BIOS modding for GPU OC fun and profit!!

Just wondering Seeker, are those A11's you posted with the unlocked BIOs?

 

I installed the 5730 800/1050 BIOS, but in windows I kept getting garbage on the screen which started getting worse, OS became slow and soon followed with a BSOD. I tried it a few times, I could only get windows to work properly without power supply because of powerplay..
I have a Studio XPS 1645 with 5730 1GB 650/800, 4GB RAM, i7-740QM.

Any help?

 

The garbage (or artifacts we called it) on the screen means the clocks are too high. Overclocking limit on each individual GPU, even though it is the same notebook & the same GPU, are not the same. First step, you'll need to obtain how high your 5730 can go using overclock software, e.g. Spphire TRIXX or MSI Afterburner. When you got the highest & stable clocks, request BIOS mod here.

 

No, just regular A11. If you really want an unlocked one, I'll get it to you, but I can't promise you when.

Like Kizwan said, and like what is mentioned many many times in this thread, you must test before overclocking. If you look through the thread, there are many posts and links that discuss the methodology for good testing for an OC. Personally, I use the AMD GPU Clock Tool to OC, and then use Prime95 + OCCT, and then 3DMark06 to stress test for temps and stability. Start small and work your way up. If you're getting bad artifacts on just using windows outside of gaming, then the memory clocks are WAY too high for your particular laptop. OCing is always a crapshoot, and it's possible that your laptop won't want to OC at all.

 

@seeker_moc

Maybe you could just add some sort of a message on the first page?..recommending that people test clocks before flashing an OCed BIOS.
Doubt anyone's gonna go through the 91 pages on the thread.
and you wont have to keep telling people to test before flashing.

 

I thought it did already... guess I forgot it. I did provide links on where to go to learn about testing, but not about how critical it is to test first. I'll add it this weekend.

 

Thanks for your help guys
I tested and 800/1000 seems to be stable on my system with software overclock. However, the strange thing is that when I use 800/1050 using software overclock my windows works fine and my display driver crashed got reloaded (only) once in about 3-4 times that I ran all the 3DMark 06 tests.. but using BIOS overclock my OS became unstable at those clocks. Any idea why?
My current BIOS is A09 if that makes any difference..

 

Even though it just crashed once or 3 to 4 times, this is proved the clocks are too high for your GPU. "Stable clocks" are clocks that doesn't causing any problem or crash at all.

 

Yes, I know. the crash was at 800/1050. I tested a lot, played games at high settings for a lot of time and everything is fine at 800/1000. Hence I said 800/1000 seems stable. (difference of 1050=unstable and 1000=stable, seems to be the limit of stability)
My question about the unstable clock of 800/1050 is that its not producing artifacts in any tests and absolutely no problems on windows outside gaming.. but using bios overclock the artifacts were all around on my windows desktop itself, slowly causing everything to lag and eventually BSODing. those are very different behaviours for the same clock settings..
so I wonder if me finding 800/1000 stable using software OC would still be stable using the BIOS

 

If you already tested 800/1000 with GPU stress test tool & doesn't crash at all, you can use BIOS overclocked 800/1000. There is no difference between software overclock & BIOS overclock, except with BIOS overclock PowerPlay works. Recommended use OCCT to stress test GPU to know for sure the overclock is stable.

 

It just a small difference. The GPU at startup/boot already running at max clocks until Windows loaded. So, if BIOS overclocked to 800/1050, GPU already stressed earlier than you realized. Of course GPU clocks will lower because of the PowerPlay but at this point GPU already unstable. Try overclock to 800/1050 using software & run OCCT. You'll get the same result.

