Throttlestop overclocking Desktop PCs

[Retrorockit]

Myself and a few others have used Throttlestop software and unlocked CPUs to overclock OEM locked BIOS PCs.
Throttlestop was designed for underclocking, and undevolting laptop computers to extend battery life. It also allows full control of Voltage and multiplier on unlocked CPUs. There is a large TS Laptop thread, but I thought I'd make a separate discussion for PC owners so the information would be easier to find. TS overclocking is a work in progress. There is a lot that isn't known. Especially what TS can control on unlocked motherboards.

Throttlestop is a full featured program. It can modify Volts, and multiplier to the extent that they are unlocked. It displays these values and CPU speed, CPU core load, and temperatures in the main window. There is also a benchmark and mild stress test window for testing your results. Changes do not require a reboot to be applied. It also controls C state and other advanced features. You don't need to know your PLL chip to use it.

Her are a couple results (there aren't many so far).
Dell Dimension E520 QX6800 4GHZ CPUZ, 3.72GHz Firestrike.
Dell Precision T3400 QX9650 4.15GHz Firestrike.
Dell Optiplex 745 QX6800 3.45GHz
To be tested Dell Precision T3500 W3690 Xeon

If you want to try this you will need an OEM computer that supports an Extreme series CPU. Most of them have a 95W CPU limit and won't boot an Extreme chip. The exceptions are the old 65nm machines that supported Pentium4, and PentiumD CPUs. They don't have the 95W limit. The others without the limit are the workstation models.

There is a learning curve with doing this and I will do what I can here to help with that.
1- Throttlestop sometimes won't save setting unless you open the program twice.
2- Winring0dll sometimes goes missing. My solution was to drag the whole folder to the desktop. Then it works. Unclewebb the TS developer has another solution below.
3- Some benchmarks (Heaven) can post inverse results VS overclock. This happens if you've opened TS and minimized it. A restart (not reboot) fixes this. Some (Geekbench) won't let it run until you do this. GB also doesn't detect the OC.
4- You need to change some Windows settings. A high performance User Profile, and turning down Windows Security to minimum.
Firestrike doesn't seem to have any issues with TS.
5- The Lock /Unlock feature can be confusing at first. It's a button that "will do " what it says.
If it says UNLOCK the CPU "is" locked until you click there. If it says LOCK the CPU "is" unlocked. This will lock in your overclock setting, not return to the default Multiplier.
6- Reset FID VID on exit is on page 2, hit OPTIONS to view it. This will save your overclock. Be careful. If you leave this turned OFF ( easy to do since it's on page 2) you could save an unstable setting. A backup bootable HDD is nice to have if this happens.
7- The very highest Voltage setting (1.6V in my case) can result in sudden power loss. This doesn't allow the program to reset the VID. I would avoid this setting. People will be tempted to do this because TS makes it so easy to set Voltage high and Then tun it back down on the run

There are a few unanswered questions.
1- It's known that TS only lowers voltage on locked computers with locked CPUs. How it behaves on an unlocked MB is not known. Can it raise Voltage then?
2- With a VID pinmod set high, can TS lower Voltage to the desired setting ( this would save a lot of trial and error VID modding)?

The lack of knowledge is due to the fact that most TS users are running laptops, which closely resemble locked OEM computers in many ways.

A couple other tips- This method of overclocking is actually very simple. There are only 3 variables.
Voltage, cooling, and multiplier. Everything else remains unaffected.
My advice is to do as many cooling mods as you can first. This includes heatsinks on the VRM MOSFETs. This makes more Voltage available. Start at the CPU itself with good TIM, and give some serious thought to lapping the CPU and heatsink.
As far as voltage goes there's not a lot to do. I have sometimes seen huge gains by moving big OEM fans off of the MB header. This usually brings a BIOS fan fail message.
I suppose I'll get some TLDR replies on this!

 

[P4-630]

I'm sure the Author of TS @unclewebb could tell you more about it.

 

[Retrorockit]

I spoke to him before I posted this. He said post the thread and he agreed to participate as needed. I think the laptop side of things has kept him quite busy.

 

[unclewebb]

Throttlestop is a full featured program. It can modify FSB

3- Can TS raise the FSB on an unlocked MB?

That is one thing that ThrottleStop cannot do. You can use ThrottleStop to raise the multipliers on the unlocked Core 2 based Extreme (QX or X) processors but you cannot use ThrottleStop to change the FSB speed.

