imsolidstate

Hi guys, I couldn't get what I wanted for my laptop so I decided to keep it. As such I've been tweaking it again. I have a 6990m crossfire setup that is constantly overheating, so I tore my machine down today to try and fix it. I think this mod will work for any GPU in an R1 / R2, as I expect the heatsinks all have a similar deficiency. Let me know if you have a different card and this mod fixed your overheating issues. Thanks! See what I did and pics here: M17x 6990m / 6970m overheating » imsolidstate

Hi everybody, just wanted to let you all know I listed my M17x for sale on eBay. I really hate to sell it, I've put so much time, money, and thought into it but my student loans are killing me. With that said, I wanted to thank all of you for being cool. I hate forums, but this forum is one of only two that I make exception for because of the users. Thanks for the help everybody!

I've been searching for something to get fan control working on my 6990m GPUs in linux. They are stuck at whatever speed was set during boot by the BIOS. HWinfo64 works great for Windows, but is there something similar for linux? Setting aticonfig --pplib-cmd has no effect. It reports success but has no effect on the fans. Reading the temps seems to work ok with aticonfig --adapter=all --odgt. Since HWinfo64 can make it work on Windows, I'm pretty sure it's an issue with the drivers writing to the wrong registers. I wondered if anyone knew something about this. I know most of the temps are pulled from the SMbus, but I don't know if the GPU sensors are available on the SMbus. I think I might start looking for an SMbus/I2C library and dump the SMbus traffic with a C program. Otherwise I would have to take the laptop apart and get hardware access to the bus with a logic analyzer, which I don't really want to do. Does anyone know more about this? Thanks.

Hey, thanks @Nospheratu. Since you mention that there usually is a flash of the desktop a few minutes into the game, I think it's the OSD telling me about CAPS on/off. The first one breaks into the game but not after that. I will try your advice.

An undervolt and some thermal mods got me down to 83C, which seems safe enough. However, I figured out something else that is bothering me. My secondary card drops out after playing a game for a few minutes. I'm currently playing Deus Ex. If I start the game and just leave it alone, crossfire is fine, both cards get hot and stay hot for as long as I leave it. If I start playing the game, after a few minutes the secondary will go to idle and the primary takes roughly 100% load. Is this because my secondary is a 6970m flashed to 6990m and the secondary card crashes when the primary tries to access a shader that isn't there? It's the only thing I can think of since it appears that GPU memory usage is dynamically allocated and so I assume the shaders must be too. Shaders appear to be the only differences between the 6970 and 6990. Any thoughts? I might try swapping the cards so that the card with fewer shaders is the primary, see if that fixes it. Thanks.

Good catch @Khenglish. I was figuring ~22A since that was the max I pulled from my 330W supply before shutdown when I load tested it. So 22A * 1.1V = ~24W. [MENTION=119]Mr. Fox[/MENTION] sounds like you've got it sorted. Good work.

Mr. Fox, your 40A regulator should be perfect. The 25A may not stand up for the long haul as it is probably rated for 25A of rectified AC current which is a much different thing than 25A of DC. You don't need to parallel anything with the 40A, that is the way to go. Forget about what I said about paralleling diodes, I said that because you are only using two of the four diodes in the bridge, but they are connected together so you can't do what I mentioned. That 40A rectifier should be fine as long as you don't overheat it. It could potentially be operating at 15 watts.

Sorry Mr. Fox. I just tried to draw the circuit and it won't work with the bridge rectifier, my mistake. You can only use half of the bridge in this configuration so you would need two bridge rectifiers to go parallel. Just check the rectifier you have to make sure it can take the current, each diode has to handle about 22A of continuous current at max load.

