Share your CPUZ Benchmarks!

[Felix123BU]

hot damn, 5Ghz on that chip, I just gave my 4790K away this week to a friend in need, but could only go 4.6 allcore on a decent voltage, and it was delidded

same, could not go past 4.6 on my late 4790k back in the day whit out insane voltages, and it was also delidded and with liquid metal on a strong AIO, temps where not the problem, voltages where

 

[storm-chaser]

I managed to coax a bit more out of my HP z820 workstation...

 

[Makaveli]

 

[storm-chaser]

Slight improvement in bandwidth after bumping up the RAM from 4166 to 4400MHz, going to see if I can run this frequency on CL18...
What about z390 BLCK overclocking? So far it seems stable at 200 Mhz, a 100% overclock. Does it make any difference at all seeing how it's uncoupled from the PCIe bus and other integral systems? Maybe just a coincidence, but right off the bat I hit 4400MHz on the RAM, I was not able to do that with 100MHz BLCK....

Edit: Have not found the upper limit of this ram yet:

 

[Deleted member 178884]

Funny how you bombard my posts. Cute scores by the way, you're yet to even get close to my z390 rig.
Mine by the way, it's been months since you've tried to beat it so have a smaller reminder:

 

[storm-chaser]

View attachment 190247
Funny how you bombard my posts. Cute scores by the way, you're yet to even get close to my z390 rig

Im not in competition with you in any regard, not sure why you are bench racing here. I'm just having fun.

Are you familiar with overclocking 9th gen intel chips? Perhaps you can shed some light on front side bus overclocking for me.

Very happy to get 4533 on the RAM... I think that's where I'll settle up @ CL18

EDIT: Moving to cache overclocking now as the memory appears to be stable.

 

[Deleted member 178884]

Im not in competition with you in any regard, not sure why you are bench racing here. I'm just having fun.

Are you familiar with overclocking 9th gen intel chips? Perhaps you can shed some light on front side bus overclocking for me.

Very happy to get 4533 on the RAM... I think that's where I'll settle up @ CL18

EDIT: Moving to cache overclocking now as the memory appears to be stable.

View attachment 190250

Push uncore, 4.4 is pretty low and I'd recommend running around 4.8-5 or so depending on the CPU bin, as for your prior statement you asked for advice to only bombard my DM, constantly spam me with reactions for months on end so it's only fair I return the favour, move on and stop bombarding my posts it's childish quite frankly. As for BCLK I would advise against that if you are running NVME SSDs as it can cause damage to them and potentially break them, SATAs should be fine though - also, I would advise moving tRFC down to the 320-360 mark.

 

[storm-chaser]

Push uncore, 4.4 is pretty low and I'd recommend running around 4.8-5 or so depending on the CPU bin, as for your prior statement you asked for advice to only bombard my DM, constantly spam me with reactions for months on end so it's only fair I return the favour, move on and stop bombarding my posts it's childish quite frankly. As for BCLK I would advise against that if you are running NVME SSDs as it can cause damage to them and potentially break them, SATAs should be fine though - also, I would advise moving tRFC down to the 320-360 mark.

I could say the same about you, but really, If you have any issues please PM me and I'd be happy to discuss in a respectful conversation. We don't need the whole world to know about our minor disagreements and I'd prefer to keep this thread clean from strife if at all possible.

As for the memory tips, I will try them and report back. Is the PCIe bus not uncoupled from the FSB? Thought I read that somewhere where this issue with the NVMe drives only effects some Ryzen chips?

EDIT: Sorry the last few posts should be in the AIDA64 benchmark thread, I will post there going forward so as not to clutter this thread up with unrelated chatter.

 

[Deleted member 178884]

Is the PCIe bus not uncoupled from the FSB?

Some boards have a separate BCLK generator for the CPU+RAM (2 BLCK generators) so it isolates it from PCIe and similar. And the NVME drive issue is universal to any platform, if you push the BCLK on NVME drives it will eventually kill them at a fast rate.

 

[Vendor]

 

[Toothless]

Dual x5570.

Dual 2680v2.

Dual x5680.

 

[storm-chaser]

Dual x5570.

Dual 2680v2.

Dual x5680.

For reference here is my dual processor dell workstation (x2 Xeon X5650)

Running a higher bus speed yielded a slight improvement in single core and a slight dip in MT.

And this is the validation:

 

[mortenserv]

Well it's at the end of its lifespan so thought I could just as well up the OC on my 8700K to 5.4 MHz (AVX0) @ 1.420V. Hyperthreading disabled on purpose.

