4850 Xfire Benchmark suite - what you guys after?

Mmmm, the disc didnt work for me either; I ended up making a bootable usb, and then putting the files from the iso onto it, which worked fine. Do you want me to send you a zip of that including the hidden files and see if that works for you, if you format as fat32 etc.

 

Yeah if you wouldn't mind Alex, that would be sweet!

 

Try formatting your USB stick as fat32 (quick format), and then copy every file from this zip onto the drive into the root folder (AKA not \clevo, just F:\ for example. If you hammer F10 as your M17 loads, you should be able to chose the drive as a USB HDD.

http://rapidshare.com/files/263822862/usbdisccontents.zip.html

If that doesn't work, follow zfactors bootable USB drive instructions, up til you'd used the HP programme and his linked bootable ME files to make a USB boot, then copy everything across again; if it asked to replace any files (which it should; the 3 hidden system files) then SAY NO, and keep the ones you just made, but copy everything else across. Then trying restarting and hammering F10 again.

 

Ok cool, thanks Alex. Now I just have to find a usb stick, I left mine at work. Unless a usb WD harddrive would work just this once?

 

What rock were you living under when ATI announced their first 40-nm GPU?

It never made it in to the ASUS K5 because of bad production yields but still you should've heard something somewhere about the Mobility 4830/4860 and desktop 4770.

 

heard about 4770, not 4860

 

Same GPU: RV740.

Mobility cards were launched first but TSMC had lousy luck manufacturing the GPU and it looks like all the working ones went to desktop cards instead.

 

I see, well mxm 3.0 ati cards excite me regardless....haha

Look what they did to nvidia GPU pricing! It is almost normal!

 

mxm 3.0 have a tdp of 75w max. i just researched at least on the m17x. so hopefully newer gpus will be at a max of 75w or lower.

 

It probably would if it was running FAT32 not NTFS. The M17 calls the larger USB sticks USB HDD from BIOS anyway. You might have to be willing to temporarily format it though, incase you have to go down the bootable usb creation road because my files dont work for you.

 

With the die shrink to 40-nm the TDP isn't a problem. The GTS 260Ms for example have about half the power consumption of the 55-nm GTX 260M.

 

if it is half the consumption...lets see twice the transistors!

haha

 

Why would they offer such a large boost all at once when they can trickle out the performance increases over a couple of years and milk the profits.

 

Sometimes having the fastest chip can be worth it simply for prestige alone... a lot of non-tech savvy buyers will buy on the strength of strongest chips (oh, they;re really good at the top; they must be good then), not to mention, if you have the absolute best; well you can charge a pukka premium for it; it might cost 10% more to make and would probably still sell for 50% profit, enthusiasts WILL pay for absolute performance if it's 'worth it', if it's half as fast again as anything out there; there will be people willing to pay the cash.

This is one of the key reasons I went M17 instead of M17x, as the power/price differential wasn't worth it to me. I don't need a quad, and for a similar spec Im running £400-500 less; especially with the price drops.

What I would really LOVE to see is 5 series mobile parts being released on a MXM2.1 part basis. If the cards themselves are similar TDP, then there's no reason they couldn't, and PCI-E 2.0 8x isn;t exactly lacking bandwidth.
It'd also give ATI a bit of a coup when it comes to mobile graphics; as there's a far wider range of chassis compatible with MXM 2.1 than 3 atm.

Of course this is all theoretical atm; I'd love to see real indications of the real power improvement of the 5 series compared to the 4. I mean the 4 series is about 20-80% *AA dependant* faster/efficient than the 3 series architecture; the 3870 crossfire vs 4850 crossfire is already demonstrating that; as even with CPU bottlenecking, a 4850 part running at 500Mhz is consistantly 10-15% faster than a 3870 mobile part running 620Mhz, and is more efficient with AA. (yes I know that's a generalisation; and I haven't checked relevant stream processor counts in comparison with the 3 series chip, the 4850 is identical in that regard to the desktop part with 800)

Saying that, lower NM should mean theoretically even if they don't go for larger dies, they should be able to go for higher clocks on a comparibly featured part, within the same thermal output range.