1- Throttlestop won't save setting unless you open the program twice.

You should not have to do this. You might have a problem like this if you run ThrottleStop in one folder and it creates a ThrottleStop.INI config file in that folder and then you drag the folder somewhere else on your hard drive. Newer versions of Windows can change the read / write file permissions when you move folders. This can turn off file write permissions on the ThrottleStop.INI file so ThrottleStop will be able to read from this file but it will no longer be able to write any changes back into this file. The best advice is to unzip the ThrottleStop folder and all of its contents to wherever you want it and leave it there. Do not move it around. I put this folder in C:\Program Files(x86). If you do need to move it around, make sure that the INI file still has read write permissions. If you do not know how to check that, just delete the INI file and let ThrottleStop create a new INI file which should have the proper permissions.

2- Winring0dll sometimes goes missing.

This is another one of those Windows safety features. It might also be hard drive speed related. If ThrottleStop is not able to load the WinRing0 file within a reasonable amount of time, Windows will have a fit and then might black list the WinRing0 files and prevent ThrottleStop from opening them up ever again. Very bizarre. You can also run into this problem if you try to start 2 different programs at about the same time that both try to open the WinRing0 driver. One program will likely succeed and the second program will fail. Some antivirus programs are also hesitant to let a person run any software that uses the WinRing0 files to access the CPU registers. I use Avast and reasonably modern platter hard drives and have not had this problem in a long time. RealTemp and ThrottleStop both use WinRing0. Some other monitoring programs also use it but the WinRing0 files are hidden. The final WinRing0 driver was released many years ago so there is not much one can do about it.

Another thing you have to watch out for is if you change the FSB speed while in Windows, this can screw up the accuracy of some of the high performance clocks that Windows and many other programs use to measure time. I wrote a separate program called WinTimerTester to check for this bug. It runs two of the main Windows system timers against each other. In theory, both timers should measure the same amount of elapsed time. If there is a problem, one timer will run much faster than the other one. These timers are used in a lot of benchmark type programs. If you start getting weird benchmark results when FSB overclocking, it would be a good idea to test your timers.

WinTimerTester 1.1
https://drive.google.com/file/d/0B0dpSo9k93jDZlZ6NWNlV0RTcUU/view?usp=sharing

http://forum.notebookreview.com/threads/setfsb-game-timing-problem-g60jx.568525/page-2#post-7364805

Using Hibernate or Stand By mode might let Windows re-calibrate these timers so they are both in sync like they should be.

Many companies are knee deep in old Dell workstations. Hopefully more hobbyists on a budget learn that you can build a cheap and reasonably speedy computer with this old hardware. I am still rocking my old QX9650 as I type.

I think the laptop side of things has kept him quite busy.

I promised the W1zzard that I would finally update RealTemp with more features and better support for the newer 7th Gen CPUs and Windows 10. Final testing is in progress. I borrowed the TS Bench and a few other goodies from ThrottleStop.

 

[Retrorockit]

Thank you for clarifying things. Someone reported FSB down clocking to me. I guess it wasn't true.
The start twice might only apply to the older TS6 version I used, and maybe just to first installation. I've had users report it won't save and this seemed to resolve that.
I knew my moving the folder to the Desktop wasn't the best solution, but it got me into the Dell overclocking game.
I get questions daily at Tomshardware about overclocking OEM computers and I thought a reliable resource for this technique would be useful. Also I'm curious if traditional FSB overclockers would find it useful also.
BTW W3690/W3680 6 core Xeons are unlocked, and if you buy workstations with them already installed you can get quite a deal.

 

[agent_x007]

LGA 775 Conroe/Wolfdale/Kentsfield/Yorkfield don't have a PCU (power control unit), add to that fully analog VRMs on MBs, and I'm pretty sure Overvolting is impossible (or requires a ASUS/Gigabyte specific program that can't work with OEM boards).
I assume OEM board doesn't have "hidden" Vcore control options.

You can downclock and/or decrease Vcore, because of C-stages (and SpeedStep), since those are supported.
Unlocked CPU's don't have locked higher than stock multipliers, that's why you can choose them in TS.

Locked OEM boards should support VID pins modding and *BSEL/tape mods (*limited to max FSB supported by board).

