Throttlestop overclocking Desktop PCs

[frankr2994]

Did a bunch of reading and it does look like there are ways around and back doors but it seems that if you have a 10 dollar programmer it just skips all the BS. It's on order and something like that I should be able to use at work too. Typically send chips out to have them flashed...maybe I'll learn how to change the tune in early fuel injected cars instead of paying for it.

 

[Retrorockit]

I'm a retired truck mechanic, so you're much more advanced than I am at this. My thinking is the back doors will point you at what needs to be done in the BIOS when you get there. If there aren't any menus there, each setting could require a BIOS mod. Somewhere at the BIOSMods site they gave instructions on how to make the checksum add up.
Early EFI ? You just hammered the copper washer under fuel pressure regulator thinner when you hung a turbo on them.
Actually that was the early Bosch injectors which were mostly mechanical with a few sensors added for emission control.
Batch fire,TBI, Lambda sensors. Aaaaaaaaaaarghh.
I went over to diesel trucks around the end of the carburetor era. Good career, move bu, ruined it as a hobby.

 

[frankr2994]

Yep and you can buy a resistor on ebay to make your mass air flow sensor show it's getting more air lol. Usually like to have better control on mixture and timing on track cars. We get a number of chips flashed and occasionally do stand alone units.

 

[frankr2994]

Ok back to TS overclocking. Waiting on a flash programmer and a atx psu bench kit to really get into this but for now.. turning hyperthreading off gives me alot of headroom. Not saying I want to leave it off but can I use this to find a stable voltage at higher clocks? Right now I can run 4.7 non hyperthreaded and I think I can get 4.8 out of it. Just trying to see if I can get an idea what I'll need.

 

[Retrorockit]

You can text edit TS to overclock even fewer cores than that.
Since you're the first one doing this system there's not much we can tell you about what to expect. Half the fun here is that we're not just copying what others have done many times already.

 

[frankr2994]

Well I more or less thought that was a general question. I guess to rephrase would voltage requirements be the same with HT on or off? If so I could find a stable voltage at high clocks but keep my current requirements below the threshold

Oh and a little update on linpak. Threads running is not relevant to hyperthreading. Running benchmarks it will only run 6 threads regardless of HT on or off. With it on it shows 50 percent utilization however it was designed to feed info into the cpu so efficient that it's really 100 percent load. Watts and temp rise the same if it was loaded 100 percent. Ive also found out that it was not a mosfet shutting down ripping my wattage up and voltage down. The hard limit seems to be 158 watts. Running at lower clocks I can see a large portion of the stress test running at 70 watts then Jumping to 120+ watts. No throttle flags. Bumping the clock watching it hit above 140 watts throws yellow throttle flags with no actual throttle happening. Going a little higher smacks the 158 watt limit resulting in red throttle flags and a decrease in voltage. The highest it will possibly go is 158 watts no matter of voltage or clock speeds so I know this is a hard limit I will need to change. Changing turbo limits and time does nothing to change this behavior and my temps stay below 80c. Usually in the low 70s. My system is fully put back together as I got my other case fan in the mail yesterday and have it installed. Lots and lots of airflow and I have the mosfets cooled with a decent aluminum heatsink leaving just the chokes to have something installed on them.

 

[Retrorockit]

There is a hardmod to spoof the VSENSE pin on the Voltage controller chip with a pull down trim pot. This also requires adding a digital Voltage display to a VRM choke so you will know what you're now getting. If the wattage is "calculated" VxA=W then this should spoof the watt limit also. If it directly measures Voltage inside the CPU it won't work. Of course the die shrunk newer CPUs may not be able to tolerate the 200+Watts that 45nm, and 32nm CPUs could.

 

[frankr2994]

Well playing with some different software yesterday I was able to unpack a bios exe from Dell downloads. Opening it up in ami configuration shows that all the menus for adjusting everything are there and just hidden from view. So if I get my programmer and a little help from the biosmods or winraid folk I should.....be able to have a fully unlocked system.

 

[Retrorockit]

The 2 security features I'm aware of are the Checksum which I believe has been solved at BIOSMods, and digital signing. Dell offers BIOS updates for many years so I'm not sure choronologically when this started, or exactly what it does. But my Optiplex 380 didn't have it until BIOS A07. So I had BIOS A05 moddded to avoid whatever it does. DCCU has a setting to allow installing older BIOS files. IDK if this can defeat that or not. Maybe the oldest BIOS for that won't have it? Maybe the chip burning method makes all this moot?
Some of the newer BIOS files lock down the unlocked Xeons. Plus the performance hit for Spectre/Meltdown protection (but may be very important to some users).

 

[frankr2994]

As far as I can tell the programmer makes all the security stuff irrelevant. I know checksums and file size are still very important.

 

[Retrorockit]

A lot of people advise to get the latest BIOS file available. But I've found that's not always the best idea lately. At least not for performance modding.
Great news that the menus are actually in there!

 

[frankr2994]

Well I haven't gotten a response over on the other forums about modding the bios. I've read to make the menus visible to set them from default to user and if that's really it i guess I'm good to go. It's just really everything is set to default so it's obviously coded for default to have a different meaning for each variable. Before I flash anything I'll go into the bios on the 5810 and just go down through making sure I set everything I don't see to User and hopefully that will be the fully unlocked bios. There are alot of options and it did take me a while to find tdp settings as they are not part of the overclocking menu

Oh and FYI I still plan to use TS for my overclock. I very much like being able to change everything as I test and not having anything saved for when it's not right and crashes. Still should save a ton of time vs only changing settings in the bios. So it should still all be relevant to this thread.

