Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
cMP OC
- Thread starter UCDHIUS
- Start date
- Sort by reaction score
New high score 
https://browser.geekbench.com/v4/cpu/9478509
[doublepost=1534654814][/doublepost]Lets back that up that bench.. with a even higher bench https://browser.geekbench.com/v4/cpu/9478564
[doublepost=1534655809][/doublepost]Can we go higher? https://browser.geekbench.com/v4/cpu/9478729
Yes we can !
https://browser.geekbench.com/v4/cpu/9478509
[doublepost=1534654814][/doublepost]Lets back that up that bench.. with a even higher bench
https://browser.geekbench.com/v4/cpu/9478564[doublepost=1534655809][/doublepost]Can we go higher? https://browser.geekbench.com/v4/cpu/9478729
I really wonder if this is one of the new feature in 138.0.0.0.0.
I tried everything I can to OC the CPU with the old firmware. The best I can achieve is the system report higher clock speed, but actual performance shows no noticeable improvement. (But I never use CineBench as CPU performance reference. I only use GB. This may make the difference)
917cb in CineBench is definitely not a W3690 can do at the default clock speed normally in MacOS. I never test this in Windows 10. Not sure if the score algorithm is a bit different. May test that later when I have more time.
But the GB4 result still at the normal range. So, I still doubt if the CPU is really overclocked.
This is what I got with my non OC W3690 some time ago (still with 10.13.2 and 0084.B00). Nothing specially done (e.g. clean installation, or kill every unnecessary process in activity monitor), just run GB4 in macOS. The performance is not that far away from the OCed score which is done in Windows 10 with newer version GB4.
If the real clock is over 4GHz, I will expect a better improvement.
Or the CPU is actually OCed, but then hit the power throttling point almost straight away. Therefore, only very little performance gain. (I can’t see thermal throttling is an issue here. So, may be the pre-determined max power is quite low, and just good enough for the CPU to run at default clock speed)
When I OC my 8700K, I remember that I have to raise the power limit, otherwise, not much extra the CPU can do. I wonder if this is the limit on our cMP now.
If CineBench is less stressful for the CPU, therefore, it use less power to compute. It’s possible that can go higher clock if the limitation is actually the power limit, but not thermal limit, or the clock speed can’t be altered.
In fact, what suprised me is when I tried OC my W3690 few years ago. The system will actually hang if I input a multiplier which is too high. Feel like the system is actaully responding to my input. However due to no noticeable improvement observed in GB, I decided that’s a fail.
I tried everything I can to OC the CPU with the old firmware. The best I can achieve is the system report higher clock speed, but actual performance shows no noticeable improvement. (But I never use CineBench as CPU performance reference. I only use GB. This may make the difference)
917cb in CineBench is definitely not a W3690 can do at the default clock speed normally in MacOS. I never test this in Windows 10. Not sure if the score algorithm is a bit different. May test that later when I have more time.
But the GB4 result still at the normal range. So, I still doubt if the CPU is really overclocked.
This is what I got with my non OC W3690 some time ago (still with 10.13.2 and 0084.B00). Nothing specially done (e.g. clean installation, or kill every unnecessary process in activity monitor), just run GB4 in macOS. The performance is not that far away from the OCed score which is done in Windows 10 with newer version GB4.
If the real clock is over 4GHz, I will expect a better improvement.
Or the CPU is actually OCed, but then hit the power throttling point almost straight away. Therefore, only very little performance gain. (I can’t see thermal throttling is an issue here. So, may be the pre-determined max power is quite low, and just good enough for the CPU to run at default clock speed)
When I OC my 8700K, I remember that I have to raise the power limit, otherwise, not much extra the CPU can do. I wonder if this is the limit on our cMP now.
If CineBench is less stressful for the CPU, therefore, it use less power to compute. It’s possible that can go higher clock if the limitation is actually the power limit, but not thermal limit, or the clock speed can’t be altered.
