Throttlestop overclocking Desktop PCs

[frankr2994]

OK for mine when I done the voltage to 2048 before the scewin together with unlimited tdp worked fine. I cannot say if for the Overclocking unlock scewin still interferes somehow since I updated the bios without reinstalling windows by can test it. To be clear if you don't follow the process of the battery and bios jumper even though the new bios is written it is not loading so you get the previous copy operationally but when you read it is as per update.

I'd love to know where the old version is stored to stay loaded since when you program with an external programmer, the first step is to completely wipe the firmware

Regardless, I believe that if you are using Ami BCP or scewin that you are either changing the question default or the selected option, any dependencies on those choices are still in place, regardless of which way you go. Amibcp was originally used by a lot of people to unhide options. With the graphical interface of the Dell bios, that seems to be impossible. I have had a few people try for me and write me modified. Bioses and replace modules to try and get stuff to show it never worked

Before I came across scewin I was able to hex edit the Intel RC setup module and make changes after flash. However, that was a lot of trial and error and a ton of bricked bioses

 

[sk111]

I'd love to know where the old version is stored to stay loaded since when you program with an external programmer, the first step is to completely wipe the firmware

Regardless, I believe that if you are using Ami BCP or scewin that you are either changing the question default or the selected option, any dependencies on those choices are still in place, regardless of which way you go. Amibcp was originally used by a lot of people to unhide options. With the graphical interface of the Dell bios, that seems to be impossible. I have had a few people try for me and write me modified. Bioses and replace modules to try and get stuff to show it never worked

I have read this process in the dell t5810 manual. It is stored in the CMOS. Could you share one of these modules? I would love to take a look "just" in case. In effect the scewin is exposing all the parameters of the amibcp and I did notice for things I could not make work in amibcp I could not change in scewin either. For example I changed the core offset voltage in both amibcp and scewin to a test of +20 mv. That didn't show up in hw monitor. As soon as I launched TS hw monitor showed immediately the offset voltage of +150 mv I had set there without sleep cycle.

 

[frankr2994]

I have read this process in the dell t5810 manual. It is stored in the CMOS. Could you share one of these modules? I would love to take a look "just" in case. In effect the scewin is exposing all the parameters of the amibcp and I did notice for things I could not make work in amibcp I could not change in scewin either. For example I changed the core offset voltage in both amibcp and scewin to a test of +20 mv. That didn't show up in hw monitor. As soon as I launched TS hw monitor showed immediately the offset voltage of +150 mv I had set there without sleep cycle.

One of my hex edited modules? I really don't think I have any of that stuff lingering around. That was a while ago.

I still believe any settings not taking affect are due to other dependencies not being met. Like there is a setting called overclock feature. I believe this was to be used together with an xtu efi module. You can find this module in supermicro boards (who I believe builds the Dell precision boards) I think after you have all that on place you can start making those changes that aren't working.

But in all honesty I had about 2 years in playing with this stuff on my Dell and finally moved on. Most of the information you'll find out there is either from me or from one of the couple ppl that asked for me to help lol. I recently got sucked into another rabbit hole of trying to enable above 4g decoding on my gigabyte board which doesn't have the option at all. I called it quits last night as I cannot figure out how to implement a new variable

 

[sk111]

One of my hex edited modules? I really don't think I have any of that stuff lingering around. That was a while ago.

I still believe any settings not taking affect are due to other dependencies not being met. Like there is a setting called overclock feature. I believe this was to be used together with an xtu efi module. You can find this module in supermicro boards (who I believe builds the Dell precision boards) I think after you have all that on place you can start making those changes that aren't working.

But in all honesty I had about 2 years in playing with this stuff on my Dell and finally moved on. Most of the information you'll find out there is either from me or from one of the couple ppl that asked for me to help lol. I recently got sucked into another rabbit hole of trying to enable above 4g decoding on my gigabyte board which doesn't have the option at all. I called it quits last night as I cannot figure out how to implement a new variable

True indeed not need to spend much time on this. I got already more than I thought possible with that machine. Top overclock, overclocking memory etc.
For the 4G decoding in dell works fine lol and you can even implement the rebar in bios (I have done but for my gpu was stuttering so dropped it)

 

[frankr2994]

True indeed not need to spend much time on this. I got already more than I thought possible with that machine. Top overclock, overclocking memory etc.
For the 4G decoding in dell works fine lol and you can even implement the rebar in bios (I have done but for my gpu was stuttering so dropped it)

Ya I know Dell has it....pisses me off. Might have to turn on my 5810 or 7810 again

 

[sk111]

Ya I know Dell has it....pisses me off. Might have to turn on my 5810 or 7810 again

Ya I know Dell has it....pisses me off. Might have to turn on my 5810 or 7810 again

It's never too late

It's never too late

By the way here is the settings I found in amibcp the current and power limits I did prior to scewin (first and third) and the rest after. The one before the last one is the attempt to change the offset as a test which did not work and not sure why though maybe the overclocking feature needs to be set to enabled which i had disabled (in the last one)... one more attempt at some point. If you got the machine feel free to try.

