Khenglish

Furmark isn't a very good test anymore due to current limiting and drivers throttling the card when simply detecting that furmark is running. I go by what my temps are with vsync off in games.

So did you make the coolers yet? I'm wondering how things turned out.

I'm thinking that win8 just lacks a driver database for it. Installing a driver manually should make it work.

Personally I recommend that you do not get an AMD APU. I tried to get an A10-4600m with a 7660g running well and I simply could not do it. The system had very bad stuttering that I simply could not fix. I doubt that the stutter was not just my system in particular since with the tweaks I did (locked NB frequency to max, locked CPU to 2.7GHz/3.2GHz, tightened memory timings including CAS)I was significantly outscoring a default APU, but the stuttering was still just as bad. I think the main reason for the stutter was terrible driver management. With GPU-Z and afterburner you can see the GPU clocks switching several times per second between 324, 485, and 685. I tried to lock the GPU frequency to max, but disabling power saving features would lock the GPU to either 324 or power-on clocks (200). The stuttering was so bad that I would definitely recommend an HD4000 over it even though in benchmarks the APU usually scores much higher. If you can get a dedicated GPU though in your price range then by all means get that over the HD4000 though.

Yeah it looks like it's just a proprietary x8 connector. On a side note the mPCI-E connector was originally designed for x2 and it has enough pins for it, but x2 mPCI-E never materialized. The only devices that need more bandwidth than 1GB/s (PCI-E 2.0 x1) are graphics cards.

12V rail is very easy to measure with a multimeter. Just stick the leads in a molex connector.

That connector has many more pins than a mPCI-E connector. It just happens to have the break in a similar spot to make it look like one. It would be suicidal to rely on only an x1 link for SLI like that since the ultrabay GPU would need to send its framebuffer and other synchronization data over PCI-E to the internal card. Adding PCI-E links only requires 4 signals each. The reason PCI-E x16 has so many lines is mainly for power and the motherboard not knowing the capabilities of the card that is being plugged in (an x4 card must work in an x16 slot), but when you know what exactly is going to be plugged in, you can cut out a lot of lines. I am guessing that the ultrabay GPU is x8 PCI-E 2.0. Possibly x4 3.0, but I doubt lenovo made a good enough connector for 3.0. You can check if it is or not in GPU-z. Just make sure you do the little render test to make sure the link is fully engaged. As for hooking up an eGPU to the slot, you can't without an enormous amount of work and resources. SLI requires 1 card to be a master, and others to be slaves. The BIOS will likely be set up to always treat this slot as a slave. And that is the least of your problems. The main problem would be to somehow design and manufacture your own custom PCB to plug into that slot and wired to a full size PCI-E slot. It takes very good connections to transmit PCI-E 2.0, let alone 3.0. You're basically making your own PE4H but without the cables and without any documentation on the pinout of the motherboard connector.

That salesman is a dick. The AX860i is a very high quality PSU. That 600W figures is assuming you're running a titan full load and have like 6 HDDs, 6 fans, several CD-ROM drives, a floppy, a ZIP drive, a high power CPU, and you lined your case with christmas lights. Titan has a TDP of 235W. That means 470W for 2 titans alone, then add 250W for a realistic conservative number for the rest of the PC. 720W, not this 1200W BS. You might have problems if you do a lot of overvolting then run furmark with current limiting disabled, but that's it. @Brian What makes you thin the 12V rail was an issue? Was the PSU shutting down? When messing around with my 580 in furmark I could drop a 550W PSU down to 11.48V. I thought it may have been hurting my overclock so I got an HX850 which never dropped below 12V, but it made no difference.

With my CPU over 4GHz and using the 1V BIOS I need to turn down by screen brightness to prevent my 240W PSU from shutting down. My GPU temps are mid 70s with fans on automatic so I will go higher voltage after getting a stronger PSU (and VRM/MOSFET) cooling). You should be able to push higher voltages than me though since I doubt your CPU will pull over 80W in games like mine.

Yes clevo BIOS options are horrendous. Luckily though you can do most things with XTU that you would normally do with proper BIOS settings. You're only running 2 RAM sticks? On the P150EM and P170EM if you only have 2 sticks you're supposed to use the memory slots under the backplate, not the keyboard. I'm not sure if it really matters though and if the P370EM is the same.

If the GPU is under light load it will automatically clock down to save power even when plugged in. You can prevent this with nvidia inspector if you want, but I haven't noticed the clock changes causing any stuttering so I don't think this is something to worry about. The card will slow down to 758 if it reaches 90C, but yours is dropping to 771 so it looks like it's just not being stressed.

How did you overclock the CPU? You got a real BIOS?

I've been trying to figure this out too since in theory 2 8mm heatpipes can transfer heat more efficiently than 3 6mm heatpipes (2*4^2 = 32 vs 3*3^2 = 27). I think the reason the 3 pipe is better is because it has more contact area. Many laptop heatsinks have the heat pipes embedded within the radiator for heat to transfer from the heat pipes to the radiator better. I think since the m18x does not do that and instead has the pipes on top of the radiator causes surface area to be a bottleneck.

