Test: CPU Core count and RAM scaling in ACC, AC1 and R3E

[RobertR1]

So I finally got a working AIO and did a back n forth in realbench at a setting that I know is close to throttling my air cooler.

Settings:
9900k
HT OFF
5.2ghz
1.305v in bios
z390 aorus pro
Turbo LLC
46x core
Case = haf x with side panel off

Dark Rock 4 results:

Alphacool 360LT results:

Summary: get a good AIO or custom loop if you want to push a thermally dense chip.

@Dan Costa also learned that I have a massive voltage wall between 5.3 and 5.4. I'm now running 5.3 at 1.345v in bios with llc turbo. Even going to 1.42v for 5.4, I can't even pass R15. GG

 

[Kek700]

Trying to run my newly acquired 9600k proved, even at 5Ghz all cores to be very problematic.
To cut a very long story short, I managed to get core voltage to 1.25v at 5Ghz, sufficient for my needs. But had big problems with over heating, just could not keep the chip in a reasonable range.
Definitely not with CPU test software, not in AC either, admittedly using a cheapish 120mm Aio.
Eventually decided to go for custom water cooling loop, that is still 120mm rad, but now I am in the sub 70deg C territory, and in the 50 deg in AC.
A simple summary of what was a lot of effort to try and get things to work, my 6600k processor certainly had so many good points in this way, even considering it is 3 generations away, it still is not relatively that far behind. Considering I ran it at 1.4v all cores 4.7Ghz with no temperature issues on a cheapish AIO.

 

[RobertR1]

Trying to run my newly acquired 9600k proved, even at 5Ghz all cores to be very problematic.
To cut a very long story short, I managed to get core voltage to 1.25v at 5Ghz, sufficient for my needs. But had big problems with over heating, just could not keep the chip in a reasonable range.
Definitely not with CPU test software, not in AC either, admittedly using a cheapish 120mm Aio.
Eventually decided to go for custom water cooling loop, that is still 120mm rad, but now I am in the sub 70deg C territory, and in the 50 deg in AC.
A simple summary of what was a lot of effort to try and get things to work, my 6600k processor certainly had so many good points in this way, even considering it is 3 generations away, it still is not relatively that far behind. Considering I ran it at 1.4v all cores 4.7Ghz with no temperature issues on a cheapish AIO.

9xxx runs hot. With a better rad, you can probably get 5.2ish which would help in low fps and VR but if you're not on VR you won't notice a difference.

 

[Kek700]

It is probably a silly statement, but some times it
seems at least in my system as it was, heat generates more heat, once the temperatures started to get into the 80 + range at that point the temperatures rose easily into the 90 + . Once I had got the temperatures to the mid 60’s. With a good quality custom loop, same size rad.
It just seemed to contain it all and temps would stabilise at 65 deg C and stay there.
As you say there is some advantage for me to clock higher, but once AC starts to demand a lot from my i5 there is very little I can do to notably offset the
situation.
I only have another 100 in the CPU, and maybe 3700 to 4000 in my ram, I at the moment cannot see the point in stressing it all, in probably a, in-vain attempt to put it all at a perfectly steady 120 hz.
I do have G-Sync on my side anyway, so that gives me a fall back.

 

[RichardHessels]

If you would do the same tests on resolutions we all do here 1440p, extra wide, triple screens, 4k or VR you will see even less differences in the result. As the videocard is the bottleneck at anything higher than full-HD.

 

[Deleted member 197115]

If you would do the same tests on resolutions we all do here 1440p, extra wide, triple screens, 4k or VR you will see even less differences in the result. As the videocard is the bottleneck at anything higher than full-HD.

Flat screen, may be, VR is very different.

 

[Dan Costa]

It is probably a silly statement, but some times it
seems at least in my system as it was, heat generates more heat, once the temperatures started to get into the 80 + range at that point the temperatures rose easily into the 90 +

Its not silly, every cooling device has a tipping point after which the temperatures start to rise very quickly.

 

[Terry Rock]

Some things seem to have become the norm in the simracing computer world.
All Intel CPUs must maintain above 5 GHz core clock...on all cores...all the time.
VR must run at 90 fps with no dips below.
Monitors must be 144 or greater frequency.
Why is everybody so caught up in having Intel CPUs running above 5 GHz?
Most probably wouldn't be able to tell the difference if you ran a much cooler 4.7 or 4.8 GHz.
It'd probably be more stable as well.

 

[RobertR1]

Some things seem to have become the norm in the simracing computer world.
All Intel CPUs must maintain above 5 GHz core clock...on all cores...all the time.
VR must run at 90 fps with no dips below.
Monitors must be 144 or greater frequency.
Why is everybody so caught up in having Intel CPUs running above 5 GHz?
You probably wouldn't be able to tell the difference if you ran a stable and much cooler 4.7 or 4.8 GHz.
It'd probably be stable as well.

Why mod your car for another 30hp if you have 500 already? People enjoy tinkering and squeezing every little bit out of it.

RAM tuning will do wonders for sim racing but requires a bit of time and effort. CPU overclock are more common and easier to compare against so it becomes a bit of a side competition for many.

Also the higher clocks mean better low FPS which helps in vr to maintain better smoothness and less dips into reprojection as often, if you’re straddling the edge of native FPS already.

 

[Terry Rock]

Why mod your car for another 30hp if you have 500 already? People enjoy tinkering and squeezing every little bit out of it.

RAM tuning will do wonders for sim racing but requires a bit of time and effort. CPU overclock are more common and easier to compare against so it becomes a bit of a side competition for many.