 

[sk111]

It's never too late


By the way here is the settings I found in amibcp the current and power limits I did prior to scewin (first and third) and the rest after. The one before the last one is the attempt to change the offset as a test which did not work and not sure why though maybe the overclocking feature needs to be set to enabled which i had disabled (in the last one)... one more attempt at some point. If you got the machine feel free to try.
View attachment 371581
View attachment 371582View attachment 371583
View attachment 371584View attachment 371585

One more thing and to follow on your previous work either with modifying the bios or using scewin after the processor is unlocked (and ideally the current too) you can use xtu but that would mean having to reload it at the startup. I came across the following script in powershell that solves this problem and automatically loads with Windows. As a matter of fact you can add the extra parameters "-WindowStyle hidden" without quotes in task scheduler add arguments and will happen completely in the background.

GitHub - Jackman1506/TaskScheduledXTU: A repository to allow users to automatically apply XTU profiles on boot with the use of Powershell.

A repository to allow users to automatically apply XTU profiles on boot with the use of Powershell. - Jackman1506/TaskScheduledXTU

github.com

 

[frankr2994]

One more thing and to follow on your previous work either with modifying the bios or using scewin after the processor is unlocked (and ideally the current too) you can use xtu but that would mean having to reload it at the startup. I came across the following script in powershell that solves this problem and automatically loads with Windows. As a matter of fact you can add the extra parameters "-WindowStyle hidden" without quotes in task scheduler add arguments and will happen completely in the background.

GitHub - Jackman1506/TaskScheduledXTU: A repository to allow users to automatically apply XTU profiles on boot with the use of Powershell.

A repository to allow users to automatically apply XTU profiles on boot with the use of Powershell. - Jackman1506/TaskScheduledXTU

github.com

Now that's a good find!

I have my kids 5810 all setup with an oc but half the time he forgets to turn it on lol. I was going to at some point put my supermicro cooler on his machine so I can crank his 1650 to the moon.

With that cooler 4.8ghz should run all day long.

Don't know if it's still the case but in that machine with the fans cranked to 100 percent I held the record on user benchmark with that CPU at 4.8. probably 3 years ago or something

 

[sk111]

Now that's a good find!

I have my kids 5810 all setup with an oc but half the time he forgets to turn it on lol. I was going to at some point put my supermicro cooler on his machine so I can crank his 1650 to the moon.

With that cooler 4.8ghz should run all day long.

Don't know if it's still the case but in that machine with the fans cranked to 100 percent I held the record on user benchmark with that CPU at 4.8. probably 3 years ago or something

Definitely. Works great for me.

Definitely. Works great for me.

The record? On a dell t5810? I am impressed. By the way conscious of the fact that the current is unlocked. Do you have any safe controls in xtu when you push higher voltages?

 

[frankr2994]

Definitely. Works great for me.


The record? On a dell t5810? I am impressed. By the way conscious of the fact that the current is unlocked. Do you have any safe controls in xtu when you push higher voltages?

If I set current to unlimited I found it I set pl1/pl2 wattage to 185 and 205 I never trip the PSU rail. Any higher and it will trip sporadically.

So I have to force a throttle unfortunately

 

[sk111]

If I set current to unlimited I found it I set pl1/pl2 wattage to 185 and 205 I never trip the PSU rail. Any higher and it will trip sporadically.

So I have to force a throttle unfortunately

Nice thanks. If it trips it justs shuts down?

 

[frankr2994]

Yep just does a forced reboot

 

[sk111]

Yep just does a forced reboot

OK. So far I play with offset voltage (that applies only when it torboes) and I leave the main voltage at default. But will see if constant voltage has any benefits.

 

[frankr2994]

OK. So far I play with offset voltage (that applies only when it torboes) and I leave the main voltage at default. But will see if constant voltage has any benefits.

I just start with 1.4v set on adaptive. Set the clock up until it's unstable, back off til stable then dial voltage back again till unstable. Set the voltage back up a touch.

I saw no point in playing with offset or setting voltage fixed. Vccin is already set to 1.8v default and I never saw any improvement to increasing this.

But again with a stock cooler I'm not able to approach anything I just said lol. To keep temps under control on my kids right now I have that CPU set to 4.2ghz but it will run 4.8 at something like 1.37v.

 

[sk111]

I just start with 1.4v set on adaptive. Set the clock up until it's unstable, back off til stable then dial voltage back again till unstable. Set the voltage back up a touch.

