Jump to content

Robbo

Registered User
  • Posts

    743
  • Joined

  • Last visited

  • Days Won

    2

Everything posted by Robbo

  1. I get the same Firestrike GPU score as you with your 675MX, except I did it through a higher core overclock and a lower memory bandwidth - I sure would like the 156GB/s bandwidth you got! Looks like core clock & memory bandwidth can sometimes be interchangeable to some extent to get the same performance, although your results still prove that core clock yields better performance/higher scores. Your 50% better memory bandwidth makes up for my 10% higher core clock, so core clock has a greater effect. Interesting to see that comparison.
  2. Point #2 could be your CPU, but if you've got a Sandy bridge i7 or later, then it's not because the CPU's not good enough, instead it may be that the CPU is getting too hot & throttling. For Point #1, someone else would need to answer that, I don't know.
  3. That's useful that you've put the height of the pads required for the different chips. Is it not dependant on whether you have the 460M/560M heatsink or 580M or 6990 heatsink? Or are all the heatsinks the same in terms of dimensions & construction? I know the 580M heatsink looks a little different to the 560M heatsink - the 580M heatsink has a copper section in one area, where the 560M just has 'normal' metal at that point.
  4. @Cloudfire: That's interesting, looks like you're probably right about that after looking at those graphs.
  5. Yeah, futuremark system info doesn't seem that reliable! I might try reinstalling that if I can be bothered, or I might just wait until the next driver update to see if that has any effect. Yes, what you say about the memory is pretty much in-line with my experiences on this laptop.
  6. @svl7: Just a quick question. I tried undervolting by 50mv to see what clocks I could run at, just for fun really. But, when I ran a 3DMark11/3DMark13 run, in the result scores it just listed my GPU as Generic VGA. It hasn't done that before, and seemed to coincide with me trying the undervolt. If I run at stock or increased volts, then I still get 'Generic VGA' listed in the scores. I have installed the 331.82 driver recently, so could be that possibly, but wondered if by undervolting it once it had somehow changed something with how the GPU was recognised? (It's still recognised at 670MX in GPUz, NVidia Control Panel, etc).
  7. Yep, it's a close call between a 7970M (even overclocked), and an overclocked 675MX (as long as you flash that modified VBIOS). You'd be happy with either, the R4 is a good laptop though, a pretty good choice.
  8. @svl7, yep that's really good information from you, and it fits with what I knew about GK104/GK106/670MX/770M already. GK104: 670MX/675MX/680M/780M. GK106: 770M. Notebookcheck have it wrong I believe when they say the 670MX is GK106 (I think they just got confused by the fact it had the same general specifications as the GK106 GTX 660). I always figured 670MX would overclock better than the 'similar' 770M, due to the fact it's a bigger chip with more surface area: better cooling for one, better heat dissipation. @Cloudfire: your memory overclock is good at 1200Mhz, I only get 1100Mhz as you know. It's just luck of the draw on the individual chips you end up with on your board. They're produced to a certain tolerance I guess, and there's inherent variations in their micro structure (if that's the correct phrase) that determines that there will be differences in their ability to run at certain frequencies at any given voltage. So, looks like you had better luck with your VRAM chips, and I had better luck with my core (differences between GK104 & GK106 aside). @svl7: you asked me if the 1100Mhz VRAM was a bottleneck in games. I guess it's hard to quantify. If I have GPUz open when I'm gaming the memory controller load is a representation of the saturation of the memory bus I believe, but I may be wrong about that? When I'm gaming the memory controller load rarely goes above 70%, and in most games hovers around 40-50%. But, if I lower the memory overclock it will lower 3DMark scores a little. When I was doing my overclocking testing though, the percentage core overclock was more influential than the percentage VRAM overclock, so I think the core plays a bigger role. I think as long as you've got 'enough' memory bandwidth then you're sorted, but if you've not got enough, then performance would tank I reckon - I imagine a bit like when you run out of VRAM, but not quite as extreme! I get 106 GB/s bandwidth at 1100Mhz VRAM, so that's 10 GB/s less than a 680M. I'd like to have a bit more bandwidth, but I don't think it's become a major issue in any games yet - a lot of that evidenced by the GPU load mostly being 97-100%. I believe that if memory bandwidth was too low, then GPU load would plumment also, because I think that is a measurement of how busy the core is. A lot of what I've talked about right there is inferred by me observing the GPUz sensor readings while gaming, combined with some relatively limited knowledge on my part, so I can't guarantee whether all this stuff that I talk about is 100% correct. Haha, that's my disclaimer!
  9. OK, sounds good, I'd be interested in hearing how you get on. You're a bit confused on the voltage thing though. The 780M (1.00V) runs at a higher voltage than the 680M (0.987V), also the 770M (0.968V) runs at a higher voltage than the 670MX (0.925V). Also, the modified VBIOS won't affect your memory clock, as the voltage that you can change is only related to the core, not the VRAM. Therefore, your maximum VRAM overclock will be the same on the modified VBIOS - the good thing about that is it's less work, you've already determined your max VRAM overclock. You might be able to get the same core overclock as mine, but your max core overclock for a stable 3DMark run at 0.968V is 915Mhz, whereas at 0.968V I can get 1006Mhz through 3DMark11. So if I was to extrapolate those results I would guess your core would end up at about 1019Mhz at 1.05V (I wouldn't push the voltage any higher though). This is just my estimate though based on comparing our 2 GPU's. Having said that, I'd still rather have your system with sli 770M's!
  10. I don't know about that, maybe someone else in this forum can chime in about that?