In fact, what suprised me is when I tried OC my W3690 few years ago. The system will actually hang if I input a multiplier which is too high. Feel like the system is actaully responding to my input. However due to no noticeable improvement observed in GB, I decided that’s a fail.
Last edited:
This is the pedal to the metal, multicore is still climbing tweak by tweak
https://browser.geekbench.com/v4/cpu/9481266
[doublepost=1534673737][/doublepost]Its been backing up, run by run "consistently" if I must say.
[doublepost=1534674285][/doublepost]Back to back 920 CB runs
I understand that. The stability is there.
I now also tends to believe that the CPU does react to your input, but just how much and why.
So, if you revert the multiplier to the default setting. The scores will drop straight away. And if you tune them up again, the score will increase. Is that correct?
I think this is good enough to prove that we can actually some how OC the W3690 on a 5,1 in Windows 10 with 138.0.0.0.0.
However, the Mac performance seems still limited by something else.
Although I like seeing these numbers, I wouldn’t recommend anyone do this unless they have some spare parts.
I have some spare boards etc. so I’ll feel a little less worried if it goes kaboom.
The last run was wide open.
are you using Intel xtu to OC in windows?
always half wondered how well the VRM where made on the MP and not relay shore what there rated at or if you can pull temps from them in windows.
looks fun to do but not for me
No, ive been pumping wayyyyyyyyy more wattage etc etc than I ever thought possible all day and nothing happened yet. *knock on wood*
No, ive been pumping wayyyyyyyyy more wattage etc etc than I ever thought possible all day and nothing happened yet. *knock on wood*
It looks like you are getting a 5 MHz overclock. Pushing the silicon from 3460mhz to 3465mhz. The chip can go much faster without harming the silicon. If you can raise the voltage, it can easily go faster.
I’m hitting 250 MHz it’s blazing fast. In all seriousness though It’s running a stable 4.1-4.2 GHz though. Some Geekbench says 5MHz some say 8.1 GHz , 6.25 GHz lol.
In all seriousness though It’s running a stable 4.1-4.2 GHz though. Some Geekbench says 5MHz some say 8.1 GHz , 6.25 GHz lol.
Sure that’s not really operating at 4.2GHz. Even we compare that to the single core turbo clock 3.73 GHz. 4.2GHz still a 12% OC, and your GB only shows 3% faster than mine.
If that’s compare to the base 3.46GHz, 4.2GHz is a 21% OC. However, your multi cores score is just 4% higher than mine.
So, no matter what you insert, the CPU definitely not running at that speed. Also, I doubt if the voltage is really changed. Again, it’s not that hard for the software to report what we entered. I wonder if hardware monitor can correctly report the CPU operating voltage.
I did this kind of OC tests quite a few years ago. Really can’t member all the details. But from memory, no parameters suggested that I can really alter the voltage (e.g. temperature won’t change, and no system instability etc).
Apart from OC, I also tried undervolt. I can at least hang the system on the OC test (not good, but at least a sign that something is changed). For undervolt, absolutely zero effect.
However, that’s with the old firmware and few years old software. The new one may be different. I tried all softwares that I can download by that time. From memory, only one (at most two) really take my input, all others not even have proper feedback to my input.
With that “working” software. All I saw is just the clock speed changed but no noticeable performance improvement. But TBH, I didn’t pay attention if the CPU did 3-4% faster. I consider anything less than 5% is just my measuring error. And not worth to OC anyway.
In all seriousness though It’s running a stable 4.1-4.2 GHz though. Some Geekbench says 5MHz some say 8.1 GHz , 6.25 GHz lol.
That's not what Geekbench is showing
A Score of 4.1 / 4.2 will deliver a much higher score.
Open Hardware Monitor from Windows display's the speed and temps for all CPU cores. Can you post a screenshot of the clocks and temps from your w3690.
It would also be helpful to see CPU and SPD tabs from CPU-Z.
[doublepost=1534698735][/doublepost]
Nope. Perhaps a misunderstanding of Geekbench results.