 

[Yukikaze]

15W Kaby Lake vs 25W Tiger Lake (Lenovo x390 vs Lenovo Thinkbook Yoga 14s):

 

[Ferrum Master]

15W Kaby Lake vs 25W Tiger Lake (Lenovo x390 vs Lenovo Thinkbook Yoga 14s):

If you divide the score to get performance per one watt... they are same...

 

[storm-chaser]

Working on stability now but I can get it to boot into windows at the very least... It's right near the limit of my custom loop which consists of two radiators: One 120 mm and one 360 mm... Question, how big of an impact does flow have? In other words, should I be running the pump at full throttle or back off a little bit? It's a freezemod pump with PWM control. BTW this is my highest OC, ever. !!!

EDIT: Looks like 5.602GHz is the upper limit for this chip.

 

[ezmafl]

10700-K, playing around with single core turbo overclocks -- 53, 53, 52, 52, 52, 52, 51, 50, 1.295v fixed, 47x Ring, LLC High on a Gigabyte Z490 Vision D. The 53x bump helped push the single core score to 629.

Running a few hours of P95 26.6 blended, forcing an hour or two on 112/112, running Forza Horizon 4 for a couple hours and trying 30 mins of R20 or R23 in loops don't cause any blue screen or WHEA errors. Not true 24-hour test by any stretch, but these little short runs initially look good.

R20 single core gets 542.

I tried 55, 55, 54, 54, 53, 53, 52, 50 at 1.35v fixed, LLC High, AVX 0. It blue-screened at 1.31v, so I bumped it to 1.35v. Haven't tried any stability tests whatsoever, but it got through CPU-Z.

Windows didn't like 57, 57, 56, 56, 1.35v. I wonder why...

It's probably better overall if I went to all core 51x. Settling at the moment with 53x4, 52x2, 51x1, 50x1 and testing stability at 1.295v.

 

[storm-chaser]

10700-K, playing around with single core turbo overclocks -- 53, 53, 52, 52, 52, 52, 51, 50, 1.295v fixed, 47x Ring, LLC High on a Gigabyte Z490 Vision D. The 53x bump helped push the single core score to 629.

Running a few hours of P95 26.6 blended, forcing an hour or two on 112/112, running Forza Horizon 4 for a couple hours and trying 30 mins of R20 or R23 in loops don't cause any blue screen or WHEA errors. Not true 24-hour test by any stretch, but these little short runs initially look good.

R20 single core gets 539.

I tried 55, 55, 54, 54, 53, 53, 52, 50 at 1.35v fixed, LLC High, AVX 0. It blue-screened at 1.31v, so I bumped it to 1.35v. Haven't tried any stability tests whatsoever, but it got through CPU-Z.

Windows didn't like 57, 57, 56, 56, 1.35v. I wonder why...

It's probably better overall if I went to all core 51x. Settling at the moment with 53x3, 52x4, 50x1 and testing stability at 1.295v.

I've been doing the same as of late. Turbo core is really fun to play around with.

Something very similar with my 9600KF: x53 x53 x53 x52 x51 x50
Here is my bench with these settings:

So I take it there are zero gains in IPC between 9th and 10th gen Intel CPUs? We are both within the margin of error, and at the same clock speed getting identical results.

Have you tried leaving the vcore settings on auto? As is the case with my MSI board, when you do that, it automatically goes into adaptive mode, indicating the board will seek out and deliver optimum voltage based on load / heat etc. This works really well for my 9600KF vs a vcore set to override mode, because the voltage will dip at idle but the clocks still remain really high, so it seems ideal for modifying turbo config.

 

[ezmafl]

I've been doing the same as of late. Turbo core is really fun to play around with.

Something very similar with my 9600KF: x53 x53 x53 x52 x51 x50
Here is my bench with these settings:

So I take it there are zero gains in IPC between 9th and 10th gen Intel CPUs? We are both within the margin of error, and at the same clock speed getting identical results.

Have you tried leaving the vcore settings on auto? As is the case with my MSI board, when you do that, it automatically goes into adaptive mode, indicating the board will seek out and deliver optimum voltage based on load / heat etc. This works really well for my 9600KF vs a vcore set to override mode, because the voltage will dip at idle but the clocks still remain really high, so it seems ideal for modifying turbo config.

Ha, too funny regarding the similarities. Good idea about trying adaptive; however, it'll be a few days before I have the extra time to tinker with this again. Was trying to have some computer fun after the kids went to bed. I want to try all-core 51x soon, as it's been a while, also want to look at finding a way to actually test the single core overclocks. Just because it's booting and "working," I'm concerned it may not be stable. Someone recommended that the new OCCT 8.0 can individually test. I also want to see if I force certain programs (handbrake) to not use certain cores, and will that then cause the processor to run said program at 51, 52 or 53 all cores?

