This is my first laptop in a long time, and my first attempt at undervolting and overclocking. I'm definitely not an expert and suggestions and corrections are welcome! I muddled my way through, but it seemed to turn out pretty well. I got ~12% boost in GPU performance and a whopping 29% in multicore CPU. Now my 2060 is as fast as a stock MaxP 2070, and my i7 Intel processor is nearly as good as a Ryzen 4600H...
On my new laptop I started with the GPU. Unlike desktop systems, laptops have strict power limits that really nerf the GPU performance. This is set by the manufacturer and usually can't be changed. Mine is set to the highest value for a mobile 2060 [115W] but that is still very restrictive. Any time it has a high load it will slam into that 115W ceiling pretty quickly, even though the temperatures are fine [~75C with the stock fan curve]. So even though the GPU is capable of running at ~2100MHz at the stock 1.063v, it never gets close because the power consumption would be over 115W. So in order to get higher performance you need to do a combination of undervolting and overclocking. I started with no changes to the default laptop configuration [ie I don't know what the power settings, Dragon Center settings, or anything else were]. This looks like a good guide for Afterburner [though I didn't use it]: //www.msi.com/blog/get-a-free-performance-boost-with-afterburner-oc-scanner
Follow that guide and do the scan, then save this scan as Profile #5. Keep that so you have a good starting point for your tweaks. This is a trial and error process, but for this laptop I'd suggest trying a max voltage in the .925-.938 range. Hit shift+left mouse to select everything to the right of that point. Then click the last point on the curve and drag it down to value of the first point in the selection [say the frequency value at the .925 voltage]. Sometimes that works, and sometimes it partially works. Be prepared to do a fair bit of futzing around. What you are trying to do is get the curve to be flat to the right of that point. This will keep the GPU from exceeding that voltage. When you like the look of it, hit the checkmark on the main panel, then save it as one of the profiles. Open HWinfo and go to the GPU section to monitor clocks and temperatures. Now fire up a benchmark [I used Superposition 4k optimized mostly] and let it run. If the temperature is below 80C, you're fine for now, otherwise you'll need to increase fan speeds. I don't know if this laptop temperature-throttles since I never pushed it that high; it was easy to stay ~75C or less. This laptop has a button to the right of the keyboard [the bottom one] that lets you set both fans to the max if you wish.
Now look at the max power. If you are getting over 110W, then drop the voltage even more [move lower on the curve]. You are going to need some headroom for increasing Vram clocks, plus if you dare you can try raising above the scan curve [increase clocks at a given voltage]. Look for artifacts in the benchmark runs [or crashes!] to let you know you went a bit too far...
I read that Nvidia switched to low voltage Vram on mobile 2060s for 2020, supposedly to improve performance by reducing Vram wattage and hence allowing more power for the GPU cores. Based on my testing I'm skeptical of that reasoning. That dropped the stock Vram clock down to 5500 from 7000... which is a lot. And I got a way bigger performance/watt boost [~5x!] overclocking the Vram vs the GPU. So I'd definitely suggest overclocking the Vram as much as you can. I was able to go up to 6750 [vs 5500 stock] and it's been stable with no artifacts in everything I've tried.
I produced curves at 4 different voltages/power settings, and checked to make sure they were stable. The summary of my results on the GPU is below. I also ran Timespy and got a score of 7312 vs 6587 stock; so ~11% improvement there similar to the Unigine benchmarks. These were before I did any CPU mods, but I checked a couple later, and it made little difference. In one of the images below you can see that I got a 6126 score in Superposition vs 6099 before CPU mods. That makes sense because the benchmarks are very GPU focused, but in games that use more CPU the gains might be even greater.
Tweaking the CPU was a bit more involved, but greater gains were had. First you need to enable overclocking in the bios. Since Plundervolt many laptops do not allow this, but MSI does.
To get into the bios, hit the delete key while booting.
1] To get into the extended bios, press these 4 keys together: Right ctrl+shift+left alt+F2.