 

alright, I'll use OCCT now. what are the optimal settings for OCCT to stress test the GPU enough for the best judgement?
I'm running only "GPU : OCCT" (GPU : MEMTEST is greyed/unavailable, hopefully thats fine for ATI 5730) with settings:
Test type: Auto(1h)
Resolution: 1920x1080 @ 60Hz
Shader complexity: Default(0)
Error check on (however it says this will reduce the temperature so it won't reach the peak)

Please suggest any changes or recommendations.. Thanks

PS. also I'd like to ask if the core clock and memory clock have individual limits or do they have a limit in combination? I mean is it possible for me to only test increasing the memory clock with default core clock to find the maximum memory clock limit and then start with increasing the core clock? or does increasing one clock reduce the limit of the other?

 

I think the best settings are:-
- set "Test Type" to Infinite
- un-tick both checkbox (running fullscreen is good idea too)

Running it for 1 hour should be enough.

You can increase the memory clock first until you found the stable clock. After that, you can increase core clock until you found the stable clock. Actually this is correct & recommended method. If I'm not mistaken, increasing core clock won't make memory clock limited & vise versa.

 

Thanks so much! Here it is on mediafire:

1645_A13W2_4670_750_900.zip

 

From my experience, it's more likely the core clocks being too high that cause the crashing, and memory clocks too high that cause the artifacts. If you can run memory at 800/1050 on OCCT for ~30min with artifact checking turned on and no errors, then your mem clocks are good. If you crash at those clocks with 3DMark06, I'd try turning core clocks down instead to improve stability. Like Kizwan mentioned, maybe you didn't let the GPU heat up enough at 800/1000, which is what gave you the impression it crashes at 800/1050 due to the memory.

For added stress, and to ensure maximum possible stability, run Prime95 (only use 2 threads, more than that and you might overheat the CPU) in the background when you run OCCT. This will heat up the CPU and the GPU. Since they both share the same cooling system, this is a more realistic stress test for the GPU to pass. Once you find stable clocks that keep GPU temps under ~80-82C and artifact free after 30min or so of testing, then you should be good. Close all other programs and run 3DMark06 for about 2 full loops. If you don't crash, you found your optimum OC clocks, and I'll make a custom BIOS for you.

 

Alright, I have tested so far using only OCCT and 3DMark 06, it seems my 5730 is stable at 845/1000. I will try with Prime95 running as well and will tell you the final clocks soon.
However, I have a few more questions.. The stock set of combinations of clocks/voltages given on the first post have 8 entries. Can each of these all be changed? Can it go above 8? And is there any point in changing those?

Also, which contributes more to heating? core or memory clock?

 

edit: thanks for information

 

Yes

Not sure

Some people like to change the power saving clocks even lower, to improve battery life, but it's harder to test them for stability. It allows you to pick the steps that your computer can choose from on it's way from power saving, to office, multimedia, then all the way up for games. However, in practice, I think the effort to test out all the different possibilities isn't worth whatever minimal power savings you might achieve.

 

Instead of modifying the .exe package, modify the bios1.wph file in the temp directory, then use the winphlash.exe utility to flash the modified .wph

 

thank you issue resolved

 

I have searched and asked, but have not found any software that will let you change voltage on an ATI mobile GPU. The only way to do it is through the BIOS, but there is no way to test it before hand. If that's something you really want to do, use very small steps, .05v at a time, and test for stability/heat at every step on the way up. One warning though: it is almost impossible to cause permanent damage to your GPU with overclocking as long as you don't touch the voltage. Once you start jacking up the voltage, you do so at your own risk.

 

I have figured out my stable clocks, but now I want to control the heating.. however there is something I have noticed. I have been testing out the heating on 650/800 and 845/1000, and I haven't really found a significant difference in the rate of heating or the maximum temperature that the GPU reaches.. is that possible? if not can you suggest some test that would help me notice the difference and thus continue testing? so far I've been using OCCT with Prime95 running in the background, however I get CPU too hot under 5 minutes even on 1 thread (blend), on both default and overclocked.

 

okay I've decided on the clocks I need. Can you please make me a custom BIOS for 700/1000?

