Throttlestop overclocking Desktop PCs

[frankr2994]

Dell DCCU (Distributed Client Configuration Utility) has an option to allow previous BIOS versions to write. Some BIOS modders use this utility to install their modded files. But screwing around with the BIOS is always risky. Especially based on advice from me. So it depends on how risk averse this user is. On my antique stuff it's no big deal to brick something.

I'd love to know if dccu would be able to write modded files to a 3610. However because I have actually never been successful is any bios mod I'm not about to walk someone else through it....if I ever actually get something done I'll write all about it and someone else can read and attempt to follow lol.

 

[Retrorockit]

In BIOS Mods.com somewhere there is an explanation of the algorithm Dell uses for the checksum. DCCU, since it can write BIOS over a network seems to sidestep some local security features.
My understanding of BIOS modding is very poor. I've used DDCU to load the same BIOS version I already had with a Xeon CPU family code. So not a big mod. I've used it but I basically have no idea how it works, or what it is capable of. I didn't want the latest modded BIOS because Dell started digital signing at that point. What you're working on will already have that. I can only point you to look in that direction.

 

[frankr2994]

Well I've been able to load a "modded" bios like 20 times with my external programmer. Not a huge deal for me.....it is a pain but whatever. The bitch is I have never successfully modded anything. Even with the assistance of people from bios mods or win raid. Dell has so many defaulted values in their code you can change the world in it and somehow change nothing. It's rather impressive.if I was just swapping an efi module out within the bios that's no sweat.

 

[Retrorockit]

Well, you might be very wrong...as it is strangely very efficient, Like 60~70°C on a full load of 105W CPU.


It is the bracket, check picture!

When testing my BTX heatsink with the TS undervolted Q9505s I forgot to turn my 150x50mm fan in my Opti 380 up to 2500rpm.
Hottest core was 42*C. Sometimes they were all 41*C.
I may actually be able to TS overclock that CPU (sort of). I'll have to fire up a different computer, but it would be fun.

 

[frankr2994]

Well with the t3610 we have not been able to lift the multiplier lock. I sent a DM to a user on here that was able to overclock 6 of the 8 cores with his. Hopefully they respond to me but they only have a couple post on this forum. Also found a YouTube video showing someone got an overclock and all I could find was in the comments that they didn't do anything special. Posted specs show bios revision A03. Current bios for this machine is A19.

 

[KLiKzg]

When testing my BTX heatsink with the TS undervolted Q9505s I forgot to turn my 150x50mm fan in my Opti 380 up to 2500rpm.
Hottest core was 42*C. Sometimes they were all 41*C.
I may actually be able to TS overclock that CPU (sort of). I'll have to fire up a different computer, but it would be fun.

My temps are from BOINC calcs turned on 100%.

Only bad thing is that Q9550S does not work on Windows 10. So I run those on older Ubuntu.
But if you put X3360, this one works with Windows 10.

 

[Retrorockit]

IDK why Q9505s won't work in WIN10? I think I stuck my spare one in an Opti XE and it runs 7 on one HDD and 10 on another.

Please click the green human to continue - UserBenchmark

Here's the theory on TS OC for the Q9505s.
My Q9505s had 1.250V. stock. It runs just fine at 1.100V. It might run even lower but that's the lowest setting available in TS. So there is some headroom for an OC.
Enter the Dell T3400 BTX workstation. Locked BIOS, locked CPU. X38 chipset. 375W PSU with GPU 6 pin. 130W heatpipe cooler. ( Bigger 2x6 pin PSU available)
The X38 has undocumented support for 400fsb. A tapemod (different pin than the Q6600 tapemod) should access that.
Set multplier in TS to 7x400fsb. Should boot at 2.8Ghz with no problem. Now start working the Multipler up with TS and see if you can get 3.4Ghz (8.5x400) This would beat an X5470 except for the Q9505s 6M cache. The Q9550s is kind of pricey. I think the unlocked QX9650 was cheaper last time I looked. You might be able to sell the Q9505s and get one of those???
If it can run at full speed on stock Voltage then the tapemod would work in Linux also, but you would need a 400fsb capable chipset (X38,X48). Maybe try that first? You could try this method with an E8600 2 core and see how far you can go with that CPU if you want.