Also the higher clocks mean better low FPS which helps in vr to maintain better smoothness and less dips into reprojection as often, if you’re straddling the edge of native FPS already.

I get the reference...I get overclocking.
My point is this...in continuing your example....
Modding the engine to that initial 500 hp is great. At that point, It is way more than stock and still reliable.
If that extra 30 hp is the difference between cruising the 'Tail of the Dragon' on the weekend or fixing a connecting rod out the side of the block, guess which one most people should take?

 

[RasmusP]

It's more like you would have 485 hp and most of your tuning buddies are having 500-520.
You just want to crush this 500 mark

But that's not the reason. The reason is that there's always a plateau when overclocking where you need a lot higher voltage for a very little increase in clockspeed.

My old i7 2600k has this plateau at 4.4 GHz and it's happily running there since years.

Thing is, most 8xxx and 9xxx Intels have this plateau at or beyond 5 GHz, only issue being the cooling.

About the difference between 4.7 and 5.0:
Most people I know become dizzy when they play with reprojection so the 90 fps are very important to them.
But even at 5.3 GHz on your 9900k you just need to add a few AI or not optimized mod cars on the grid and that was it with the constant 90 fps...

So squeezing out the maximum is important for them.

Btw, I'm running limited 60 fps in all games with gsync so no 144 Hz for me.
But I have 100 Hz on desktop and it's a massive difference in fluidity. Don't really feel a difference when going for 120 Hz though...

 

[RobertR1]

I get the reference...I get overclocking.
My point is this...in continuing your example....
Modding the engine to that initial 500 hp is great. At that point, It is way more than stock and still reliable.
If that extra 30 hp is the difference between cruising the 'Tail of the Dragon' on the weekend or fixing a connecting rod out the side of the block, guess which one most people should take?

Here's a recent test that goes over the potential differences. This is a 9900k at 5.0ghz and 5.2ghz

https://www.tomshardware.com/uk/features/intel-special-edition-core-i9-9900ks-benchmarked

The piece to focus on are the minimum fps which as @RasmusP stated in his reply can make a difference in the VR experience. Dropping below the desired FPS creates stuttering/choppiness/re projection which you want to avoid as it's more jarring in vr than on a flat screen. Thus if the difference between 5.0 and 5.2 is an extra 10fps on the low end, I'll take that all day in VR.

If you're on a flat screen with gysnc/freesync then you can handle frame rate fluctuations a lot better so the value can be trivial.

 

[Deleted member 197115]

Another thing, frame time is a compound of CPU and GPU time, the less time CPU spends preparing draw calls, the more left for GPU, which means room for more SS and higher graphics settings.

 

[Durge Driven]

Who is going to buy a topend CPU and 208Ti to run PC2 or any other sim @ 1080p
I understand why they limit res in these kind of tests but it's not real world higher the Res less gap right ?
Other game tests there were closer 3900X@ PB faster then stock 9900K in a couple

KS is 110AU dearer and the 3900X cooler I could sell for $40AU easy ...so $150 less which is exactly 20% ergo I should expect 20% less performance but it's not quite that much

 

[Durge Driven]

Wow still over 60% in Steam use 1080p and i5 and GTX1060 is average hardware ...never realized

 

[RobertR1]

Wow still over 60% in Steam use 1080p and i5 and GTX1060 is average hardware ...never realized

A lot of fps genre gamers prefer high hz at 1080p. 240hz and such are heavily cpu dependent.

 

[2stains]

I have a 1080ti/7700k for 1080p but x3 for triples and it's still barely enough for 60 fps in ACC

 

[RasmusP]

Who is going to buy a topend CPU and 208Ti to run PC2 or any other sim @ 1080p
I understand why they limit res in these kind of tests but it's not real world higher the Res less gap right ?
Other game tests there were closer 3900X@ PB faster then stock 9900K in a couple

KS is 110AU dearer and the 3900X cooler I could sell for $40AU easy ...so $150 less which is exactly 20% ergo I should expect 20% less performance but it's not quite that much

It's also very interesting to see what a CPU can run at in theory as swapping out the CPU is mostly a case of buying a new motherboard, new ram, installing a fresh windows, reinstalling all programs etc.
Swapping out a gpu every 2-3 years is a 5 minute job nowadays.

That's why I'm still on my i7 2600k I bought in 2011 but I already had a:
560 ti, 760 (warranty replacement for the 560), 970 and now 1070.
And I'm planning to buy a 3070 since I upgraded from 27" 1080p, 60 Hz to 3440x1440, 120 Hz.

But now the CPU is also too weak to run higher fps than 50-60 (thanks gsync, I love you, lol).
So I'll upgrade the CPU too.

But that CPU will have to run 90-120 fps (+ vr maybe) for at least 5 years to be a fitting investment for me.
Probably outliving a 5070 gpu and a 7070 gpu too.

For people like me, who love to have everything running fluently but aren't willing to swap their CPUs as frequently as GPUs, knowing the true limit of the CPUs is very important to be future proof.

Same as it's important to see how a rtx 2060 runs at 8k in comparison to a 2070 for example.

So people who will run some less extensive games on their future TVs know what to buy. Invest 100€ now to be able to run 8k okayish or not etc.

 

[Kek700]

Eventually became very pleased with my i5 9600k
5 GHz all 6 cores for £200,
Mother bd to run said, £200
Cooling, it was my main problem, it drives ACC at 90 FPS + with full field online race, 3440x1440 monitor, until stuttering took hold.

That reminds me,
must repeat looking at my temps.

 

[dud]

Is there any program for windows that displays CPU usage like top(1) in Unix does, which is 100% = one core pegged?

I want to see how many times I have single core doing a single thread workload being a holdup.