CPU upgrade needed!

[RasmusP]

CPU has returned to 2 - 5% normal idle. Whatever it was up to for 10 minutes on launch i don't know.

ok figured it out. you can change the threads in 'preferences'

Not sure what the 3 vs 4 thread comparison tells me?

3 and 4 threads in cinebench simulate the load of ac/acc/rf2/raceroom quite well.
1 thread would mean that the cpu boosts one core to the maximum but when you play one of these Sims, the boost won't be as high (if you don't lock all cores manually ofc).
So Single thread Benchmarks distort the true gaming performance.

Full multicore Benchmarks obviously give better results with more and more cores.
But more than 5 cores give you absolutely nothing for sim fps.

Are your 553 points for 3 or for 4 threads?

Here are some numbers from my excel sheet:
10600k @ 4.9 GHz = 628
3600 "oc" = 606
5600x stock = 743
5900x stock = 796

From early benchmarks, the 11900k might have 805 points. But that's some guesstimation

Now with some simple rule of three you could calculate the fps with every cpu. You just need to know the fps for one of the CPUs. Then you know what all the others would be able to push out

EDIT:
I actually dig out my sheet and put in your 7600k. Here's how it looks:

CPU/Score	10600k 4.9	7600k 4.5	3600 OC	5600x stock	5800x stock	5800x 2nd	5900x stock
3 threads	628​	553​	606​	743​	724​	742​	796​
4 threads	821​	719​	807	961​	921​	973​	1030​
fps 3 threads	90,0​	79,3​	86,8​	106,5​	103,8​	106,3​	114,1​
fps 4 threads	90,0​	78,8​	88,5​	105,3​	101,0​	106,7​	112,9​

For the 11900k I searched for Cinebench R20 scores but that benchmark uses AVX CPU instructions so it's not really comparable to simracing.
I did some wild calculations from values I've found and in the end:
It's slightly better than the 5900x in single threading. The 11600k will be a bit worse and based on past generations, I will probably gain about 19% compared to my 10600k in fps.

In the table above that would mean:
10600k: 90 fps ; 11600k = 107,1 fps

Not really worse the money but if you need the magic 90 fps for VR, it's worse everything!

 

[Interslice]

As i was saying earlier I was only getting 65 fps during the race last night.
30ish online at monza.

On the plus side i think i've sorted my video editing issues just by setting windows update to only run during the night. I can finally make decent quality again from the replay files.

 

[RasmusP]

As i was saying earlier I was only getting 65 fps during the race last night.
30ish online at monza.

On the plus side i think i've sorted my video editing issues just by setting windows update to only run during the night. I can finally make decent quality again from the replay files.

65 fps seems a bit low when compared to the table.
Can you run the replay in vr and check your fps from start a second before the start for one minute?
I'll do the same (from your cockpit).

Maybe there are some settings that would like get optimized!

Can you share your csp settings? Afaik you can save a preset and share it!

Only if you're interest! Great that the recording is working again.

We could also read out voltages, temperatures etc and then try to optimize the overclock manually. Depending on how well the cooler works, we might get 4.7-5.0 GHz out of it.

 

[pattikins]

I was getting 65 fps on my oddysey + last night according to the last session fps thing in CM. I was ok to use but you would definetely notice the difference in the smoother 90fps i was getting in practice server. No idea if it was inserting warp frames or if it was a true 65 fps..

Your Odyssey+ has two settings - 90hz & 60hz* - they can be selected in Windows Headset Display settings. The reported 65fps by CM was most probably an average, so there was a good chance that you had a number of frames below that figure (no 'warp frames'). As Rasmus argues, it's a bit low. Did the low framerate effect your race pace?

* I did a few 60hz & 90hz tests back in September using a Steam program called fpsvr. Switching to 60hz had a significant positive impact on the cpu usage, but I ultimately preferred 90hz - so I had to put up with the frame spikes, stuttering etc.,.

GT4 @ Kyalami - Wednesday 23rd September 2020 Race 1 (16 cars) Odyssey @ 90hz

GT4 @ Kyalami - Wednesday 23rd September 2020 Race 2 (14 cars) Odyssey @ 60hz

 

[Interslice]

65 fps seems a bit low when compared to the table.
Can you run the replay in vr and check your fps from start a second before the start for one minute?
I'll do the same (from your cockpit).

Maybe there are some settings that would like get optimized!

Can you share your csp settings? Afaik you can save a preset and share it!

Only if you're interest! Great that the recording is working again.

We could also read out voltages, temperatures etc and then try to optimize the overclock manually. Depending on how well the cooler works, we might get 4.7-5.0 GHz out of it.

Content Manager

acstuff.ru

I think these are my settings using the share function
denied!

see attached CSP settings

 

[RasmusP]

https://**********/s/302bAF

I think these are my settings using the share function

Link is snipped... Can you put it in a link box?