2] Go the the Advanced tab, scroll down to Overclocking Performance Menu: Overclocking Feature [Enabled]
3] Hit the back button, and now scroll down to Power and performance: CPU Power management control: CPU Lock configuration: CFG Lock and Overclock Lock [Disable]
4] Save and exit
Unlike the GPU, the CPU isn't really power limited [after mods in Throttlestop]; rather it's frequency and possibly temperature limited. The 6 core turbo limit appears to be locked at 4,290 MHz and my machine temperature throttles at 95C. I don't think I can change either. Using undervolting and other settings in Throttlestop I was able to get all 6 cores to run at the 4,290 Mhz max frequency at ~91C [fans on full bore], indefinitely. That consumes ~68W vs >80W on stock clocks; which would only be seen for a few seconds before it throttled.
The Throttlestop Fivr settings I settled on are shown below. Some "guides" will tell you that Core and Cache settings need to be the same, but I saw unclewebb mention that this is not true. On this processor it seems you can undervolt the Core a lot more. I didn't even try a more than -.200v offset because I thought that was huge, and it was stable. I tried -.085v on the Cache and it crashed in Prime95, but it seems stable at -.075.
I was frustrated for awhile because the CPU behavior seemed to be hardcoded to throttle to 45W after a few seconds even if temperatures were fine. Then I read about "Disable and Lock Turbo Power Limits" in Throttlestop; that did the trick! I was able to get the CPU to use my TPL settings instead. Like I said earlier, I can run at max frequency and ~68W "forever" instead of dropping down to 45W, which makes a huge difference. My highest Cinebench 20 score was 3377 vs 2620 stock. The single core score was unchanged at 475; not too surprising I guess since a single core isn't throttled by power or temperature at the stock settings. I may change these settings depending on what applications or games I'm running, and how tolerant I am of fan noise.
It was kinda fun seeing how I could push the hardware and there is really no downside. It wasn't that hard even with no experience, and the gains were substantial. I highly recommend giving it a try.
On my new laptop I started with the GPU. Unlike desktop systems, laptops have strict power limits that really nerf the GPU performance. This is set by the manufacturer and usually can't be changed. Mine is set to the highest value for a mobile 2060 [115W] but that is still very restrictive. Any time it has a high load it will slam into that 115W ceiling pretty quickly, even though the temperatures are fine [~75C with the stock fan curve]. So even though the GPU is capable of running at ~2100MHz at the stock 1.063v, it never gets close because the power consumption would be over 115W. So in order to get higher performance you need to do a combination of undervolting and overclocking. I started with no changes to the default laptop configuration [ie I don't know what the power settings, Dragon Center settings, or anything else were]. This looks like a good guide for Afterburner [though I didn't use it]: //www.msi.com/blog/get-a-free-performance-boost-with-afterburner-oc-scanner
Follow that guide and do the scan, then save this scan as Profile #5. Keep that so you have a good starting point for your tweaks. This is a trial and error process, but for this laptop I'd suggest trying a max voltage in the .925-.938 range. Hit shift+left mouse to select everything to the right of that point. Then click the last point on the curve and drag it down to value of the first point in the selection [say the frequency value at the .925 voltage]. Sometimes that works, and sometimes it partially works. Be prepared to do a fair bit of futzing around. What you are trying to do is get the curve to be flat to the right of that point. This will keep the GPU from exceeding that voltage. When you like the look of it, hit the checkmark on the main panel, then save it as one of the profiles. Open HWinfo and go to the GPU section to monitor clocks and temperatures. Now fire up a benchmark [I used Superposition 4k optimized mostly] and let it run. If the temperature is below 80C, you're fine for now, otherwise you'll need to increase fan speeds. I don't know if this laptop temperature-throttles since I never pushed it that high; it was easy to stay ~75C or less. This laptop has a button to the right of the keyboard [the bottom one] that lets you set both fans to the max if you wish.