Hello, looking for an opinion on how to fix the overheating issue on my 6990M crossfire. The primary card gets quite hot, the secondary one doesn't. Idle temps are approximately 50C, but I reach 90C in games and occasionally get what I'm guessing is a thermal shutdown. This is the shader temp, as reported by HWiNFO64. Other temps on the card are slightly lower than the shader but track the shader temp. One of my cards has an out of spec core voltage regulator, I can't remember if I put it in the primary or secondary. I hacked the power good signal so that it would post and run since the voltage is only slightly off, I checked it with a meter. It's something like 0.1V high when unloaded, and comes into spec when loaded. So my question is should I undervolt just the primary card or reduce core clocks, or both. If I reduce core clocks do I need to match the two card frequencies? How much should I underclock? Thanks.

Sorry Mr. Fox, I was busy. As you found out, connecting the supplies in parallel with a diode for each will work, but there will be no load balancing. There will be a lot of current on one side, so make sure the diodes can handle the current. Since the original mod was to use a rectifier, you could parallel the diodes for better current handling. Good luck.

@Nospheratu, yes, I think that would work. As long as the motherboard can read the ID chip and the power supply sees a low level on the ID line, you should be fine.

Ok, sorry to everybody that has been waiting around for a fix for this. I've been busy and haven't had time to dig into it. I did figure out what is causing the power supply to trip at 240 watts. It's a simple fix, and my PSU put out 440W before shutting down once I fixed it. See here: 330W power supply for M17x update » imsolidstate

I was hoping someone would answer that question. I keep getting emails asking about doing the conversion but if I do that I really need to know what works and what doesn't. What causes the high power consumption with 920XM/7970m crossfire? Or is it something else? It doesn't make sense that my 840QM/6990m crossfire uses about half the power consumption.

Well I guess we have to assume that the power supply current limit is not present on the R2 motherboard, or at least not on the particular rev that StamatisX has. Unfortunately though this makes the problem even more difficult to solve. The various regulators on the motherboard for all the different voltage rails have feedback through resistors to the PWM regulator and the higher load of the 7970m crossfire / 920XM may be pulling a rail low enough for the regulator to drop out, causing a shutdown. There could also be other current sensors on the board somewhere that we don't know about. Maxim makes some really tiny current sense amplifiers. My advice to those having this issue would be to try scaling back clocks one at a time to see if they can get around the problem. For example reduce the CPU clock and see if the GPU still causes a failure, then try just reducing GPU clocks. I realize this isn't really a very good solution but the only way to really figure out what is going on is to use some specialized equipment to monitor each rail and find the one that droops. You would need something like a logic analyzer with a front end for each rail to connect it to it's own channel so that you can compare all of the voltage rails and see which one fails first. You have to be able to compare the rails or you won't know which one drops first. - - - Updated - - - Wait a minute, forget all that about monitoring the rails. I just realized StamatisX said 16 Amps. That's 320 Watts. On the output side of the adapter. The input side is necessarily higher. That means that he is exceeding what the 330W supply is able to deliver, and he doesn't have that problem because he has effectively a 480W supply. This correlates to emails that I've received from someone else saying that he noticed the adapter would shutdown, the indicator light would go off and he would have to unplug the adapter to reset it. It would appear that the 330W supply is not enough for a really maxed out M17x, i.e. 7970m crossfire and 920XM, dual hard drives, etc.

Radji, I think you are right. I measured current draw on my system as I haven't had any problems, and I'm using around 200W. However I had a look at the R1 schematic, and there is a current sensing circuit that detects overcurrent conditions for the battery and different adapters. The system will trip at 11.69A on the 240W adapter. Of course this is the R1 schematic, I don't have an R2 schematic. I assume that something similar is on the R2 MB. Since the trip limit is set by a few resistors it is possible that the trip limit could be raised but this isn't an easy fix as you have to change out resistors on the motherboard.