 

[Retrorockit]

I started the TS thread because it was the only game in town for overclocking locked BIOS computers. But I'm open to other methods as long as it doesn't turn into another aftermarket overclocking forum. Modding office computers, workstations, and other locked down systems is hard to find information on. The little bit that exists gets lost or "talked over" in other forums. So it's welcome here. Lex has a Dell workstation thread for more conventional users, but pretty much anything goes here.

 

[frankr2994]

Totally get that. I was thrilled when I found this thread and would love to convince more to take a more economical route to having an awesome computer.

 

[Retrorockit]

One thing I do ask is if you try something, and it doesn't work, post that too. With these computers it's easy to bang your head against the wall trying things that don't work. What works on one system may not work on another. Especially with Dells that don't always follow industry standards.

 

[frankr2994]

One thing I do ask is if you try something, and it doesn't work, post that too. With these computers it's easy to bang your head against the wall trying things that don't work. What works on one system may not work on another. Especially with Dells that don't always follow industry standards.

Hasn't most of my post been about it not working

 

[Retrorockit]

Hasn't most of my post been about it not working

I just like to get that in the thread every now and then. When I was answering Dell Optiplex questions at Tomshardware I found that many things that failed on one system would work just fine on a different one. What didn't work for one guy may solve somebody else's problem.

 

[frankr2994]

Got the programmer. Still haven't received any input from other forums so I'm going ahead. I pulled power and ground from the usb socket,soldered pins and attached an old fan lead into my atx breakout on the 3.3v rail. The 3.3v regulator on the board dropped this down to 2.4v though.....grabbed a jumper from and old hard drive and jumped the 5v and 3.3v pin on the programmer. I now have a totally controlled 3.3v programmer capable of supplying the amperage of the backfeed on the motherboard when hooking up to the bios chip. Next step is learning how to do a bios dump.

Used a program called asprogrammer . Was able to successfully pull 3 bios dumps and all of them have the same file size and all of them open up just fine in amibcp which is an ami bios configuration tool. I read something about a boot guard that won't let the pc boot with a modded bios and it has to do with intel me. Still reading up on what I have to do with that. But bright side my voltage mod on the programmer worked great! I have to check the files with a hex editor to verify the dumps are exactly the same and if so I'll start modding one of them. Hopefully by the end of the weekend I'll have an unlocked workstation.

 

[Retrorockit]

The subject of Supermicro servers came up. I found this news story. Kind of OT but may be of interest to some.

Supermicro spy chips, the sequel

- It really, really happened, and with bad BIOS and more, insists BloombergServer maker says latest article is 'a mishmash of disparate allegations'Following up on a disputed 2018 claim in its BusinessWeek publication that tiny spy chips were found on Supermicro server motherboards in 2015...

www.nexusnewsfeed.com

 

[frankr2994]

Ok flashed my modded bios. Didn't brick it but I also didn't accomplish anything......bios menu is identical to how it was before. I pulled another dump and verified my new modded bios was indeed installed. Not sure what the hell is going on.

 

[Retrorockit]

Some of the locking happens in the PLL chip. Whether it's the BIOS that activates this, or it's burned in at the factory, or hardwired IDK. There's a user from China at BIOSMods.com named Genius239 that does some Dell BIOS modding. He modded an early Optiplex/ Xeon BIOS for me to avoid digital signing. Perhaps if you start a thread there, or PM him directly you can get one of those guys interested. I know on the PLL chip there is a locking pin that gets a locking signal (input) at startup, and then has an output for the USB bus after that to avoid spoofing the signal. Perhaps the BIOS chip has a similar scheme, or may be hardwired to activate a locking feature to do this. The datasheet for the chip itself may reveal this.
One approach could be to mod some memory modules for reduced latency using RWEverything instead of raising the bus speed. If you can get a low Voltage module to call for standard Voltage there should be some headroom there. On the one PLL I investigated there was one setting to lock the PLL, but another to select which FSB it was locked to. No one here has tackled Dell BIOS modding. It's there to be done. I wish I could be more help, but it's outside of my very limited experience. I do TS overclocking because it's usually very simple.

Did you change any settings when you flashed the BIOS, or just try to unlock the menus?

 

[frankr2994]

I did not change any settings. Well one I defaulted enable oc settings. That just should make menu items not gray if I actually could see them. I have threads on biosmods and winraid. The only 2 places I've found that deal with this. It seems that the main people over there that seem to help everyone have not been active over there for a couple months. I received my first reply at one of them about hex editing. Need more info and hopefully I'll get it. I started both threads about a week prior to trying anything as I was hoping to get some info and tips before my first attempt

 

[Retrorockit]

I've tried to find books about Hex editing and got nowhere. The little I've picked up has been from datasheets, and looking at SetFSB screens.
SetFSB PLL list hasn't been updated since the late LGA775 era. I read the datasheet and look at the screen and I can see what's going on, but how to fix it in the BIOS is beyond me. Once TME is activated on the PLL chip it's not going to raise the FSB. I've seen videos on Youtube that unlock the PLL TME in Clockgen on some very old Dells (XPS9200, Dimension E520 from 2005) but I don't understand what's being done there.
One approach I've thought of is to find the FSB setting which will be in Hex, and transpose whatever digits you find there into something better, so as to preserve the checksum value. Maybe you can get some better values locked in at boot?

 

[frankr2994]

You know one thing I didn't think to look through are all the IT books I bought for college. Some were given to me and most were not needed for a class. I got no answer idea how far some of them go in depth but I'm going to dig them out.

 

[Retrorockit]

"When" a BIOS signal is sent is important too. Some like the TME locking are sent before POST is run. Others like the bus speed outputs, and maybe FSB speed limit selected are applied after the Power Good signal. So the locking signal, and the speed selected as the limit may need to be approached separately. If you can't remove the limit, but can raise the limit selected then it should become possible by some means to select any speed lower than the new limit.