Now look at the max power. If you are getting over 110W, then drop the voltage even more [move lower on the curve]. You are going to need some headroom for increasing Vram clocks, plus if you dare you can try raising above the scan curve [increase clocks at a given voltage]. Look for artifacts in the benchmark runs [or crashes!] to let you know you went a bit too far...
I read that Nvidia switched to low voltage Vram on mobile 2060s for 2020, supposedly to improve performance by reducing Vram wattage and hence allowing more power for the GPU cores. Based on my testing I'm skeptical of that reasoning. That dropped the stock Vram clock down to 5500 from 7000... which is a lot. And I got a way bigger performance/watt boost [~5x!] overclocking the Vram vs the GPU. So I'd definitely suggest overclocking the Vram as much as you can. I was able to go up to 6750 [vs 5500 stock] and it's been stable with no artifacts in everything I've tried.
I produced curves at 4 different voltages/power settings, and checked to make sure they were stable. The summary of my results on the GPU is below. I also ran Timespy and got a score of 7312 vs 6587 stock; so ~11% improvement there similar to the Unigine benchmarks. These were before I did any CPU mods, but I checked a couple later, and it made little difference. In one of the images below you can see that I got a 6126 score in Superposition vs 6099 before CPU mods. That makes sense because the benchmarks are very GPU focused, but in games that use more CPU the gains might be even greater.
Tweaking the CPU was a bit more involved, but greater gains were had. First you need to enable overclocking in the bios. Since Plundervolt many laptops do not allow this, but MSI does.
To get into the bios, hit the delete key while booting.
1] To get into the extended bios, press these 4 keys together: Right ctrl+shift+left alt+F2.
2] Go the the Advanced tab, scroll down to Overclocking Performance Menu: Overclocking Feature [Enabled]
3] Hit the back button, and now scroll down to Power and performance: CPU Power management control: CPU Lock configuration: CFG Lock and Overclock Lock [Disable]
4] Save and exit
Unlike the GPU, the CPU isn't really power limited [after mods in Throttlestop]; rather it's frequency and possibly temperature limited. The 6 core turbo limit appears to be locked at 4,290 MHz and my machine temperature throttles at 95C. I don't think I can change either. Using undervolting and other settings in Throttlestop I was able to get all 6 cores to run at the 4,290 Mhz max frequency at ~91C [fans on full bore], indefinitely. That consumes ~68W vs >80W on stock clocks; which would only be seen for a few seconds before it throttled.
The Throttlestop Fivr settings I settled on are shown below. Some "guides" will tell you that Core and Cache settings need to be the same, but I saw unclewebb mention that this is not true. On this processor it seems you can undervolt the Core a lot more. I didn't even try a more than -.200v offset because I thought that was huge, and it was stable. I tried -.085v on the Cache and it crashed in Prime95, but it seems stable at -.075.
I was frustrated for awhile because the CPU behavior seemed to be hardcoded to throttle to 45W after a few seconds even if temperatures were fine. Then I read about "Disable and Lock Turbo Power Limits" in Throttlestop; that did the trick! I was able to get the CPU to use my TPL settings instead. Like I said earlier, I can run at max frequency and ~68W "forever" instead of dropping down to 45W, which makes a huge difference. My highest Cinebench 20 score was 3377 vs 2620 stock. The single core score was unchanged at 475; not too surprising I guess since a single core isn't throttled by power or temperature at the stock settings. I may change these settings depending on what applications or games I'm running, and how tolerant I am of fan noise.
It was kinda fun seeing how I could push the hardware and there is really no downside. It wasn't that hard even with no experience, and the gains were substantial. I highly recommend giving it a try.
@rruff - The FIVR window shows that when 6 cores are active, the maximum multiplier is 43 so your CPU will max out at 4290 MHz [43.0 X 99.767 MHz]. That is as good as it gets. Your Cinebench results confirm that your CPU is running much better now. Pretty bad when you buy a new and improved laptop with a powerful 10750H and it initially runs slower than a well tuned 8750H. No more worries that an old school 8750H or 9750H will be kicking sand in your face.