A HUGE HUGE thanks to Brian! The 6990m vBIOS you posted works perfect. And since the failure mode was identical to my first 6970m, I tried flashing it to that card as well. Works. So I wonder if the card I originally had was a 6990m, but someone flashed it to a 6970m for some reason? Or are some 6970ms capable of flashing to 6990m? Maybe they are speed binned and I got lucky, who knows. I didn't try it under load since I don't have the heatsink/fan so maybe it won't work under load. We'll see. Once I get a crossfire cable and secondary heatsink/fan, I'll be testing out 6990m crossfire.

I just tried diff/dhex with the file I downloaded from here and the file at notebookreview and they are showing as the same. I'm not sure where my problem is. The Dell vBIOS is fine on both cards. When I try to run the Clevo 6970CX0.bin vBIOS its fine in the POST/BIOS screens but as soon as windows loads I get an ugly screen. I've tried different Windows drivers but right now I'm running the mobility driver. The problem exists in Linux as well, the screen looks the same and it crashes at the login once I restart after installing the proprietary driver.

Okay I had the same nasty screen issue flashing the Clevo vBIOS on the new card. Does anyone have a hash of the Clevo vBIOS that is downloadable from this site? I must have a corrupt vBIOS.

Hi everybody, been really busy for a while but finally got another card for a potential crossfire setup. I have a Dell 6970m that will only run the Dell vBIOS, the Clevo vBIOS gives me a really ugly picture. I hate that the computer won't shut down by itself, so I wanted to test it as a secondary with another card running the Clevo vBIOS. I got a Dell 6990m, which had the Dell vBIOS shipped. What I want to do is run the Clevo vBIOS on the 6990m as the primary. Problem is I am having trouble finding a crossfire cable to test it. eBay and google are only turning up the old cable and they still want like $60 for it. The heatsink and fan are also really expensive at eBay prices. Can anyone tell me where I might be able to get this stuff for relatively cheap? I searched the marketplace but didn't find anything. Thanks.

Okay, thanks svl7. If I find a good deal on a 6970 and right side heatsink I'm going to try it.

If you have an R1 I can tell you exactly what to do but I don't have an R2 schematic, and I don't plan on tearing down my computer for a while so I can't tell you exactly where the inductor is. Otherwise you have it correct.

Hello everyone, just wondering if someone can tell me if crossfire will work properly with two different vBIOSes or two different manufacturer's cards. I have a Dell 6970m that has some kind of issue where it only works on the Dell vBIOS, not the Clevo vBIOS. The Dell vBIOS doesn't work well in the R2, so I was wondering if it would work to run that card as the secondary and run another 6970m as the primary with the Clevo vBIOS.

Yeah, BS in electronics engineering and MS in computer science. It's okay though, I like talking about this sort of thing because you can always learn something new. I guess that's how I ended up here anyway, I was bored and needed a new project.

Well, we were talking about diodes in a VFD. A VFD has a three-phase rectifier as it's input stage, which is passed to an RC filter to create DC. The drive then synthesizes a sine wave by rapidly cycling the DC rail to the motor windings and varying the duty cycle to create an equivalent AC voltage. The switching creates lots of noise and back EMF that is hard on the semiconductors. In comparison DC loads are easy to calculate and design for. If the diode is rated at 15A, it will work up to 15A. In a laptop there aren't any inductive loads that will produce voltage spikes, so the diodes will be fine as long as they are rated for the current they are operating at.

I agree that using diodes won't balance the load. One power supply puts out slightly more voltage than the other, and it will shoulder most of the load until the voltage droops enough that the second one starts sourcing current. However as long as neither supply is driven into overload, it shouldn't be an issue. The fact that the voltage should drop as current increases means the loads won't balance, but they shouldn't go into overload either. It kind of works like an overload protection. VFDs are a lot different, they handle large switching currents and inductive spikes, which is probably why the diodes failed. In this application it is a straight DC load so the diodes should be fine. The only way to truly balance the loads is to build a switching circuit with a couple MOSFETs controlling the power supply inputs. Let the power supplies switch charging a capacitor that feeds the motherboard. Overkill for this though really.