TDP of CPU doesn't matter. If BIOS has ucodes for it, CPU will boot.
With TS, you can prevent VRM from blowing up tho

 

[Retrorockit]

I used Throttlestop 6.00 to overclock my Dell Dimension E520 QX6800 to 4GHz. The CPUZ link is in my sig. You can check the Voltage there. You can raise Voltage on the unlocked CPUs. You can also see the 266 fsb.
TDP of CPU does matter. Chipsets have a TDP limit and won't run CPUs that are higher than rated. If you get them to boot then you can run more power through them. If you go to Delidded.com you will see various chipset limits posted there. That's why Q9650 will run in many OEM computers but not the QX9650. Most OEM MB with 1333fsb have 95W limit.
The only reason my E520 booted with a 130W chip is it supports 135W Pentium Ds.

 

[agent_x007]

Well, I never had a BTX board, so maybe things are different on them, BUT on ATX ones you can't raise the Vcore by 3-rd party software (at least I never heard of someone doing that).
Running 1,6V through QX6800 isn't good for MOSFETs on OEM boards
Still, I run 1,4V through Pentium 4E @ 3,53GHz on PGA 478 that uses three phases (six MOSFETs in total, but I added a heatsink to it's three chokes tho) : https://valid.x86.fr/gtfwfk

There is no such thing as "Chipset TDP limit", because it's VRM and BIOS that decide what can and can't be supported. Example of G31 chipset based board, compatible with 130W Quad Core, but not with Pentium D 130W TDP CPU's : LINK.
Again, I don't know if this applies to BTX OEM boards.

Not running QX9650 on OEM boards :
Anyone tried to manually inject ucodes for Extreme versions BEFORE installing them (ie. BIOS moding) ?
Official BIOS may simply not have a ucodes to recognize such CPUs.
From AMD side :
You can run (boot/use) 130W TDP Phenom II 965 on 95W TDP only MB (even on official BIOS).
You may damaged the VRM in worst case, but it is possible.

 

[Retrorockit]

I created this thread to let people know about this method. BTX makes no difference. That's the big thing about TS vs. the FSBware. It does Voltage.
1.6V was just for a validation run. If you click the image at CPUZ you can see the Voltage for 3.72GHz which is my benchmarking speed. The E520 has 4 phase VRM with very good airflow to them. 3,8GHz was a standard Pentium 4 speed.

You say it's possible. I've never seen it done. And I've been looking a lot! I did see Dells with 95W limit and G31 so you're right about it not being the chipset. But to compare a G31 to a Pentium 4 chipset is not relevant. There is no 1333fsb chipset that supports P4. They're 2 generations apart. The P965 supports 65nm Core2Dou, and P4. Some 945 boards got upgraded also. AMD is a whole unrelated topic.
I did see a Q9650 running 3.6GHz in an Optiplex 780 (95W) at userbenchmark.com so it looks like the fsb got unlocked. I think I know how it was done.

The microcode idea sounds like it's worth a try. It's possible they leave out the code for over 95W CPUs. As far as modding a Dell BIOS goes it's a tough nut to crack. But maybe microcode can be done. It would be nice if it works I have 2- QX9650s I can't use.

Here's my take on the QX9650 microcode mod for OEM computers. If anybody did try it and it ran, if they didn't know about Throttlestop then they would just have an expensive Q9650. So maybe it's been done and forgotten.

 

[Jstn7477]

I just tried the latest TS on my Dell Dimension E520 and QX6800 (and pretty sure I have before) and the multiplier selection doesn't go beyond 11. Am I missing something?

 

[agent_x007]

You say it's possible. I've never seen it done. And I've been looking a lot! I did see Dells with 95W limit and G31 so you're right about it not being the chipset. But to compare a G31 to a Pentium 4 chipset is not relevant. There is no 1333fsb chipset that supports P4. They're 2 generations apart. The P965 supports 65nm Core2Dou, and P4. Some 945 boards got upgraded also. AMD is a whole unrelated topic.

I give you GA-EP45-UD3P (LINK), a board with P45 chipset that supports both 1600MHz FSB/PCI-e 2.0 AND has official support for everything from Celeron D to Core 2 Extreme QX9770 (except 130nm Gallatin P4EEs).
There are also boards like ASRock 775i65G R3.0, that support 65nm Core 2 Quads on old 865G chipset (with slower FSB or 1066MHz with 400MHz CL2.5 DDR1) : LINK.
And I play around a 945G based board, with PGA 478 socket and PCI-e slot

PS. I usually use SetFSB for Windows FSB OC (a close enough PLL chip name is needed tho).
Altho I don't use it often (since I prefer BIOS overclocking, and I do not use OEM boards).