For cache offset voltage, most 10th Gen CPUs like the 10750H seem to be stable somewhere around -75 mV. The older 6 core CPUs were usually good for -125 mV but not the 10th Gen. For most mobile CPUs, a maximum core offset request approximately -100 mV beyond the cache request seems to be where temperature or performance improvements in R20 stop. That means somewhere around -175 mV for the core offset for your 10750H. It is OK to go beyond this number. The CPU ignores any excess. You can set the core offset to a crazy number like -1000 mV. Most of that will be ignored.
In the TPL window you have your power limits set to 55W and 75W and then you set the turbo time limit to more than 3 million seconds which is ~42 days. That means if you run your CPU at full bore for 42 days, your CPU will finally switch the power limit from PL2 - 75W down to PL1 - 55W. I guess after running at full power for 6 weeks straight, the CPU needs a break.
So I have to ask, why do people do this? Why do people set the turbo time limit to this value? I know lots of YouTube guides recommend this but to me it makes no sense. If you never want your CPU to drop down to 55W, set both power limits to 75W and set the turbo time limit to the default value which is 28 seconds. To the CPU and to me, that request makes sense. Whenever I see a setting of 3 million seconds I always think, there goes another guy that has no idea what that setting does.
I thought about limiting this to 448 seconds. For some recent CPUs, Intel says that is the max value. I knew there would be complaints from the YouTube experts if I changed this so I just left it as is.
Thanks for sharing your results.
For cache offset voltage, most 10th Gen CPUs like the 10750H seem to be stable somewhere around -75 mV. The older 6 core CPUs were usually good for -125 mV but not the 10th Gen. For most mobile CPUs, a maximum core offset request approximately -100 mV beyond the cache request seems to be where temperature or performance improvements in R20 stop. That means somewhere around -175 mV for the core offset for your 10750H. It is OK to go beyond this number. The CPU ignores any excess. You can set the core offset to a crazy number like -1000 mV. Most of that will be ignored.
In the TPL window you have your power limits set to 55W and 75W and then you set the turbo time limit to more than 3 million seconds which is ~42 days. That means if you run your CPU at full bore for 42 days, your CPU will finally switch the power limit from PL2 - 75W down to PL1 - 55W. I guess after running at full power for 6 weeks straight, the CPU needs a break.
So I have to ask, why do people do this? Why do people set the turbo time limit to this value? I know lots of YouTube guides recommend this but to me it makes no sense. If you never want your CPU to drop down to 55W, set both power limits to 75W and set the turbo time limit to the default value which is 28 seconds. To the CPU and to me, that request makes sense. Whenever I see a setting of 3 million seconds I always think, there goes another guy that has no idea what that setting does.
I thought about limiting this to 448 seconds. For some recent CPUs, Intel says that is the max value. I knew there would be complaints from the YouTube experts if I changed this so I just left it as is.
Thanks for sharing your results.
Thought I'd submit an update...
I've been running the "TS Bench" a lot lately and found I was getting an occasional error. So I moved the cache voltage up slowly, and settled at -.061v. That's seems to be error free. The CPU consumes slightly more power now, but only ~1W. 69W vs 68W running all cores at 4,290MHz. Speed is the same, since throttling is still not happening.
Is TS Bench the best thing to use for this? It's very easy anyway.
EDIT: Another thing, that probably no one but me will care about, but I thought I'd mention it anyway. I noticed that after running a video benchmark, power would continue to be fed to the GPU even though it was idle. ~9W constant, so it was not a small amount. If I played a video the iGPU would be used but the dGPU was still just sitting there consuming 9W. I intend use this laptop in a vehicle with a solar panel and batteries, so gratuitous watts are bad, even plugged in. The reason why normal people wouldn't care is because as soon as I unplugged the power, the 9 watts disappeared... and it stayed off after plugging back in. So, it would never happen while using the internal battery.