To overclock my P150EM I need to edit the BIOS with AMIBCP to force higher turbo ratio limits. Each multiplier change = BIOS flash. It's annoying, but it works. PM me if you are interested. lawtq did you find another method or are you doing the same thing I'm doing?

It has to do with the VRM and MOSFETs overheating. I replaced the MOSFET cooling pad with a copper shim and it lets me sustain another 10 watts or so. The system will toggle turbo on and off if I run over 75W for too long. Before the shim turbo would disable in the low 60W range. This will occur with the CPU temp in the upper 80s. The shim made a direct impact with no improvement in CPU temps so it is definitely voltage regulation related.

Clevo does provide VRM schematics if you want to try to figure out a hardmod to reduce voltage. There did not seem a to a practical one for my P150EM but perhaps that is not the case for you. Also BCLK overclocking does not automatically raise the VID like raising the multiplier does. I'm not sure if FTIC will work for your chipset but if it does then you can follow http://forum.techinferno.com/general-notebook-discussions/2091-lets-enable-overclocking-all-6-7-series-laptops.html to get BCLK overclocking. I only got 104.17 fully stable so it doesn't get you very far, but that's around .05V off VID at the same clocks as just raising the multiplier.

OK I'll look into that. I thought about that stuff but I never saw someone try it on a raw processor die, but you did with the 680ms so it looks like it's OK. I also thought of indigo since it would discourage me from taking the heatsink off and wasting time messing with the laptop, but not sending $9.50 down the drain each HS removal is nice. Update: Just ordered liquid ultra. It looks like people who didn't have good results with it either got a bad tube or applied it wrong (used the q-tip).

The memory speed increase will help. Your current memory can likely run much faster though if you force it with XTU or flash it with thaiphoon burner. As for CPU I will occasionally get VRM temperature throttling in BF3. When it cuts from 4.3GHz to 2.4GHz my fps will sometimes drop from the low 60s to the high 30s, so CPU speed matters.

I hear ya about the Vcore for mild overclocks. Same problem with a p150EM. Only solution I see is to make 4.3GHz thermally stable, which is no small task. Fixing this would require having real BIOS settings, which it seems clevo is committed to never doing.

Are those temps with the fans forced high? Great CPU temps but it makes sense if you think about it. Games have a hard time using 4 cores let alone 6, so it's like you're running a regular mobile quad core but the cooling is overkill. What does prime95 and the like look like? Looking at P570 cooling gave me an idea for something that I think will work on a P150. Move the fans back and put in an extra radiator. Buy extra GPU cooler and add the core radiator to the CPU side and memory/VRM radiator to GPU side. That HAS to make a difference assuming I don't mess up.

Those are great temps for that room temperature. I'm just worried that XTU won't work so you'll need to use AMIBCP to overclock, which is a PITA.

So how are temps? Can you overclock it with XTU?

I'm tired of thermal pastes drying out on me. I'll get good performance at first, but then the paste starts drying out and temps start going up 3-5C. shin-itsu would start drying out after a month, and PK-1 after 3 days (although the PK-1 beat the shin-itsu substantially before drying out and tied it while dried out). AS5 dried out too bad when I tried it years ago so I'm under the impression that pretty much any paste will dry out, so I figure I'll try something that's not a paste. There's also the plus of it outperforming everything else. I am just 2-3C away from being able to run BF3 with the CPU fan on medium at a constant 4.3GHz and I think that if I get indigo to work then I'll get that 3C if not more. So indigo xtreme is designed for each individual socket for desktop systems. Installation is to place the thing on the IHS, then seal it on the motherboard. Clevos have a fairly high surface area heatsink, so I think it may be possible to apply and seal on the heatsink. I don't see why applying it to a heatsink instead of an IHS would make a difference. So does anyone here have experience with indigo xtreme? Do you think there's a chance of being able to apply it properly to a laptop?

Yes your current CPU will be holding you back in some games like BF3. You'd need to be able to overclock the 2920xm for it to be a significant improvement. I don't know if XTU will let you overclock or not even if you flash on a 170HM BIOS, so it may be necessary to set overclocks by flashing the BIOS for each multiplier change (which I have done with my P150EM). Another problem is the 150HM cooling unmodified won't be able to handle a 2920xm at any higher than stock. You will get headroom though if you lap the CPU heatsink to get rid of the bends clevo put into it during manufacturing, but it won't be much room since the 150HM has an aluminum CPU radiator (I paid extra for copper). And the final problem will be power. I'm surprised that you're not getting PSU shutdowns already with your 180W PSU. I have a 240W and it shuts down on me with the 680m under load and the CPU pulling 75+ watts. If I use the 180W the PSU will cut out with any overclocking at all. So it's up to you. You can get significant gains with an xm CPU, but it will be a lot more complicated than just getting a new CPU.

What PCI-E are you running with these cards? I assume both cards are at least x8?