 

[Retrorockit]

I just tried the latest TS on my Dell Dimension E520 and QX6800 (and pretty sure I have before) and the multiplier selection doesn't go beyond 11. Am I missing something?

There is a setting for "C states off" that keeps it from downclocking, and gives more control to Throttlestop. Unclewebb disagrees but mine sometimes won't save until I open the program a 2nd time.

I give you GA-EP45-UD3P (LINK), a board with P45 chipset that supports both 1600MHz FSB/PCI-e 2.0 AND has official support for everything from Celeron D to Core 2 Extreme QX9770 (except 130nm Gallatin P4EEs).
There are also boards like ASRock 775i65G R3.0, that support 65nm Core 2 Quads on old 865G chipset (with slower FSB or 1066MHz with 400MHz CL2.5 DDR1) : LINK.
And I play around a 945G based board, with PGA 478 socket and PCI-e slot

PS. I usually use SetFSB for Windows FSB OC (a close enough PLL chip name is needed tho).
Altho I don't use it often (since I prefer BIOS overclocking, and I do not use OEM boards).

You seem to be looking for an argument. You were wrong about Voltage raising on unlocked CPUs and I may be wrong about the 95W limit. Please understand that I have specialized in overclocking locked BIOS computers, and especially BTX where there are no aftermarket parts. There is a lot I don't know about normal overclocking, and I'm always finding out something new in my area of interest. I like to do original work and not something that others have done dozens of times before. There may be many people who installed Core2Extremes in OEM computers (like myself with QX6700) and got nothing out of it. Most people don't post their failures. This is how you can succeed.
I'm actually more interested in the Dell T3400 BTX workstation with Locked BIOS, X38 chipset with hidden support for 400fsb, and 130W CPUs. On a 3 phase VRM BTW.
Most of what you're posting seems to be more about proving that you're an expert at something else, and not about the subject of this thread.
Many OEM computers like the E520 have a locked PLL chip and crash when SetFSB is applied.

 

[P4-630]

@Retrorockit learn to use the "Multi-Quote" button instead of double posting!

 

[agent_x007]

I play with everything other than OEMs (and BTX), so we actually have no common ground...
I saw you comment "searched a lot", but I haven't thought that you were talking about OEM only boards.
So is this program OEM only then, ie. no one used it for regular MBs ?
I'm asking this, because I'm interested in using this on ASRock 4CoreDual-SATA2 R2.0.

 

[Retrorockit]

@Retrorockit learn to use the "Multi-Quote" button instead of double posting!

This my first thread here so I'm learning. Different forums have different rules. Thanks.

I play with everything other than OEMs (and BTX), so we actually have no common ground...
I saw you comment "searched a lot", but I haven't thought that you were talking about OEM only boards.
So is this program OEM only then, ie. no one used it for regular MBs ?
I'm asking this, because I'm interested in using this on ASRock 4CoreDual-SATA2 R2.0.

I didn't make the thread OEM, or BTX because usually I just get trolled by people who don't care about such things. Also if someone were in a SetFSB, or VID mod situation it might be possible to VID mod too high and use TS to drop back down instead of a series of VID mods. Maybe BIOS mod too high and test on the run to get where you need to go. I don't know what it might be good for. I'm just letting people know it exists. It seems like a good companion to SetFSB, and with C2X it also works when SetFSB doesn't. The program was originated for undervolting laptops. The OC aspect is a side feature that's not well known. Laptops are basically OEM computers with even more limitations on the PSU side. Actually it's not been used very much for OEMs either.

The WG864 CPUZ postings here are myself, and another guy from Tomshardware who tried it also. He also did a Dell T3400 QX9650@4.15GHZ
http://valid.x86.fr/top-cpu/496e746...d652043505520513638303020204020322e393347487a
Someone posted a "Mission Impossible" At Reddit about overclocking a Dell Optiplex 745. I had it @ 3.45GHz in about 30 minutes using spare parts I had sitting around. Since it wasn't as good as the 4 phase 520 I didn't load test it.
If anyone wants to get an OEM to play with the Dimension 9200 is the same as the E520 but has room for a full size video card. The T3400 has room for 2 GPUs and supports 1333fsb. All Dells form this era have an issue with GCN3 not able to display the Dell BIOS resolution. This is all R9-3xx (and 285/380) GPUs and newer. Older HD7xxx are OK.