I tried changing all sorts of settings, power plans, etc... with no luck. Figured I could just unplug the cord or reboot if that was the only thing that worked. Then I started messing with Afterburner. First thing I noticed was that AB had a bunch of monitoring functions polling the GPU regularly. I turned these off, but it didn't fix the issue. Shutting down AB doesn't work either; the dGPU still stays at 9W. Then while I was randomly pushing buttons, I noticed that if I selected my 3rd profile [2nd fastest one], the dGPU power would drop to zero. It didn't happen with any other profile and works with that one every time! Weird...
Turning off monitoring is still a good idea because AB will wake up the dGPU on a regular basis and waste watts if they are on.
I've been running the "TS Bench" a lot lately and found I was getting an occasional error. So I moved the cache voltage up slowly, and settled at -.061v. That's seems to be error free. The CPU consumes slightly more power now, but only ~1W. 69W vs 68W running all cores at 4,290MHz. Speed is the same, since throttling is still not happening.
Is TS Bench the best thing to use for this? It's very easy anyway.
EDIT: Another thing, that probably no one but me will care about, but I thought I'd mention it anyway. I noticed that after running a video benchmark, power would continue to be fed to the GPU even though it was idle. ~9W constant, so it was not a small amount. If I played a video the iGPU would be used but the dGPU was still just sitting there consuming 9W. I intend use this laptop in a vehicle with a solar panel and batteries, so gratuitous watts are bad, even plugged in. The reason why normal people wouldn't care is because as soon as I unplugged the power, the 9 watts disappeared... and it stayed off after plugging back in. So, it would never happen while using the internal battery.
I tried changing all sorts of settings, power plans, etc... with no luck. Figured I could just unplug the cord or reboot if that was the only thing that worked. Then I started messing with Afterburner. First thing I noticed was that AB had a bunch of monitoring functions polling the GPU regularly. I turned these off, but it didn't fix the issue. Shutting down AB doesn't work either; the dGPU still stays at 9W. Then while I was randomly pushing buttons, I noticed that if I selected my 3rd profile [2nd fastest one], the dGPU power would drop to zero. It didn't happen with any other profile and works with that one every time! Weird...
Turning off monitoring is still a good idea because AB will wake up the dGPU on a regular basis and waste watts if they are on.
Using P95 for stability testing will significantly lower your attainable OC. Most folks use an older version of P95 so they don't hurt their hardware when AVX and other newer instruction sets are present. That mean you OC is not necessarily stable when they are. I have had 24 hour stable P95 OCs fail in 20 minutes under RoG Real Bench. You can lower temps by about 7C or more with RB and therefore potentially get an extra 0.1 OC or more.
Another tip, custom build laptops usually don't have limiters found in popular gaming brands. Remember back when if to tried to run certain stress tests and the lappie would refuse to run them ?... I was shocked when I tried and they all ran fine. back then you'd even have an array of base chassis designs with desktop versions of CPUs and GFX cards.
Don't see that much anymore ... got an email this morning, a friend of my son is planning on having this built for him when either tax refund or $1400 Stimulus check arrives ... I'll be installing OS and programs doing basic set up ... Im expecting problems ... may take me a week or 2 to get it just right
10th i7-10870H Processor
HM470 Chipset
0 Dead Pixel warranty
RTX 3070 8GB6 Max-Q Design
Thermal Grizzly Kryonaut Thermal Compound - CPU + GPU
32GB Dual Channel DDR4 SDRAM at 2933MHz - 2 X 16GB
500GB Samsung 970 EVO Plus M.2 PCIe NVMe SSD Boot and Program Drive
2TB SAMSUNG® 970 EVO Plus PCIe NVMe M.2 SSD [Slot 2] Data Drive
Intel® Wi-Fi 6 AX201 M.2 AX + Bluetooth® 5.1 Combo Card
Power Supply 180 watts
Will come in at about $1850 delivered
Another tip, custom build laptops usually don't have limiters found in popular gaming brands. Remember back when if to tried to run certain stress tests and the lappie would refuse to run them ?... I was shocked when I tried and they all ran fine. back then you'd even have an array of base chassis designs with desktop versions of CPUs and GFX cards.
