• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

CPU IPC Gaming Performance over the generations

Caporegime
Joined
17 Mar 2012
Posts
50,004
Location
ARC-L1, Stanton System
An actually informative video from HUB.

All CPU's clocked to 4.2Ghz and 4 cores 8 thread.

A couple of interesting things, for me anyway... the 2600K and 3770K are almost identical, i thought this might be true given that the 3770K is the same CPU but on 22nm vs 32nm for the 2600K.

The 4770K is significantly better than the 3770K, i was surprised at this as i didn't think that again there was any difference between them, well... i was wrong about that.

The 3800X is actually very similar to the 9900K, that is clock for clock core for core, so very similar IPC in games.
I'm a little bit surprised about this, i thought Coffee Lake was still a bit stronger in gaming IPC but it seems not, its just the clock speed difference.

What's not at all surprising is the FX series from AMD, that thing is just plain bad.

I think some of these slides are still far more GPU limited than CPU, which is typical HUB but it is what it is and there is some interesting stuff about it.

A couple of highlights.

This one i think is a GPU limitation, if you look at the slide below it the 10900K is 15% faster than the 7700K so i think you can add that to the 130 it scores here, so 150 FPS. a solid win for the 10900K here, i think with the GPU bottleneck removed the 5800X would have got to about 130 FPS.

O4KL1TI.png


Zen 3 here kicking arse and taking names

9w5Lzfj.png


All put together hmmm.... i think with half of these slides being GPU limited a slide like this is utterly useless but here it is.

wPoevJ5.png


 
They could have done more work and run all those with Ampere 3090 too.

Would have been interesting how the older CPUs cope with feeding Ampere.

BTW: IB is not just a shrink of SB from 32nm to 22nm. Although both did - almost unbelievably - use the same socket.
 
They could have done more work and run all those with Ampere 3090 too.

Would have been interesting how the older CPUs cope with feeding Ampere.

BTW: IB is not just a shrink of SB from 32nm to 22nm. Although both did - almost unbelievably - use the same socket.

This is a really good point, i have no doubt using the 3090 would have resulted in quite different outcomes, Nvidia rely heavily on the CPU for scheduling, it makes quite a lot of difference so with just 4 cores 8 threads it would have push these CPU's a lot harder, using the 6900XT instead of the 3090 was probably quite deliberate.
 
Should have used a 3090 because Nvidia needs more cpu power than amd does because of its driver scheduling overhead
 
Real world FS2020 with 3080 ... my old 6700k wasn't hacking it. Still a great cpu though.

I found the performance in gaming with the 6700K wasn’t half bad with a Radeon RDNA card. AMD graphics cards seem to leverage a lot more performance from that system.
 
I found the performance in gaming with the 6700K wasn’t half bad with a Radeon RDNA card. AMD graphics cards seem to leverage a lot more performance from that system.
I just think it was a combo of bottlenecking/ lack of cores for FS2020 CP2077 etc. Surely it would bottleneck a 6800/6900XT though?
 
If the current Nvidia 3000 cards are more bottlenecked than AMD 6000 cards does this mean in a year or two time with the newer more powerful CPUs we'll see them pull away from AMD in performance?
 
If the current Nvidia 3000 cards are more bottlenecked than AMD 6000 cards does this mean in a year or two time with the newer more powerful CPUs we'll see them pull away from AMD in performance?
Not sure, but I don't think current 6+ core AMD/Intel will bottleneck 3000's .... just the old stuff.
 
Not sure, but I don't think current 6+ core AMD/Intel will bottleneck 3000's .... just the old stuff.

Depends on the game. There are quite a lot of Youtube Vidoes of people with 8600K's with horrendous micro stuttering in BFV using RTX 2000 series.

Adding HT to the CPU, so 8700K cures that issue but we are talking about RTX 2000 series and a 2 year old game at this point.

Games are making ever more demands of the CPU, So CPU's have to keep getting ever faster especially if you're running an Nvidia GPU because unfortunately they will chew up a good chunk of your CPU just for its self.

Bottlenecked at 1080P on an RTX 3090 and these are not high frame rates.

0eTJlne.png
 
Ah.... that makes sense. ta :)

I think these extensions are linked to dragon ball z. E.g:

MMX = kaio ken
MMX2 = kaio ken x10?
SSE = super saiyan
SSE2 = super saiyan 2
SSE3 = super saiyan 3
SSE4 = super saiyan 4
AVX = ssj god (red)
AVX2 = ssj god 2 (blue)
AVX512 = ssj god super saiyan evolution (vegeta)

And im sure AMD used FUSION which was also used in dbz. I think there is enough of a coincidence there to suspect something going on.
Still waiting for the ultra instinct instruction set. :D
 
I think these extensions are linked to dragon ball z. E.g:

MMX = kaio ken
MMX2 = kaio ken x10?
SSE = super saiyan
SSE2 = super saiyan 2
SSE3 = super saiyan 3
SSE4 = super saiyan 4
AVX = ssj god (red)
AVX2 = ssj god 2 (blue)
AVX512 = ssj god super saiyan evolution (vegeta)

And im sure AMD used FUSION which was also used in dbz. I think there is enough of a coincidence there to suspect something going on.
Still waiting for the ultra instinct instruction set. :D

I'm sure that's really funny but its gone over my head, probably because i don't play Dragon Ball Z.
 
It doesn't seem like 2 minutes ago that AMD fans were arguing that multi-threading was the future and IPC wouldn't matter. Bulldozer architecture was so poor, they should have just kept building on Phenom II X6.
 
It doesn't seem like 2 minutes ago that AMD fans were arguing that multi-threading was the future and IPC wouldn't matter. Bulldozer architecture was so poor, they should have just kept building on Phenom II X6.

Well, its both, not one or the other.
 
An actually informative video from HUB.

All CPU's clocked to 4.2Ghz and 4 cores 8 thread.

A couple of interesting things, for me anyway... the 2600K and 3770K are almost identical, i thought this might be true given that the 3770K is the same CPU but on 22nm vs 32nm for the 2600K.

The 4770K is significantly better than the 3770K, i was surprised at this as i didn't think that again there was any difference between them, well... i was wrong about that.

The 3800X is actually very similar to the 9900K, that is clock for clock core for core, so very similar IPC in games.
I'm a little bit surprised about this, i thought Coffee Lake was still a bit stronger in gaming IPC but it seems not, its just the clock speed difference.

What's not at all surprising is the FX series from AMD, that thing is just plain bad.

I think some of these slides are still far more GPU limited than CPU, which is typical HUB but it is what it is and there is some interesting stuff about it.

A couple of highlights.

This one i think is a GPU limitation, if you look at the slide below it the 10900K is 15% faster than the 7700K so i think you can add that to the 130 it scores here, so 150 FPS. a solid win for the 10900K here, i think with the GPU bottleneck removed the 5800X would have got to about 130 FPS.

O4KL1TI.png


Zen 3 here kicking arse and taking names

9w5Lzfj.png


All put together hmmm.... i think with half of these slides being GPU limited a slide like this is utterly useless but here it is.

wPoevJ5.png




Except that you don't measure IPC using a game or just games
So the headline is misleading
 
Back
Top Bottom