 

[frankr2994]

Hey was wondering if this is any sort of good practice? Can I rely on single core testing for voltage stability? Like set voltage up high enough to run 4.5 or 4.6ghz and just let the CPU throttle down below the amp limit when all the cores are up? That way anything single threaded could have some more speed. I don't recall ever getting different clocks based on cores in the fivr screen actually working.

 

[lexluthermiester]

Only bad thing is that Q9550S does not work on Windows 10.

What? Sure it does, it'll even run Windows 11. I've actually tested this in one of my own machines.

 

[unclewebb]

I don't recall ever getting different clocks based on cores in the fivr screen actually working.

Check if the C states are enabled. Intel Turbo Boost does not work correctly if the C states are disabled.

Here is a test of a 10850K with the turbo ratios set to their default values.

While running a TS Bench 1 Thread test with the C states enabled, it is easy to see that the CPU is using the correct 52 multiplier the vast majority of the time.

When the core C states are disabled, now the CPU is limited to the 48 multiplier during the exact same TS Bench 1 Thread test.

All Core i based CPUs produced since 2008 work the same way.

Only if you have an unlocked K series CPU and you are using the same multiplier no matter how many cores are active, then having the C states enabled or disabled does not matter. At default settings, most Core i CPUs use staggered turbo ratios that decrease as the number of active cores increases. Any Core i CPU that uses staggered turbo ratios has to have at least core C3 enabled so the CPU can boost up to the highest turbo multiplier when lightly loaded.

 

[frankr2994]

Check if the C states are enabled. Intel Turbo Boost does not work correctly if the C states are disabled.

Here is a test of a 10850K with the turbo ratios set to their default values.

View attachment 277947

While running a TS Bench 1 Thread test with the C states enabled, it is easy to see that the CPU is using the correct 52 multiplier the vast majority of the time.

View attachment 277948

When the core C states are disabled, now the CPU is limited to the 48 multiplier during the exact same TS Bench 1 Thread test.

View attachment 277949

All Core i based CPUs produced since 2008 work the same way.

Only if you have an unlocked K series CPU and you are using the same multiplier no matter how many cores are active, then having the C states enabled or disabled does not matter. At default settings, most Core i CPUs use staggered turbo ratios that decrease as the number of active cores increases. Any Core i CPU that uses staggered turbo ratios has to have at least core C3 enabled so the CPU can boost up to the highest turbo multiplier when lightly loaded.

Ya I have an e5-1680v3 so it's like a k series CPU. Whatever I set 1 core in fivr is what all my cores can run. I have to also use the set multiplier option on the main page. The highest number on set multiplier is the highest number in fivr. if I don't use fivr and set multiplier together then I cannot get an overclock. So I've accepted that's how that works. I was really curious on can I tell that a voltage is stable based on single thread stressing. For instance it appears that I can run 4.5ghz single threaded at 1.27v. if I push a workload that is multi threaded I'll hit my amp limit and it will automatically clock down core speed and voltage to be under the limit. Which all core stable is 4.2. so my hope would be 1.27v at 4.5 single and 1.21v 4.2 all core .

Maybe I can do this...tried TS bench and cinebench r23. Single threaded and multi threaded seemed great. Single rode 4.5ghz and multi ran 4.3...which is higher than what I've been able to get. Ran 4.3ghz for 30min on R23. Figured what the hell so I started up F@H and it maxed out at 4.3 for about 5 minutes and crashed. So I think I may have to set my voltage higher to force the CPU to clock down to 4.2 instead of 4.3. another surprising thing was TS bench 16 threads ran at 4.5ghz...made way more heat than cinebench. Max temp was 81c. Started to make my cooling solution not so great. Homemade vrm heatsink also got...hot. don't know how hot because I haven't invested in an IR thermometer for at home but the heatsink fins could damn near burn me. Even the chokes and the caps for the vrm got quite hot.