Full specs: Dell Studio XPS 1645, i7 740QM, 4GB 1333MHz DDR3, ATI 5730, 1080p WLED, 500GB 7200rpm HDD

 

I get the CPU overheating problem too. It's actually worse if you only use 1 thread, try using 3. Using only 1 thread lets "turbo boost" kick in. The overall CPU doesn't get as hot, but the one turbo boosted core shoots up in temp rapidly, causing throttling. You can also try using ThrottleStop to limit your CPU multiplier to 8, which will let it warm up without causing it to overheat and throttle.

As far as GPU temps go, the difference between stock and OC for me is only 2C (78C vs 80C), so if you don't notice much of any difference that's normal. The 5730 seems to handle higher clocks fairly well, but you still have to test, as every individual chip can vary. I find the vRAM to be the real limiting factor for the 5730. If 700/1000 is what you want, I can make that up tonight.

 

ah I see, yes please make me a 700/1000, and if its possible for you to make two, I'd also like a 845/1000. if not, 700/1000 will do. thanks a lot

 

I've flashed the original a12 bios by Dell but I really cannot install those drivers. You know, the point is that I've been experiencing overheating, and hence threshold, after having installed the 11.6 driver. It may be just a coincidence, but it would be interesting to make a comparison chart.

 

I've wondered why the i7's aren't smart enough to switch which core is the 'active' one especially in the case of single-core turbo mode as the core gets near the throttle limit. They're supposed to be so advanced.

 

Windows is supposed to manage that. I've seen it work at times but sometimes it doesn't seem to notice that it's maxing a single core.

 

Hi All,

This is modified BIOS from seeker_moc.
1645_A13W2_5730_700_1000.zip
( MD5: 5417A28AEB63D074FD468129E0BC7D0D)

 

after all that testing, I still get a small amount of artifacts on my screen :/ not in windows (or I may have caught one or two just after startup), but in ubuntu.. so my mem clock of 1000 was still not stable enough. could you please make just one more for me? I want all the cool features such as the unlocked BIOS, microcode update, A09-like GPU throttling in A13, but with my clocks mildly overclocked to 700/950 for 5730.
(or if you have any suggestions for me about the clocks, that'd help)

 

I can't really make any clock suggestions, as each chip is different. Though if you can't run 700/950, then you have the worst performing 1645 I've ever heard of, and may want to check your cooling system (upgrade thermal paste, remove dust filters, clean out fan and heatsink).

If you run Linux, I can make you my Linux-optimized version. It fixes lots of Linux-specific errors, like not finding the HPET (non-modified BIOS treats anything non-Windows XP SP2 or newer, including all Linux, as "other" OS, disabling advanced BIOS functions like the HPET for compatability), missing _BCQ function (LCD brigntness control query; present for the 4670, but not the 5730 for some reason), and errors taking control over PCI-e. There's no negative side effects in Windows, except that it will show some non-removeable pci-e devices when you click on the 'safely remove hardware' button.

 

Nooooo mine barely handles 780, 950. even one more on the memclock gives me artifacts. Sad, but true. though i Agree 780 is significantly higher than 700 in overclocking land.

 

For thermal paste, I want to suggest Shin-Etsu. Shin-Etsu works wonderfully on my Dell with i7-720QM. I use Arctic MX-4 before, CPU cores easily up to 70s Celsius with light usage. Now, with Shin-Etsu, CPU cores only at 50s Celsius with light usage & idles at 47C to 50C.

I think, like ICD7, Arctic MX-4 also need enough pressure to perform best. I noticed on my Dell, heatsink didn't put much pressure on the processor though. ICD7 vs. Shin-Etsu; Shin-Etsu is the perfect choice if pressure is the issue.

 

It goes to show that all the cards are different. On mine (4670) I can max out the "safe" area in Rivatuner for the core (somewhere around 855, haven't tried going further) but the memory can can only go to around 940 for "100%" stability.

One interesting thing is that before I re-pasted, setting the memory clock to around 940 would cause instant artifacts even though the GPU temperature was in the 50-52C range, but after a re-paste 965 is fine even when the card goes well above 80C (in games).