If you want to use it please do so. But realize you might be the first one doing it. So there is no forum or expert to hold your hand. I hope you'll post what you find here. What works, and what doesn't. Unclewebb said he added the TS testing suite to RealTemp 4.00

I was actually at the Gigabyte page you posted recently and told someone who owned that board that it supported all those CPUs and he insisted that it only supported up to QX6800.
We both ended up not sure.It's nice to know I was right. But the motherboard was dead so it didn't matter.

I just tried the latest TS on my Dell Dimension E520 and QX6800 (and pretty sure I have before) and the multiplier selection doesn't go beyond 11. Am I missing something?

You might try TS6. Throttlestop gets rewritten with every new family of CPUs, and probably operating systems also. TS6 is a lot simpler due to this.

I'm going to throw out some Dell part numbers you might need also.
D9729 Pentium4 PentiumD heatpipe cooler. Bolts right into the E520. This is the go to bolt in cooler for Optiplexes also.
T9303 Similar but different heatpipe layout and better airflow. Cools better. If you buy it with the cover you can cut the top 1/2" off and glue it onto the E520 cover and use it as a guide to move the holes.
Delta AFC1512DG, Dell#NC466, Dell# DG168 These are all the same Delta 150mm X 50mm 1.8A fan. Sometimes cheaper under the Dell numbers. This replaces the whole fan and housing assy. Velcro and a little duct tape will get it in. Plugs into the Dell 5 pin MB header. The original fan speeds up too late and gets noisy also. This is a workstation "quiet" fan.
GFB1212VHG Delta 2 motor 3.4A 8 wire fan. Fits in the stock housing. Obviously not running off of the MB header. Power leads to Molex 4 pin, Pair the PWMs to MB and run one rpm wire to the MB. There are 4 wire versions of this fan also (2+,2-). Throws fan fail message. Requires reset in BIOS to clear. This fan @ 100% (pull PWM wires) was how I got 4GHz.
Nidec TA350DC This is a 92mmX38mm fan. It comes in ratings from .4A up to 1.8A It controls speed by air outlet temperature. This could be put on the back of the D9729 to pull air though the close fins and can have the CPU heat ducted out the back separately.
You can expect 3.45GHz with the fan on the MB. This assumes a SLACP chip. B3 stepping is harder to work with.

1.45V stock, 1.5V with VRM cooling, 1.6V with fan power form Molex. 1.6V is the highest setting Intel wrote chipset calls for. It's usually unstable, 1.588V is max. useful.

You can browse my Retrorockit gallery at Overclock.net for pics. of some of this.

I just tried the latest TS on my Dell Dimension E520 and QX6800 (and pretty sure I have before) and the multiplier selection doesn't go beyond 11. Am I missing something?

There is one more thing. I've added it to the Learning Curve section. The LOCK, UNLOCK button can be confusing at first. If it says UNLOCK the CPU is locked until you click it. Then it will say LOCK and the settings will be "unlocked" until you lock them by clicking there again.

 

[Retrorockit]

I've added couple more cautions to the learning curve section. One of them is a known issue. The other one is based on my own experience with this.

 

[Jstn7477]

My mistake was that I didn't press UNLK to unlock the voltage and multiplier. I can do so now, but my remote machine froze at 12x and the default 1.337v so I'll have to play around with it later when I have access to it. My QX6800 is a G0 core, by the way.

 

[Retrorockit]

My E520 maxed out at 1.45V. w/o mods. i would set that first and see how far you get. Then back down while load testing for stability. I would expect 3.45GHz stable w/o other mods. Do you have one of the heatpipe coolers. The stock low profile copper base finsink isn't much good for overclocking. It will run the SLACP at base clock just fine.

 

[unclewebb]

but my remote machine froze at 12x and the default 1.337v

When overclocking these CPUs using ThrottleStop, it is best to start with lots of voltage before adjusting the multiplier. If you start adjusting the multiplier and the CPU does not have adequate voltage, you are going to be seeing a lot of BSOD screens. After you do some stability testing at the higher MHz, then you can go back and see if you can still run reliably with less voltage.