However, I think Cinebench R20's single core test is utilizing the turbo speed -- at 5.2 or 5.3. Trying to get a baseline for other CPUs, online shows 539 is also what an i9-10900K stock runs, since that can turbo to 5.3. Fun stuff.

 

[storm-chaser]

Ha, too funny regarding the similarities. Good idea about trying adaptive; however, it'll be a few days before I have the extra time to tinker with this again. Was trying to have some computer fun after the kids went to bed. I want to try all-core 51x soon, as it's been a while, also want to look at finding a way to actually test the single core overclocks. Just because it's booting and "working," I'm concerned it may not be stable. Someone recommended that the new OCCT 8.0 can individually test. I also want to see if I force certain programs (handbrake) to not use certain cores, and will that then cause the processor to run said program at 51, 52 or 53 all cores?

However, I think Cinebench R20's single core test is utilizing the turbo speed -- at 5.2 or 5.3. Trying to get a baseline for other CPUs, online shows 539 is also what an i9-10900K stock runs, since that can turbo to 5.3. Fun stuff.

You could also disable 4 cores and set the two remaining active cores @ 5.3 or 5.4. Sometimes, you can get away with this despite the fact the same system may crash if you have an all core 5.4GHz OC.

I tried 55, 55, 54, 54, 53, 53, 52, 50 at 1.35v fixed, LLC High, AVX 0. It blue-screened at 1.31v, so I bumped it to 1.35v. Haven't tried any stability tests whatsoever, but it got through CPU-Z.

Another thing. If you are running close to your thermal margins with your turbos set this high, you can drop LLC down to the lowest level possible and re-run some torture tests to find the optimal setting. Slowly raise your LLC until stability is reach.

This was the case with my 9600KF because I had LLC set on max, which was pumping over 1.430 volts into the CPU under load, greatly reducing the performance of my custom loop and adding huge amounts of heat, which narrowed the margins even further.

So I methodically went through each LLC option until I came across the one that could pass torture tests and maintain stability. This means instead of 1.430 vcore it turned out to be more like 1.373 volts, enough to keep the OC stable.

You could also disable 4 cores and set the two remaining active cores @ 5.3 or 5.4. Sometimes, you can get away with this despite the fact the same system may crash if you have an all core 5.4GHz OC.

Wow. I am super impressed with the adaptive vcore on this motherboard. and After fixing LLC temps are substantially lower. matter of fact out of curiosity I have all my fans set at 800 rpm, the pump at 2000 rpm and each radiator fan the same, 800rpm. This is essentially an idle setting for this rig. Hard to believe but with the AIDA64 stress test, temps peak at about 163*F... totally rock solid! and that's at 5.0GHz all core turbo speed! That's right, these fans DONT ramp at all. Damn I'm beginning to like LGA 1151 more and more.

 

[mclaren85]

I know the core voltages are way too high but that's how AMD works. Settings are all default

 

[storm-chaser]

Ha, too funny regarding the similarities. Good idea about trying adaptive; however, it'll be a few days before I have the extra time to tinker with this again. Was trying to have some computer fun after the kids went to bed. I want to try all-core 51x soon, as it's been a while, also want to look at finding a way to actually test the single core overclocks. Just because it's booting and "working," I'm concerned it may not be stable. Someone recommended that the new OCCT 8.0 can individually test. I also want to see if I force certain programs (handbrake) to not use certain cores, and will that then cause the processor to run said program at 51, 52 or 53 all cores?

However, I think Cinebench R20's single core test is utilizing the turbo speed -- at 5.2 or 5.3. Trying to get a baseline for other CPUs, online shows 539 is also what an i9-10900K stock runs, since that can turbo to 5.3. Fun stuff.

I just tried an experimental setting with 53x 53x 53x 53x 51x 50x set for turbo boost.

This substantially improved my single core result as you can see below. Prior to this I had only three cores running at 53x, like the screenshot above....

 

[freeagent]

Nothing too crazy..

 

[mclaren85]

I just tried an experimental setting with 53x 53x 53x 53x 51x 50x set for turbo boost.

This substantially improved my single core result as you can see below. Prior to this I had only three cores running at 53x, like the screenshot above....

View attachment 192202

How could Intel manage to get so low cpu core voltage? Damn Intel is way better! My voltage hungry 3900x takes up to 1,45V. !

 

[freeagent]

How could Intel manage to get so low cpu core voltage? Damn Intel is way better! My voltage hungry 3900x takes up to 1,45V. !

He is probably up there too..