 

It's not that I can't run 700/950. I was asking for 700/950. I have tested my card to work fine for 845/1000. But that configuration brings about slight thorttling very soon, so I wanted a mild overclock along with the other features. Also, it turns out that 1000 mem clock is just a little above my limit. so just to be 100% safe, I asked for 950.

If there are no negative side effects in Windows and I can play my games just as well, I'd like a linux-optimized version

 

Hey Seeker,

I'm interested in the default 13.5 for the 1645 with 4670 (1645_A13.5W_N C.zip), however I would like 1 change.
Could you make it like this?
100 125 0.9
300 300 0.9
300 500 0.9
300 400 1
450 600 1
300 800 1.2
400 800 1.2
675 800 1.2

...or whatever lowest clocks are stable for the 4670. Maybe even lower voltage if possible, too? I'm interested in make the powersave-state as power-saving as possible.

Thanks

 

I'm more interested in the features you mentioned in that first post rather than overclocking itself, so please make me a linux-optimized BIOS for 5730 overclocked to a safe 700/900

 

I agree. I did alot of research before trying the Shin-Etsu G751. It's really good stuff, and one of the few things I'd recommend over AS5. Plus, it's non-conductive, non-capacative, and has no burn-in time.

I can make what you requested, or even something with lower voltage. However, I haven't experimented any with undervolting, so I don't know what would be stable or not. It will take trial and error on your part if you're interested in trying.

Sure, I'll make them up tonight and send them to Kizwan for posting. Sorry it took me so long, but I've been busy lately.

 

Sure I can test.. I tried downclocking with rivatuner to 100/125 but it changes the voltage back to 1.2v rather than 0.9. 100/[email protected] would already be better. Maybe I can then try 100/125 @ 0.7v ?

 

Um, guys .. im planning on swapping out my i7-720qm with a better processor.
Any problems i might run into that i should know of?

 

Overheating xD

 

haha .. yes i spent my day researching this. lol .. and eventually dropped the idea ..

 

Currently on my Dell, i7-720qm max temp are around 70C to 74C when playing any games. I think there is enough room for extra heat if I swap it with extreme edition processor. After overclock it, I think it might goes up to 90C. Throttling only start at TZ01=95C. Still OK I think.

 

Well not for me.. If I watch a movie with the XPS on my bed it get's SO HOT that it shuts itself down instantly. And this is without dustfilter and clean fans.
It just cant get the air out with LCD open and its too weak to suck enough air from underneath

 

Mine is even worse.. i've got some fancy thermal compound applied.. and since i have an external monitor i laptop lid stays half shut. Even then while gaming the temps go up to 80+ C .. i think if i plug in another cpu, it'll catch fire lol.

 

That what happen to me but not watching movies. It happen when playing games. I didn't overclock CPU or GPU at all. To make matters worse room temperature around 34C to 36C. It keep happening even after re-apply thermal paste using Arctic MX-4. I got it fixed by changing to Shin-Etsu thermal paste. This is the comparison between Arctic MX-4 & Shin-Etsu: http://forum.notebookreview.com/7911260-post1823.html (EDIT: With Shin-Etsu, when playing games, CPU temp only around 70C to 74C, not even touch 80C. In the Arctic MX-4 picture, the max temps are what I usually got while playing games before changing to Shin-Etsu.)

BTW, make sure to clean the fan/exhaust grill. Hot air won't be able to escape if dust block it.

 

I use throttlestop profiles to disable turbo while watching a movie in bed etc, and then it won't shut down. Pathetic isn't it?
Btw when it does shutdown and I turn it on 10secs later it will BSOD like CRAZY or give other errors. I need to leave it for 5min or it won't boot HAHHA

 

Yeah, definitely the same like I got before. The BSOD is not like regular BSOD, the text only a couple of lines, right? Now I got it sorted out, no more problem, no more overheat when playing games or running benchmark.

 

Hey Seeker_moc!

Wondering if I could get my hands on a 725X1050 A11 bios for the 5730 (1647).

 

Hmm...REVIVE thread!