I saw no point in playing with offset or setting voltage fixed. Vccin is already set to 1.8v default and I never saw any improvement to increasing this.

But again with a stock cooler I'm not able to approach anything I just said lol. To keep temps under control on my kids right now I have that CPU set to 4.2ghz but it will run 4.8 at something like 1.37v.

OK will try this and see if any better. Is there any easy modification on the cooler like adding extra fan without having to buy new one? I am looking for the 1660 v3 one. Currently I have it at 4.3 stable (with the offset method) and temps are very contained up to 80 C (85 under 100 load) - have reached 4.5 but backed off as needs better cooling. Your record at 4.8 was it with the 1650 or 1660?

 

[frankr2994]

OK will try this and see if any better. Is there any easy modification on the cooler like adding extra fan without having to buy new one? I am looking for the 1660 v3 one. Currently I have it at 4.3 stable (with the offset method) and temps are very contained up to 80 C (85 under 100 load) - have reached 4.5 but backed off as needs better cooling. Your record at 4.8 was it with the 1650 or 1660?

Record was with a 1650, my 1680 was stabile for about a year at 4.75 but I'm down around 4.6 now. Guessing high voltage degraded it.

Nothing simple you can do with a cooler that I know of. Narrow ilm is the devil. My cooler was sort of a complex job that doesn't really fit it the case lol

 

[sk111]

From what I see some Noctua heatsinks have an extra fan and same number of heatpipes (3). I will give it a shot by attaching an extra fan.

 

[frankr2994]

From what I see some Noctua heatsinks have an extra fan and same number of heatpipes (3). I will give it a shot by attaching an extra fan.

If you yank your 5810 sink you'll see it's pretty poorly made. Heatpipes don't make good contact, big gaps around them that thermal paste displaces into. The sinks they made for the 7910 are designed WAAAY better. Unfortunately square ilm.

You can try adding some more fans. Find an 80mm something high pressure fan.

 

[sk111]

If you yank your 5810 sink you'll see it's pretty poorly made. Heatpipes don't make good contact, big gaps around them that thermal paste displaces into. The sinks they made for the 7910 are designed WAAAY better. Unfortunately square ilm.

You can try adding some more fans. Find an 80mm something high pressure fan.

I ordered a nice fan. Let's see if I am able to fit it on.
By the way I found a part that could fit the vrm.

 

[sk111]

I ordered a nice fan. Let's see if I am able to fit it on.
By the way I found a part that could fit the vrm.
View attachment 372161

An update on this. The alienware vrm cooling was excellent. Didn't fit the holes on the board but created a riser and attached it. Then as a fan I used a dell t5810 front case fan directly attached on the cooler..
Results are great on 1660v3. 4.6 Ghz oc passes the xtu stress test at 1.31v. Also 4.7Ghz @ 1.33 passes the cpu z stress test which is lighter.. 4.8Ghz at 1.34 just does the benchmark and crashes on the stresstest meaning can be fine tuned with more voltage. The fan keeps the temps below 80C at the stress test with tdp limit 185 watt.

 

[Retrorockit]

Something I found (like all my stuff a very long time ago) is the Dell CPU fans ran off of the same power rail as the CPU. I suppose you guys could check this with a Voltmeter. Getting the stock fan (1.6A. in my instance) off of the MB header did 2 things.
It reduced the Volatge I needed to get the same overclock, and it increased the Voltage available from the VRM for higher overclocks. This was by the same amount in both directions. If you just move the fan you will probably get a fan fail message in the BIOS that needs a manual reset. A small 2 wire chiset fan to the fan header stops this. The RPM and PWM wires can stay on the CPU fan.
I was running a 2 motor Delta fan GFB1212VHG with 3.4A max. current. So this mod wasn't really optional. But the increase was very real.
Whether this is still true or not IDK. A lot of what I did was BTX stuff, so it may have been specific to that, or just a Dell thing in general?????
But they do make that sort of fan in 90mm.

DELTA ELECTRONICS GFB0912HHG DIMENSIONS AND INSTALLATION INFORMATION Pdf Download

View and Download Delta Electronics GFB0912HHG dimensions and installation information online. GFB 92 x 92 x 50.8 MM SERIES Exhaust Fan. GFB0912HHG fan pdf manual download. Also for: Gfb0912shg, Gfb0912vhg, Gfb0924hhg, Gfb0924vhg, Gfb0948hhg, Gfb0948vhg.

www.manualslib.com

 

[frankr2994]

Mb power on the 5810 is separate from the CPU rail.

Per my math my PC was power cycling right at the 18 amp rail limit (spikes around 210 watts). And when that was happening I had 4 92mm system fans running full tilt off the board.