Getting info from additional apps will help define what @UCDHIUS is seeing, as there are plenty of referenced results on Geekbench browser.
Last edited:
If you look at the difference between the 3.33ghz cpu and the 3.47ghz cpu should offer some indication of 15mhz overclock diffrence. Plus all them scores would be much better on a bare system with everything turned off. The problem with benchmarks are you can always tweak the system to improve scores. Set geek bench to high priority in Windows and it will gain score. Unless the overclocking software offers over volting the cpu and memory and changing the multiplier with out that only marginal over clocks will be possible. Changing clock speed on the fly is possible through FSB only. And the ram will only go so far.
I be done a lot of pc over clocking and have a 6950x at 4.4ghz from 3.0 ghz
[doublepost=1534700517][/doublepost]https://www.3dmark.com/spy/2841513
I be done a lot of pc over clocking and have a 6950x at 4.4ghz from 3.0 ghz
[doublepost=1534700517][/doublepost]https://www.3dmark.com/spy/2841513
Last edited:
I have validated CPU -Z links of it running @4.1 GHz in the original post. I can post another if you want.
I made sure to post, it when I first made a the thread.
[doublepost=1534721306][/doublepost]https://valid.x86.fr/8rt3e6
The posts you highlighted that actually show a higher geekbench score are clearly Windows machines that are overclocked. Note the gigabyte brand motherboard.
Plenty of people have OCd these chips with sometimes incredible results. But no one has ever done it on a cMP. Because you need access to a bios that doesn't exist.
Plenty of people have OCd these chips with sometimes incredible results. But no one has ever done it on a cMP. Because you need access to a bios that doesn't exist.
That was a joke hence the emicon
I see improvements in the software, I use games etc.
Take it for what it, I was honestly shocked my self with limited control that is available.. that I could even take it remomtly close to 4.1 GHz.
I thought I couldn’t even to get it to hold over turbo boost for a extended amount of time.
(At first, It would boost up to 4.1 GHz on 1-3 cores when it hit four or more throttle down to 3.80 GHz.) but I got over that hump.
I’ve been doing this for close to a day straight. I wouldn’t post photoshopped images on here just to say “look at me” “look at me”.
Once I get my 1080Ti back in, (my SFF PSU gave up the ghost.) I’ll do multiple screen recordings.
Temp is a clear indication of something is happening.
89-93C under load. Would you say, that’s “way to hot what are you doing”.
Here’s is the thing I have multiple spare parts. I honestly don’t care if something gave up and when kaboom while I was doing this.
Here is a validated link to CPU-Z of it running at 4.1 GHz.
https://valid.x86.fr/8rt3e6 please look at it.
[doublepost=1534723983][/doublepost]This is the jankiest OC I have ever done, do I recommend do this? Nope.. it’s took me a day and some to get where’s it’s at.
There is so much stuff to get around, to even get to 4.1GHz
I bought the machine with intention to try and see if I could even get over turbo boost and OC it.
At first I couldn’t get it to hold even turbo boost clocks, but I kept on trying.
That’s the reason why I bought 2 more trays and spare parts. (Just Incase it went kaboom)
On the flip side, IF someone else came out of nowhere, I wouldn’t believe it myself.
So I can understand where others are skeptical of this “since I just came out of no where” and said “look at this I have a overclocked 4.1/5,1”.
Last edited:
Thanks for the details on CPU-Z. Your CPU temp is way too high. I'd suggest installing the windows version of Macs Mac Control. To confirm, you are running a 4x4x1 ram configuration?
Can you post a screenshot of the CPU clocks and temps from your w3690 with open hardware monitor? Seeing your CPU core speeds and memory timings will help to better understand the hardware config in greater detail.
https://openhardwaremonitor.org
Link to download the latest version: Download Open Hardware Monitor 0.8.0 Beta
Release Version 0.8.0 Beta
- Added support for Intel Skylake, Kaby Lake and Airmont CPUs.