Don't see that much anymore ... got an email this morning, a friend of my son is planning on having this built for him when either tax refund or $1400 Stimulus check arrives ... I'll be installing OS and programs doing basic set up ... Im expecting problems ... may take me a week or 2 to get it just right
10th i7-10870H Processor
HM470 Chipset
0 Dead Pixel warranty
RTX 3070 8GB6 Max-Q Design
Thermal Grizzly Kryonaut Thermal Compound - CPU + GPU
32GB Dual Channel DDR4 SDRAM at 2933MHz - 2 X 16GB
500GB Samsung 970 EVO Plus M.2 PCIe NVMe SSD Boot and Program Drive
2TB SAMSUNG® 970 EVO Plus PCIe NVMe M.2 SSD [Slot 2] Data Drive
Intel® Wi-Fi 6 AX201 M.2 AX + Bluetooth® 5.1 Combo Card
Power Supply 180 watts
Will come in at about $1850 delivered
Thanks rruff for the detailed guide. Helped me with a MSI GL65 Leopard same cpu but 2070 Super.
Just a heads up for anyone reading this post. I followed this guide exactly but couldn't reach the 68W draw PL1 at first. It always went down to 45W after the sustained PL2 of 75. Looked everywhere turns out the MSI Dragon Center was set to Balanced. Changed it to Extreme Performance and now I can get that 68W sustained power draw instead of defaulting to 45W.
Besides that I dont know how are you hitting a 3300 score on Cinebench R20. Could it be because I haven't undervolted my GPU? [although it doesn't make sense since Cine doesnt use the GPU]. Im hitting 2800 on an average of 5 tests [was 2700 on the same average before Undervolting]. Can you tell me what tests you used and for how long each you used for testing stability? Thanks in advance.
PS. Disabling Prochot and amping the limit from 95 to 99 didnt increase the score on R20 just upped my overall temp limits [like it should I know]
Screens from my TS settings and HWINFO64 when running R20 Multi.
Just a heads up for anyone reading this post. I followed this guide exactly but couldn't reach the 68W draw PL1 at first. It always went down to 45W after the sustained PL2 of 75. Looked everywhere turns out the MSI Dragon Center was set to Balanced. Changed it to Extreme Performance and now I can get that 68W sustained power draw instead of defaulting to 45W.
Besides that I dont know how are you hitting a 3300 score on Cinebench R20. Could it be because I haven't undervolted my GPU? [although it doesn't make sense since Cine doesnt use the GPU]. Im hitting 2800 on an average of 5 tests [was 2700 on the same average before Undervolting]. Can you tell me what tests you used and for how long each you used for testing stability? Thanks in advance.
PS. Disabling Prochot and amping the limit from 95 to 99 didnt increase the score on R20 just upped my overall temp limits [like it should I know]
Screens from my TS settings and HWINFO64 when running R20 Multi.
@EEAM.Peru - Most people with bad benchmark scores have too much stuff running in the background on their computer. When your computer is idle at the desktop with only ThrottleStop open, what does ThrottleStop report for C0%? Here is what an idle computer reports.
I guess up to 0.5% is OK but if your computer is way higher than that, your Cinebench scores will not be great.
Do another test and post a screenshot while Cinebench is in progress.
If you are using MSI Dragon Center to manage your turbo power limits, maybe in the ThrottleStop TPL window you can check, Disable Power Limit Control. You do not need both MSI Dragon Center and ThrottleStop fighting over your CPU power limits. If Dragon Center works, you do not need ThrottleStop managing your turbo power limits.
I guess up to 0.5% is OK but if your computer is way higher than that, your Cinebench scores will not be great.
HWiNFO64 is a wonderful program, lots of useful information, but I would not leave that running when benchmark testing. When running Cinebench, exit HWiNFO and only run ThrottleStop with the Limit Reasons window open. HWiNFO interferes with the Limit Reasons data within the CPU. Watch Limit Reasons when testing. Does anything light up red? Any power limit throttling or maybe thermal throttling? Perhaps this is the reason your Cinebench scores are low.EEAM.Peru said:Screens from my TS settings and HWINFO64 when running R20 Multi.Click to expand...