 

[sk111]

Mb power on the 5810 is separate from the CPU rail.

Per my math my PC was power cycling right at the 18 amp rail limit (spikes around 210 watts). And when that was happening I had 4 92mm system fans running full tilt off the board.

I also have the cpu fan and the 3 fans. Now what I done is unplug one of the front fans and bought another dell front case fan which I plugged in the plug of the front case fan I unplugged. Then strapped that extra fan on the CPU..So my cooler in effect has the internal fan plus the extra fan I added. Now the beauty with this setup is that I have massive airflow directly on the cpu and it speeds up with the power requirement along the main fans so as soon as the temps start rising this fan increases speed too as the 2 remaining front fans. The CFM of this fan is a massive 110 at max... so lots of airflow on the CPU. Frank I kind of copied your beefed up setup but instead larger cooler I put a larger fan. Keeps temps below 80C at 185watt power limit throughout the stress test. To add I got quite lucky with my 1660v3 cpu so can overclock it at lower voltage and so less wattage is needed. And the idea is to have high single core speed for gaming 500-520 in cpu z and when avx instructions hit in stuff like video editing etc then this is limited by the 185watt tdp limit.in effect reducing the some of the core clocks to stay within the limit... something I don't particularly care very much as the idea is to max for gaming.

 

[:D:D]

Hey I'm going to post in here even though it's a laptop lol. So I got a Dell e6540 with an i7-4610m. It has a partially unlocked multiplier and can go 4.0ghz all core 4.1 single core. But as much as I've pissed around with TS in the past...I'm getting basically a frozen windows after going to strong with the settings.

Did you disable internal VR protection? Had the previous version 4600m OC'd to 4.8GHz. Not much use for 24/7 but fun to beat desktop CPU's on LN2 and take global first place with XTU. OC'd 4700MQ too which saw it compete with 4930mx and beating it in some benches but degraded after some time. Can't help think while the 4930mx didn't reach a lot higher clocks it probably had better silicon, less imperfections to help degradation. The BGA to LGA Haswell CPU's were fun too as they could be run on desktop boards and allowed bclk straps to which although were shown as supported never seemed to work on mobile chipset. At least not that I have seen

TS came out of laptop undervolting to start with. The OC stuff got buried in the laptop threads, and ignored in the major OC forums. So I started this dedicated thread. So even though it's laptop TS overclocking it works here. it got mentioned in the laptop threads but lost in the thousands of pages of UV stuff.

I think originally it wasn't biased to laptops and targeted at throttling through modulation but maybe I remember it wrong. Perhaps it was snubbed by those desktop users that only recognize BIOS OC? While there are some things that need to be done before BIOS POST there's still a lot of other things that can be achieved in software.

I recently got sucked into another rabbit hole of trying to enable above 4g decoding on my gigabyte board which doesn't have the option at all. I called it quits last night as I cannot figure out how to implement a new variable

Above 4G isn't a HW setting but just tells the BIOS it can allocate addresses above 4GB. For my X99 I run BIOS with above 4G disabled and reallocate GPU BAR1 above after BIOS POST if wanting to use a large resizable BAR. Also used a legacy BIOS without such support https://www.techpowerup.com/forums/...-resizable-bar-capability.316443/post-5153102

 

[sk111]

I also have the cpu fan and the 3 fans. Now what I done is unplug one of the front fans and bought another dell front case fan which I plugged in the plug of the front case fan I unplugged. Then strapped that extra fan on the CPU..So my cooler in effect has the internal fan plus the extra fan I added. Now the beauty with this setup is that I have massive airflow directly on the cpu and it speeds up with the power requirement along the main fans so as soon as the temps start rising this fan increases speed too as the 2 remaining front fans. The CFM of this fan is a massive 110 at max... so lots of airflow on the CPU. Frank I kind of copied your beefed up setup but instead larger cooler I put a larger fan. Keeps temps below 80C at 185watt power limit throughout the stress test. To add I got quite lucky with my 1660v3 cpu so can overclock it at lower voltage and so less wattage is needed. And the idea is to have high single core speed for gaming 500-520 in cpu z and when avx instructions hit in stuff like video editing etc then this is limited by the 185watt tdp limit.in effect reducing the some of the core clocks to stay within the limit... something I don't particularly care very much as the idea is to max for gaming.

Here is the CPU-z results with 4.7G. https://valid.x86.fr/5hctu8. As mentioned high single cores and multicore limited to around 4.55G average at 100% usage from the TDP limit imposed. 4.7 is not very stable at this voltage neither I want to push it higher for too long. So 4.6G @ 1.31V is the sweet spot for me and that is where I operate it on daily function. Gaming averages around 100-120 watt so not even close to full 185 TDP.