makes 5950x way faster than other lower cache zen 3 members in certain situations. just a theory of course, i dont mean anything by this

 

[JediFragger]

He's running 4K, max settings but motion blur off and he's using FSR Ultra Quality.

I'll try this later using 3840x1600 (AW3821DW) and see if it gives me a warning.

 

[LtMatt]

I didn't but I was swapping between FSR on/off. I did quickly start a new game and while legging it through the intro I did find a spot where the FPS tanked with per app VRAM showing at 9431MB!. At the ladders that take you back out of the sewers if anyone wants to check the spot out

Appreciate that Stooeh.

Image won't always show much difference for texture. For ultra high texture the higher the resolution you running the game will have the greater impact on quality.

Yes, HD Texture pack shines at 4K.

Which other games is it an issue? Just catching myself up on this. And also, why does the settings screen only estimate 7gb of vRAM usage when that is apparently not the case?

Godfall is another one that I have personally seen results of very low FPS with 'certain' GPUs.

No tech sites have covered it in depth though, game was just not well received so nobody cared to look into it properly.

I think the in game estimated video memory usage is wrong, because I use significantly more and it contradicts with Ubisofts own recommendations.

ltmatt 5600x seems to be huge bottleneck for 3080 in that video. min. fps is also probably affected by cpu tbh. cpu bottlenecked games usually run at horrendous frametimes. best framepacing is usually at when you have perfect %99 gpu usage across all times

6900xt high %1 lows are with 5950x

its known that 5950x have way more cache than 5600x and it is also known that that large cache increases single thread cpu bound performance in games greatly (there are some odd benchmarks i've seen before. fc6 probably likes the extra cache 5950x provides)

game is clearly single thread bound, 5600x is no slouch i know, but 5950x cache is massive. any singular core can tap into that large cache and do not even touch dram by large amounts of time.and this makes 5950x way faster than other lower cache zen 3 members in certain situations. just a theory of course, i dont mean anything by this

5600X should not bottleneck a 3080 tbh, but I won't argue with you on this point as the built in benchmark is not ideal for real world performance IMO regardless and the 5950X is a bit of a beast when paired with a RX 6000 series GPU with Smart Access Memory. It really helps the minimum FPS at times.

But I did think it was worth mentioning since the game was informing the user that his graphics card is running low on available video memory and performance may be impacted.

I'll try this later using 3840x1600 (AW3821DW) and see if it gives me a warning.

You should be fine Si with 12GB tbh, but see how it goes when you spend a few hours in the open world and let us know.

 

[Bill Turnip]

I'll try this later using 3840x1600 (AW3821DW) and see if it gives me a warning.

Same screen as me. Feel a 3080 matches perfectly, don't expect any issues with Far Cry 6 when I get round to it (Not paying full price for it).

Also connected to 4k TV downstairs, so will be interesting to see how they compare.

 

[stooeh]

Curious if anyone has a 10900K or similar to see how the CPU bottlenecks compare.

 

[dualsense1673]

Appreciate that Stooeh.

Yes, HD Texture pack shines at 4K.

Godfall is another one that I have personally seen results of very low FPS with 'certain' GPUs.

No tech sites have covered it in depth though, game was just not well received so nobody cared to look into it properly.

5600X should not bottleneck a 3080 tbh, but I won't argue with you on this point as the built in benchmark is not ideal for real world performance IMO regardless and the 5950X is a bit of a beast.

But I did think it was worth mentioning since the game was informing the user that his graphics card is running low on available video memory and performance may be impacted.


You should be fine Si with 12GB tbh, but see how it goes when you spend a few hours in the open world and let us know.

Oooh but now when you think of it, we also have to consider nvidia taking up more cpu performance, right? maybe you can take a look at, for example, 5600x+3090 benchmarks. if it still pushes low %1 lows, maybe we can relate the issue to nvidia driver being buggy and wonky =))


https://youtu.be/P0VIJaByg2k?t=111

this one for example, min. fps seems too low

i think i was wrong yeah. but maybe i'm onto something. after all, that driver constantly doing extra work behind the scenes must affect some games somehow. but again, as u said, its a theory. but low min. fps is also observable on 3090. it seems amd's hardware native scheduler that does not rely on CPU to assign workloads to GPU compute cores does wonders here

 

[LtMatt]

Oooh but now when you think of it, we also have to consider nvidia taking up more cpu performance, right? maybe you can take a look at, for example, 5600x+3090 benchmarks. if it still pushes low %1 lows, maybe we can relate the issue to nvidia driver being buggy and wonky =))


https://youtu.be/P0VIJaByg2k?t=111

this one for example, min. fps seems too low

i think i was wrong yeah. but maybe i'm onto something. after all, that driver constantly doing extra work behind the scenes must affect some games somehow. but again, as u said, its a theory. but low min. fps is also observable on 3090. it seems amd's hardware native scheduler that does not rely on CPU to assign workloads to GPU compute cores does wonders here

Yes you make a good point. I did not consider CPU overheard when I was comparing minimum FPS so that probably explains it.

 

[LtMatt]

Oooh but now when you think of it, we also have to consider nvidia taking up more cpu performance, right? maybe you can take a look at, for example, 5600x+3090 benchmarks. if it still pushes low %1 lows, maybe we can relate the issue to nvidia driver being buggy and wonky =))


https://youtu.be/P0VIJaByg2k?t=111

this one for example, min. fps seems too low

i think i was wrong yeah. but maybe i'm onto something. after all, that driver constantly doing extra work behind the scenes must affect some games somehow. but again, as u said, its a theory. but low min. fps is also observable on 3090. it seems amd's hardware native scheduler that does not rely on CPU to assign workloads to GPU compute cores does wonders here

Yes you make a good point. I did not consider CPU overheard when I was comparing minimum FPS so that probably explains it.

I think you hit the nail on the head. Look at the comment in the video you posted, user with 3090 and 5950X only getting 73 FPS average at 1440P.

Spoiler

Yet I get 107 average FPS at 1440P, max settings same CPU but different GPU, so there has to be some CPU driver overhead there surely.

The benchmark is okay but should not be used as a true measure of actual performance in my opinion unless you run at 4K max settings.