Do another test and post a screenshot while Cinebench is in progress.
If you are using MSI Dragon Center to manage your turbo power limits, maybe in the ThrottleStop TPL window you can check, Disable Power Limit Control. You do not need both MSI Dragon Center and ThrottleStop fighting over your CPU power limits. If Dragon Center works, you do not need ThrottleStop managing your turbo power limits.
EEAM.Peru said:Thanks rruff for the detailed guide. Helped me with a MSI GL65 Leopard same cpu but 2070 Super.
Just a heads up for anyone reading this post. I followed this guide exactly but couldn't reach the 68W draw PL1 at first. It always went down to 45W after the sustained PL2 of 75. Looked everywhere turns out the MSI Dragon Center was set to Balanced. Changed it to Extreme Performance and now I can get that 68W sustained power draw instead of defaulting to 45W.
Besides that I dont know how are you hitting a 3300 score on Cinebench R20. Could it be because I haven't undervolted my GPU? [although it doesn't make sense since Cine doesnt use the GPU]. Im hitting 2800 on an average of 5 tests [was 2700 on the same average before Undervolting]. Can you tell me what tests you used and for how long each you used for testing stability? Thanks in advance.
PS. Disabling Prochot and amping the limit from 95 to 99 didnt increase the score on R20 just upped my overall temp limits [like it should I know]
Screens from my TS settings and HWINFO64 when running R20 Multi.Click to expand...
Made an account just to comment on this post. I managed a score of 3384 in R20 with a -0.090 undervolt on the CPU and -0.050 on the cache. And this was multiple, back to back passes that all resulted in the same exact score. Temps hit a max of 89C with cooler boost on. Score was down to about 3150ish with cooler boost off and the fans on 'silent' mode, so the CPU was thermal throttling to about ~4ghz[cooler boost is 4.3]. The problem is, it's not stable whatsoever in games. I did the TS benchmark on every setting and got 0 errors, and ran probably 20 R20 benchmarks, no crashing or other issues. As soon as I fire up a game, it just BSOD's with a DPC watchdog violation and reboots. In games, I could only manage a 0.030 undervolt on the CPU/Cache. I guess some units undervolt better. YMMV.
Thanks for the quick replies and sorry for taking so long to answer. A relative of mine got ill. She is better now.
1. Thanks ruff I didnt know I had a button for cooler boost that sets the fans to the max. Although the laptops becomes a Boeing 747 and temps improve I don´t know how viable is this for the longevity of the fans.
2. MSI Dragon center doesn't interfere with the voltage, any internal undervolting or power limits as far as I can tell and researched. It just gives fan profiles. Using Extreme provides the best scores.
3. Thanks unclewebb I didn´t know that about R20 and background processes.
4. Now for scores. I removed the TPL modifications from Ruff and only kept the cpu and cache undervolt for now. Remembered that Uncleweb mentioned something about speedshift and speed step in other posts [dont remember exactly whay I would need to check on that again] so also didn't use Ruff's modifications on those options.
Tested stability with The witcher 3 Novigrad 30 minutes all on ultra no crashes [2070s 115w no undervolt] + Aida 64 Extreme and Unigine heaven 90 minutes
A. Extreme profile + no cooler boost + no Undervolt: 3000 score [5 passes]
B. Extreme profile + no cooler boost + Undervolt -0.01590 undervolt on the CPU -0.059 on the cache : 3165 score [5 passes]
[numbers following Uncleweb recommmendation from other posts on this cpu model]
C. Extreme profile + cooler boost + Undervolt -0.01590 undervolt on the CPU -0.059 on the cache : 3170 score [5 passes]
[Barely any difference probable because of TPL limits]
The question is will removing power limits increase the performance while keeping temperatures or lowering them? If the answer is no, the performance will increase, but the temperatures will too I don't know if it's worth it in the long run. I've read Unclewebb's argument that pushing the temps a bit more wont matter to the cpu since they are designed to go up to 100 C° and I agree on that but putting more heat around all the components and the mobo helps the longevity of all the laptop? I think better scores and less temps is the holy grail most are looking for.