- Added support for Intel Xeon E5-26xx v4 and Xeon D-15xx CPUs.
- Added support for Intel Intel i5, i7 5xxC (14nm) CPUs.
- Added support for AMD family 15h model 30h APUs.
- Added support for ITE IT8620E and IT8628E super I/O chips.
- Added support for Nuvoton NCT6102D/NCT6106D super I/O chips.
- Added more Nvidia RAM sensors (free, used and total).
- Added more sensors for Samsung and Plextor SSDs.
- Replaced the ‘Temperature Difference From 100’ with a simple ‘Temperature’ sensor on hard drives.
- Fixed wrong Nvidia GPU clock min and plotting values.
- Fixed issues with Nuvoton NCT6791D super I/O chips after wake from S3 sleep state.
- Fixed incorrect OS version showing up in reports for Windows 10 and 8.1.
- Added customizable pen colors for the plot window.
- Improved the automatic plot pen color assignment strategy.
Last edited:
For everyone to see... this because MR likes to merge posts so some possibly miss it.
Below
This is the jankiest OC I have ever done, do I recommend do this? Nope.. it’s took me a day and some to get where’s it’s at.
There is so much stuff to get around, to even get to 4.1GHz
I bought the machine with intention to try and see if I could even get over turbo boost and OC it.
At first I couldn’t get it to hold even turbo boost clocks, but I kept on trying.
That’s the reason why I bought 2 more trays and spare parts. (Just Incase it went kaboom)
On the flip side, IF someone else came out of nowhere, I wouldn’t believe it myself ethier.
So I can understand where others are skeptical of this “since I just came out of no where” and said “look at this, look at this, I have a overclocked 4.1/5,1”.
It’s hard to believe yes, as I couldn’t believe it myself at first.
I wouldn’t come on here to macrumors.. just to say “look at me, look at me” to “show off”.
This post/thread wasn’t meant to do that. It was meant to document this “journey” of improvements (if any)
The chip won’t let me go past 4.2GHz stable for extended amount of time. Could it be the chip?, the limited controls that’s available?, the tray, the main board, power supply? Who knows.
[doublepost=1534726178][/doublepost]@handheldgames I missed this part of your post
click on all the links on geekbench, with my name.
Look at the max reported clocks
That’s where I’m getting where “geek bench” says “5MHz, 6.25 GHz, 8GHz etc.” as max reported clocks. It’s inaccurate at best for max reported clock speed.
Below
This is the jankiest OC I have ever done, do I recommend do this? Nope.. it’s took me a day and some to get where’s it’s at.
There is so much stuff to get around, to even get to 4.1GHz
I bought the machine with intention to try and see if I could even get over turbo boost and OC it.
At first I couldn’t get it to hold even turbo boost clocks, but I kept on trying.
That’s the reason why I bought 2 more trays and spare parts. (Just Incase it went kaboom)
On the flip side, IF someone else came out of nowhere, I wouldn’t believe it myself ethier.
So I can understand where others are skeptical of this “since I just came out of no where” and said “look at this, look at this, I have a overclocked 4.1/5,1”.
It’s hard to believe yes, as I couldn’t believe it myself at first.
I wouldn’t come on here to macrumors.. just to say “look at me, look at me” to “show off”.
This post/thread wasn’t meant to do that. It was meant to document this “journey” of improvements (if any)
The chip won’t let me go past 4.2GHz stable for extended amount of time. Could it be the chip?, the limited controls that’s available?, the tray, the main board, power supply? Who knows.
[doublepost=1534726178][/doublepost]@handheldgames I missed this part of your post
click on all the links on geekbench, with my name.
Look at the max reported clocks
That’s where I’m getting where “geek bench” says “5MHz, 6.25 GHz, 8GHz etc.” as max reported clocks. It’s inaccurate at best for max reported clock speed.
Last edited:
Thanks for all these info. It seems something really changed. May be the new firmware, may be the OC software. But anyway, it’s good to see that’s possible.