  11. I don't have the laptop you're talking about, but I have seen it reviewed on notebookcheck.net. My understanding of the bridge between the CPU and the GPU, is that it allows for more efficient cooling of the GPU or CPU if only one of them is heavily loaded. So, for instance, if you were playing a game, and your GPU was at 100% load, and your CPU was 'only' at 50% load, then the bridge would allow an additional path for the heat to be transferred away from the GPU -> through the CPU's heat pipes, in addition to through the GPU's heat pipes. Now, if you didn't have that bridge there, then in the load scenarios I described above, the GPU would only be able to transfer heat away through the GPU's heat pipes. It's basically a way of making sure that both the CPU's & GPU's heat pipes are being utilised to the maximum extent regardless of how the compute load is split between the CPU and GPU. I think the bridge is a good feature for a one fan laptop. I think the ideal situation, if space allows in a laptop design, is to have 2 heatsinks & 2 fans, one for CPU, one for GPU. That's the most effective cooling method, but takes up more space & probably more cost.
  12. 770M and 675MX are nearly the same chips, just a wider memory bus in the 675MX, and the 770M has higher stock clock frequencies. Therefore, the 770M should run hotter than the 675MX at stock settings, but overclocked they will operate at about the same temperatures. This is because the 675MX is likely to overclock to very similar final frequencies as an overclocked 770M, this I've learned from experience with my own card (very similar to 675MX) and by looking at other peoples overclocks on their 770M's. EDIT: Actually, the 770M operates at a slightly higher voltage than the 675MX/670MX, so the 770M will be a bit hotter in all instances.
  13. SVL7 often seems to be on the lookout for new VBIOS versions to mod, so he might mod your VBIOS for you (& others) if you post it up here. Use GPUz to dump the VBIOS to a file (it's the little button to the right of the VBIOS number in GPUz), and post it up here. I followed your link to that thread, you managed to work out your max overclock at stock voltage, looks pretty good, very similar to my maximum overclock on stock voltage. That's a good omen for you getting some more performance out of your cards - as you can see mine get up to 1124Mhz on the core (which is applicable to you, because the 670MX is pretty much the same card as the 770M). What voltage is your stock voltage? 0.925V??
  14. @Cloudfire. That's interesting, looks like you've worked out your maximum overclock at stock voltage quite sucessfully. Your 770M's overcock pretty much like my 670MX, which I guess isn't that surprising considering they have an almost identical layout apart from the stock frequencies. You managed 100Mhz more on your memory than me, but I managed to get a little more on the core, 928Mhz on the core. Is your core running at 0.925V? (That would be the same as my 670MX). Should handle more voltage quite easily if your cooling is good? Ah, just seen that your temps are 70 degC, that's not bad, looks like it will handle some more voltage & OC quite well.
  15. Haha, yes, that setup with the bag does sound kind of crazy, no wonder you couldn't live with that on your desk! It's fun to mess around with these things though!
  16. I'm not sure, but have you tried comparing the VBIOS version of your card with the VBIOS version numbers that you listed? The one that is going to be compatible, is likely to be the one with the same VBIOS version number. You can use GPUz to find out what your VBIOS version number is. Cloudfire, I recognise you from the NBR forums, I'd be interested to see how well your 770M's overclock. I think they'll overclock pretty well, the cooling system in the A18 is pretty good right?
  17. I've shown you the link to the forum in the previous post. You have to buy it from the guy for a small fee, so I can't share the program with you. Also, it's only of use to you if your 560M while gaming (P0 state) is less than 1V. It it's at 1V already, you're already maxed out unfortunately. You can use GPUz to check what voltage your card operates at during the P0 state while gaming.
  18. Well, I tried to edit it via a program from someone called SVET. I think it was SVET tuner or something. For the 560M, that program allowed voltage to be assigned up to 1V - some 560M's were operating at a stock voltage below 1V, so those people could use the program to mod the VBIOS to change it to 1 Volt. I remember reading that it couldn't be increased over 1 Volt - and I have a vague recollection that it might have been hardcoded into the chip somehow, but I'm really not too sure on this last point about the hard coding. Here's a thread on that program: http://forum-en.msi.com/index.php?topic=148273.0 EDIT: yes, post #564 in that thread, SVET says it's hardware limited to 1V max
  19. I used to have the 560M in mine. I believe it's impossible to run the card higher than 1V, and mine was already at 1V at stock. I just used MSI Afterburner (any program would do) to overclock it - only managed a 20% overclock up to 900MHz on the core, even though temperatures were only 65 degC, it just wouldn't overclock higher and be stable. Yeah, for some reason I think the chip can't be assigned to run more than 1V, it just won't accept it. If anyone knows differently, then let me know, because I'm not an expert on this topic.
  20. Haha, that's good! I googled that 'Forced Air Laptop Cooler' - that is quite extreme, reminds me very slightly of a setup on my own car! Would be interesting to see some pics of your setup, what kind of difference does it make to the temperatures?
  21. With notebook cooler, although the significance of the low temperature is down to the cooling system in M17xR3, and doing a pretty good job with the pasting & padding. - - - Updated - - - Can you not run NV Inspector with admin rights & use the overclocking part of that program to increase the voltage?
  22. Well, have a look in my sig! It's a massive gain, about 87%, but the cooling system in the M17xR3 is fantastic and the GPU only gets up to 65 degC max, that helps in being able to overclock & overvolt. If your cooling system can't keep it cool, then that will be your limiting factor, otherwise you should be able to get a good overclock, maybe not 87% (maybe better, maybe worse).
  23. I have a different system to you, but with a 670MX, using svl7's modified VBIOS. Overclocked & overvolted to the max, when gaming my whole system only uses up to 160W (as long as the battery isn't charging - that can add 80W). So, I would think you'd be alright with your 180W power supply while overclocking & overvolting.
×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use. We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.