The other thing to keep in mind is if you unlock the multiplier and increase it and then change your mind and want to go back to a lower value, you need to first push the ThrottleStop Reset button to reset the maximum CPU multiplier to its default value. After you Reset, then you can increase the multiplier to whatever value you like.

Some motherboards like my Asus board completely ignores the VID voltage setting in ThrottleStop. On this motherboard, the voltage can only be adjusted within the bios.

 

[Retrorockit]

I like testing with Prime95 because it just drops threads when it detects errors instead of BSOD. I suppose each CPU, and motherboard will respond a little differently. There's a lot that isn't known about this yet.

 

[Jstn7477]

Yep, it's been a while since I've overclocked the older Core 2 chips, and my dual core X6800 topped out around 3.4GHz last time I tried, but that was on a refurbished and voltage locked ASRock 4CoreDual-SATA2 that I bought 5 years ago for $25 on Geeks.com and it finally died in April. I'm not too concerned about overclocking this 10yo PC since the chip is already power hungry enough as it is, but it's pretty neat that I can and I'd be happy even with just 12x and ~1.4v on the stock cooling. I already have a nice C2Q setup I acquired four years ago for $200 (Q9550 @ 3.65GHz, ASUS Maximus II Formula P45 and 8GB Corsair Dominator DDR2-1066) and my other working C2D machine has an ASRock 775i65G R3.0 I just had to replace a few caps on last week, PDC E5800 @ 3.75GHz, Sapphire 3850 AGP single slot and 2GB of decent vintage DDR-400 that I still need to OC once I find what voltage that board is giving it (no BIOS option).

 

[Retrorockit]

A question came up about overclocking unlocked i7, and unlocked Xeons (W3690, W3680).
The procedure is to set the multiplier to Base Clock +1 This applies maximum turbo boost. Then look for the Turbo Power Limits window and set the limit higher. This should be possible with an unlocked CPU unless the BIOS is programmed to block this. Since very few have tried this yet which systems will or won't respond are unknown. If you try this and it does or doesn't work please post here to add to the knowledge base.

Yep, it's been a while since I've overclocked the older Core 2 chips, and my dual core X6800 topped out around 3.4GHz last time I tried, but that was on a refurbished and voltage locked ASRock 4CoreDual-SATA2 that I bought 5 years ago for $25 on Geeks.com and it finally died in April. I'm not too concerned about overclocking this 10yo PC since the chip is already power hungry enough as it is, but it's pretty neat that I can and I'd be happy even with just 12x and ~1.4v on the stock cooling. I already have a nice C2Q setup I acquired four years ago for $200 (Q9550 @ 3.65GHz, ASUS Maximus II Formula P45 and 8GB Corsair Dominator DDR2-1066) and my other working C2D machine has an ASRock 775i65G R3.0 I just had to replace a few caps on last week, PDC E5800 @ 3.75GHz, Sapphire 3850 AGP single slot and 2GB of decent vintage DDR-400 that I still need to OC once I find what voltage that board is giving it (no BIOS option).[/QUOTE

At CPUz 3.45GHz is the typical overclock on aftermarket MBs. The E520 can go much higher. The steps in this method with 1066fsb 65nm Extreme CPUs are 1 full multiplier. So it's 3.72GHz or bust. Powering the fan off of the PSU, and a bigger fan will be needed. But 3.45 is respectable and easy. You could cut the PWM wire to the stock fan for testing @ 100% speed, but you won't want to be in the same room with it.

 

[agent_x007]

With VID mod I overclocked QX9770 to ~3,8GHz stable on my 4CoreDual-SATA2 R2.0.
Here's a Valid of not stable OC : http://valid.x86.fr/ij08g9

 

[Retrorockit]

With VID mod I overclocked QX9770 to ~3,8GHz stable on my 4CoreDual-SATA2 R2.0.
Here's a Valid of not stable OC : https://valid.x86.fr/ij08g

The link didn't work for me. Did you use TS at all? TS should control Volts also on a C2X. QX9650 can run 4.15GHz stable on a Dell T3400 with TS only. I think there's more in it.

 

[agent_x007]

The link didn't work for me. Did you use TS at all? TS should control Volts also on a C2X. QX9650 can run 4.15GHz stable on a Dell T3400 with TS only. I think there's more in it.

FIxed : http://valid.x86.fr/ij08g9
No TR, just tinfoil (it changes CPU VID from 1,2875V to 1,3875V).
ASRock board supports BIOS multiplier change, no Vcore change support.