 

[Interslice]

The replay in VR is 87 - 89 fps... Not sure where it got 65 fps from. Maybe from sitting in the pits after the race - that might have been counted as the last session... So I don't think I'm too far off. As I said it only felt ever so slightly worse than practice which was 90fps so high 80's sounds correct.

@pattikins No I wouldn't say it really effected my race pace. Hard to know with the race nerves versus practice calm

 

[Interslice]

Link is snipped... Can you put it in a link box?

Ha it's not letting me write acstuff.ru acstuff.com but .com = .ru

 

[Interslice]

I'm trying this auto tune again here. It can go up to 41%... Well over the 5Ghz

 

[Interslice]

Looks better. Would I need more fans and a better cooler to go much higher? Maybe that's a bad idea. Not sure at what point you risk bricking the CPU.

 

[Interslice]

597 3T and 778 4T were the new cinebench scores

I remember with my old case I 'hacked' it even higher by opening the window on a winters night, and taking the side case panel off ... But that gave me the fear of cooking it so i reset it to stock. Wasn't needed at the time anyway.

 

[RasmusP]

I'll reply tomorrow more thoroughly but for now:
For most Intels it's recommended to not go over 1.35v for 24/7 voltage, better stay at around 1.3v and not go beyond 1.4v.

However I've never seen a cpu dying from a bit of playing around with 1.5v (maximum in most bios).
My cpu mostly stays around 1.27v and shortly pushes to 1.35 when getting loaded up etc.

If you want to push your cpu, do it all manually.. These auto oc are okayish for some quick and dirty mild oc but not for really going for it...

 

[Interslice]

I'll reply tomorrow more thoroughly but for now:
For most Intels it's recommended to not go over 1.35v for 24/7 voltage, better stay at around 1.3v and not go beyond 1.4v.

However I've never seen a cpu dying from a bit of playing around with 1.5v (maximum in most bios).
My cpu mostly stays around 1.27v and shortly pushes to 1.35 when getting loaded up etc.

If you want to push your cpu, do it all manually.. These auto oc are okayish for some quick and dirty mild oc but not for really going for it...

It's sitting at 1.376 V when I run the 4 thread render in cinebench. Idles at 0.7. 1.44 - 1.456v then running the VR replay. That all a bit high?

 

[RasmusP]

It's sitting at 1.376 V when I run the 4 thread render in cinebench. Idles at 0.7. 1.44 - 1.456v then running the VR replay. That all a bit high?

A bit yeah.. Might work for years but it's a bit.. Unsafe... Kinda

 

[pattikins]

A bit yeah.. Might work for years but it's a bit.. Unsafe... Kinda

 

[RasmusP]

It's a bit tricky since these values come from extreme overclockers that are in contact with Intel but as far as I know barely anyone ever killed a CPU. These Intel i-series CPUs are pretty tough!

But the stock voltage for me is about 1.15 volt so pumping 1.4v into it is a bit massive..

What's kinda known though is that you get degradation over time when running the cpu hot and with high voltage for years. Depending on the settings it's not much though.

My i7 2600k ran at 4.4 GHz at 1.34v for 5 years. Than I got some bsods and put it to 1.38v. Ran for another 2 years.

Now it's running at 1.15v and stock setting (3.6 GHz) in my mom's homeoffice.
She can finally watch YouTube videos in 720p without stuttering. Hurray...
The amd x2 4400+ dual core went to 90% load when selecting anything higher than 480p...

 

[Interslice]

I've gotten a good 4 years value out of it so I'm fairly willing to push it on... Also in the back of my mind is the 3 laps from race end crash to blue screen of DNF! The fact I just need 2 FPS has me thinking it should be stable once i don't add any load to it..

 

[pattikins]

I've gotten a good 4 years value out of it so I'm fairly willing to push it on... Also in the back of my mind is the 3 laps from race end crash to blue screen of DNF! The fact I just need 2 FPS has me thinking it should be stable once i don't add any load to it..

I agree. However, it's worth bearing in mind that the AC video replay doesn't tax your cpu to the same degree as it does when you're playing AC... I think that is where the 65fps came in.

 

[RobertR1]

If you want to bench AC, then run the AC bench and compare your fps before and after changes.

CPU-Z, Cinebench are cache friendly and not sensitive to RAM so you might look great in those two synthetics but once you actually play the game, you might not notice much of an improvement. That's because your memory performance is the bottleneck.

 

[pattikins]

...

​

For the 11900k I searched for Cinebench R20 scores but that benchmark uses AVX CPU instructions so it's not really comparable to simracing.
I did some wild calculations from values I've found and in the end:
It's slightly better than the 5900x in single threading. The 11600k will be a bit worse and based on past generations, I will probably gain about 19% compared to my 10600k in fps.

In the table above that would mean:
10600k: 90 fps ; 11600k = 107,1 fps

Not really worse the money but if you need the magic 90 fps for VR, it's worse everything!

Now the benchmarks are in, do you plan on taking the plunge?