 

[KLiKzg]

What? Sure it does, it'll even run Windows 11. I've actually tested this in one of my own machines.

No, it will not - without the mod.

So, what did you change?

 

[frankr2994]

No, it will not - without the mod.

So, what did you change?

Most of the links are dead. But at the most all you would need with the install is the F6 Sata or chipset drivers for win 7 64bit. They were available from Intel but those links are dead. I'm sure if I put more than the little effort I just did I'd find them.

 

[Retrorockit]

I suppose a Dell T3400 might be hard to find in Zagreb.

 

[lexluthermiester]

No, it will not - without the mod.

So, what did you change?

Yes it will and DOES. I have a Lenovo desktop PC running a Core2Quad. It currently is dual booting Windows XP AND Windows 11. Runs perfectly. If it'll run Windows 11, it'll run Windows 10. If you can't install Windows 10 on that Q9550, you're doing something wrong.

 

[frankr2994]

I can report that you can majorly piss off a t5810 with overclock. From what I was trying the other day I had the multiplier set at 4.5ghz with my 1680 and voltage cranked a over 1.27. when I booted up last night I forgot folding at home was going to start up automatically. So when I opened up TS it started the overclock but powercut cannot auto enable without me going into fivr. So it set every throttle flag all at once and hard locked my core speed to 800mhz. I got F@h stopped but ultimately had to restart the PC to fix it. Wattage reading got to 185w which is above PL2 so I'm guessing that's what made it go nuts. Briefly thought I ruined everything lol.

 

[KLiKzg]

Yes it will and DOES. I have a Lenovo desktop PC running a Core2Quad. It currently is dual booting Windows XP AND Windows 11. Runs perfectly. If it'll run Windows 11, it'll run Windows 10. If you can't install Windows 10 on that Q9550, you're doing something wrong.

Well, then explain this: https://www.intel.com/content/www/us/en/support/articles/000006105/processors.html

 

[frankr2994]

Well, then explain this: https://www.intel.com/content/www/us/en/support/articles/000006105/processors.html

View attachment 278162

Doesn't officially support it. And why would they. How many people really care . Now the 8th gen and newer official support for 11 is a little upsetting. But since win 10s release people have basically been able to install it on damn near everything. Maybe you end up with a bunch of unknown devices in device manager but most of that could be taken care of with a little looking around too. Also there have been workarounds for win 11 on older stuff since the beginning. I noticed recently you can make your install media with Rufus to bake the workarounds in.

 

[lexluthermiester]

I can report that you can majorly piss off a t5810 with overclock. From what I was trying the other day I had the multiplier set at 4.5ghz with my 1680 and voltage cranked a over 1.27.

Did you make sure to increase the power limits in the TPL screen? You have to bump the power limits when OCing by more than 10ish%.

Well, then explain this: https://www.intel.com/content/www/us/en/support/articles/000006105/processors.html

View attachment 278162

I will answer with the following;

There you go. Just because Intel, microsoft or anyone else says it's not "compatible" does NOT mean it will not work and run perfectly. It should also be noted that this install of 11 is on a WesternDigital 640GB 7200RPM mechanical hard drive from 2012. Drive performance is acceptable(read good). (forgive the notepad mistake, I moved a sentence and it pasted in the wrong place, didn't realize it had happened until after taking the screenshot)

How many people really care .

Anyone who doesn't want to throw away a perfectly good and working PC? Just because it's older does NOT mean it's useless.

 

[frankr2994]

Did you make sure to increase the power limits in the TPL screen? You have to bump the power limits when OCing by more than 10ish%.