PS. From Jarrod's Tech review of this unit and his R20 scores [which are almost identical to mine] saw that he gained the most while dropping temps from using a laptop cooler since this model has good vents underneath it. Will get one this saturday.
Added screen is Extreme profile + no cooler boost + Undervolt -0.01590 undervolt on the CPU -0.059 on the cache without using Ruff's TPL modifications
1. Thanks ruff I didnt know I had a button for cooler boost that sets the fans to the max. Although the laptops becomes a Boeing 747 and temps improve I don´t know how viable is this for the longevity of the fans.
2. MSI Dragon center doesn't interfere with the voltage, any internal undervolting or power limits as far as I can tell and researched. It just gives fan profiles. Using Extreme provides the best scores.
3. Thanks unclewebb I didn´t know that about R20 and background processes.
4. Now for scores. I removed the TPL modifications from Ruff and only kept the cpu and cache undervolt for now. Remembered that Uncleweb mentioned something about speedshift and speed step in other posts [dont remember exactly whay I would need to check on that again] so also didn't use Ruff's modifications on those options.
Tested stability with The witcher 3 Novigrad 30 minutes all on ultra no crashes [2070s 115w no undervolt] + Aida 64 Extreme and Unigine heaven 90 minutes
A. Extreme profile + no cooler boost + no Undervolt: 3000 score [5 passes]
B. Extreme profile + no cooler boost + Undervolt -0.01590 undervolt on the CPU -0.059 on the cache : 3165 score [5 passes]
[numbers following Uncleweb recommmendation from other posts on this cpu model]
C. Extreme profile + cooler boost + Undervolt -0.01590 undervolt on the CPU -0.059 on the cache : 3170 score [5 passes]
[Barely any difference probable because of TPL limits]
The question is will removing power limits increase the performance while keeping temperatures or lowering them? If the answer is no, the performance will increase, but the temperatures will too I don't know if it's worth it in the long run. I've read Unclewebb's argument that pushing the temps a bit more wont matter to the cpu since they are designed to go up to 100 C° and I agree on that but putting more heat around all the components and the mobo helps the longevity of all the laptop? I think better scores and less temps is the holy grail most are looking for.
PS. From Jarrod's Tech review of this unit and his R20 scores [which are almost identical to mine] saw that he gained the most while dropping temps from using a laptop cooler since this model has good vents underneath it. Will get one this saturday.
Added screen is Extreme profile + no cooler boost + Undervolt -0.01590 undervolt on the CPU -0.059 on the cache without using Ruff's TPL modifications
Thanks for the reply,rruff said:I disabled Dragon Center and the main MSI process [don't recall the name] simply because they are unnecessary. TS and AB and Silent Option give you great control over speed, frequency, temperature, fan speed, noise, etc.
A cooling pad will help your CPU performance if you are thermal throttling, but you shouldn't be thermal throttling with a UV... not on a pure CPU test. Is your core undervolt really -.159 [rather than -.0159]? If so your UV is very similar to mine. If not, then you should set it to -.159 and make sure it is working in FIVR. I can run all cores at 4.3GHz indefinitely at ~69W. C20 scores are ~3360 [max 3390]. Since your CPU maxed at 86W I question whether any undervolt is being applied at all. With no UV mine is about that.Click to expand...
Yes the undervolt for the cpu core is -.157 amd cpu cache is -.059. Thought that maybe you meant also cache to -.159, chaged it and it went up straigth to a bsod haha fun times.
Disabled Dragon Center, but dont know about the process [there are many MSI processes] and also applied your TPL settings according to your screens on your first post.
How are you getting a max 69W if your TPL2 is set to 75W [which can be seen on the screens below].
Double checked and the only option different from yours is IccMax which I have on default. What it does I dont know so I didn't touch it [I should read about it], but don't think is the one keeping my max watt at 74.9.