I will answer with the following;
View attachment 278185
There you go. Just because Intel, microsoft or anyone else says it's not "compatible" does NOT mean it will not work and run perfectly. It should also be noted that this install of 11 is on a WesternDigital 640GB 7200RPM mechanical hard drive. Drive performance is acceptable. (forgive the notepad mistake, I moved a sentence and it pasted in the wrong place, didn't realize it had happened until after taking the screenshot)


Anyone who doesn't want to throw away a perfectly good and working PC? Just because it's older does NOT mean it's useless.

Using the powercut option. TPL is irrelevant. Just didn't have time to get powercut set before I knew what was going on. I'm pulling around 250watts so there is no setting for PL1 PL2 to make that happy

And you mis understood about who really cares. I wasn't knocking the older PC exactly. What I meant was Microsoft isn't going to get backlash for not supporting something that old. No one is going to care enough to write a nasty letter lol. I have no issue myself with anyone using old hardware..as long as it can still be practical for something. If it was low end junk new then nothing makes it good later. Most of us rocking old stuff the equipment would have been thousands new making it acceptable today for well something

 

[lexluthermiester]

Using the powercut option. TPL is irrelevant. Just didn't have time to get powercut set before I knew what was going on. I'm pulling around 250watts so there is no setting for PL1 PL2 to make that happy

Fair enough. I was having problems with a soft OC using TS and realized I needed to up the power limits at the same time. OC was stable after that. Thought I'd mention it just in case you hadn't tried it.

And you mis understood about who really cares. I wasn't knocking the older PC exactly. What I meant was Microsoft isn't going to get backlash for not supporting something that old. No one is going to care enough to write a nasty letter lol. I have no issue myself with anyone using old hardware..as long as it can still be practical for something. If it was low end junk new then nothing makes it good later. Most of us rocking old stuff the equipment would have been thousands new making it acceptable today for well something

Ah yes, thinking I did misunderstand you. Companies have their own bottom lines to consider and everything they say to the public is geared in that direction, including lying about compatibility.

 

[frankr2994]

Yep. Just remember it's never my intention to be a dick lol. I try not to bash anyone or their equipment. All the threads I monitor pretty much revolve around "outdated" equipment.

 

[lexluthermiester]

Yep. Just remember it's never my intention to be a dick lol. I try not to bash anyone or their equipment. All the threads I monitor pretty much revolve around "outdated" equipment.

Oh no worries! That's not how it came off. It just seemed like you were taking a particular position when you actually meant something else. Thank You for explaining.

 

[Retrorockit]

Well, then explain this: https://www.intel.com/content/www/us/en/support/articles/000006105/processors.html

View attachment 278162

Sometimes those lists are only what they bothered to test and certify. The Q9505s R0 stepping is the latest of the LGA775 CPUs. I would look at a BIOS update for a possible solution.
Dell is notorious for listing RAM cpacity based on what t hey offered at the time. Usually you can double that based on what the chipset supports. But Dell would rather sell you a newer computer..
Then there is this-
Tracking (PPT) envelope, which is a measurement of the maximum amount of power delivered to the socket. The Core 2 Quad Q9450's PPT tops out at 95W, while the motherboard can pump up to 142W to the Core 2 Quad Q9505S at peak performance. That opens up much more aggressive boost behavior, on both single and multiple cores, that could widen the performance gap beyond what we see on the spec sheet.

 

[frankr2994]

Well I literally have a team of people working on a t5810 bios mod right now haha. looks like the use of TS will continue as it seems to be impossible to make the bios menus visible. However I should end up with any and all limits gone and allow me to clock to the moon. Going to buy another motherboard and a 1650v3 to test on when ready. I'd hate to burn up my expensive 1680v3 if the board is going to end up shorting out or something. Also someone is working on using the JTAG connector to start programming with a raspberry pi. Might make changes easier in the future.