 

I know, was just joking lol

It is quite ok with me that we still get more performance and less power usage/heat - even if they want me to sell my children to have it......

edit: and sadly, Radeon HD 5000 series on mxm 2.1 will NOT happen...that is stepping backwards....MAYBE the lowend stuff, barely better than integrated, but nothing high end. The M7740 FirePro is MXM 3.0b.....it is 800 sp and 1 GB GDDR5 (3 guesses what the GPU really is....haha)

 

Hmmm, sounds like it could be my desktop 4870 pretending to be a mobile part with lower clocks

Actually GDDR5 isn't a major loss. People with 4850s/4870s were only attributing about a 5-10% performance rise to it at comparable clocks [700/900 (1800) vs 700/900(3600)]; its the core clocks that really raised the bar with the 4 series architecture; as the architecture seems to scale almost exponentially with core clocks. If you check out some reviews of the Atomic 4890 running a 1GHz core compared to a stock 4890 you'll see what I mean, sometimes its going from figures like 60fps to 85, just on a 20% approx clock rise. The DDR3 may have held that back a bit; but part the reason I think they went GDDR5 was cost/prestige/availability.

 

Agreed, the GDDR3 (need the G, is actually a different memory architecture) was only a slight bottleneck if that. The additional SP (320 to 800) and higher clocks really made the performance difference.

And sadly, there have been massive GDDR5 shortages, thus the disappearance of 4770 desktop parts (all out of stock)

 

Sorry there, typo.

I think the dissapearance of the 4770 part was actually down to the relative failure of TSMC's 40nm process, it was leaky and produces a lot of waste; thus increasing the cost per core of the 4770? 4870/4890 parts are in abundance and they both use GDDR5, so whilst there may be some relative shortages; I dont think that's the reason for the missing 4770 SKUs.

 

hmmm, was what I had heard at other sites, I hadn't heard about TSMC having trouble. The new Foundry spinoff from AMD is supposed to have a very promising 40nm process though - heard even Nvidia is looking into them, so Nvidia and ATI chips coming from the same fab haha

 

TSMC's poor 40nm is the reason Nvidia are considering using Global Foundries (and has also been the reason for some other nm shrinking problems; its attributed to why the 2900XT was comparatively hot and lower power than expected; the card was intended to come on a lower nm with higher clocks, which TSMC messed up at the time; largely because their R&D was minimal and just enough to keep up, rather than to improve and innovate) - can you see them using an AMD owned Fab if they had another option. The ultimate irony in this is that if they do switch to GF and its much higher R&D/standards then AMD will make money on every dedicated GPU sold on the home market; be it ATI or Nvidia, allowing them to shovel more money into continually produce comparable/better technology than Nvidia...that's gotta sting as a catch22
I bet GF can't come fully online, fast enough for AMD!

 

It's good to be AMD haha (although they only own like 40% of GF.....Kuwait or some middle east company/country has the majority I believe)

If Intel wanted some extra change, they would start contracting out their fab abilities, as I do not think anyone would argue they are inferior haha (32nm GPU's anyone?)

 

From what I've read Global Foundries won't be ready in time for 40-nm fab and will go right to 32-nm or 28-nm.

http://www.fudzilla.com/content/view/14811/34/

Also ATI believes that their next batch of 40-nm GPUs from TSMC will have better yields from TSMC because of what they've learned from the RV740 GPU.

 

Interesting times a'coming
TSMC really need to pick up thier game though, they've mishandled a lot of nm shifts in recent years; and if they don't tangibly repair this issue; then they'll struggle to compete with GF once they get going.

 

Well, it's been a bit of a rollercoaster since I came home to my new chip, out of excitement and habit, I immediately discarded the thermal pad from the northboard and got pasting... DOH!

Initial X9100 temps were...extreme shall we say. Maxxing at 94 degrees on the CPU and 102 degrees on the NB!! You could actually see minor throttling coming into play, as the time taken to calculate the IBT run would increase slightly as the chip got hotter

Not happy with this I gently took apart my M17 and reapplied a new thermal coating; being a lot more precise and restrictive with the application. Also to prevent air bubbling, I did the CPU heatsink up in a slightly different ordered to the recommended, tightening the northboard screws 5/6 up first of all, to ensure as best contact as I could get without buying a new set of thermal paste.

Max NB temp of 92..much better. CPU hitting a maximum of 82 degrees, again much better. Given these were IBT temps, I was relatively happy...so proceeded to up the multiplier.

Soon I was happily up and running at 3.32Ghz. Maximum NB of 99 degrees, and maximum CPU temps of 88 degrees; high, but I was fairly confident no other everyday programme would ever push the NB/CPU temps quite so high!

I then proceeded to run some re-benchmarks, and well...what can I say. A new Challenger approaches. An M17 with this setup seems to be leapfrogging 260M SLI and doing its best to get within a stones throw of the 280 SLI; which is absolutely brilliant; this laptop has cost me £1900, including ALL upgrades. When you consider I could get it for £200-300 LESS now by re-purchasing and not including Bluray as the base price has died; Im very happy, my performance is what you'd expect to see from a reasonably well equipped M17x, and has cost me FAR less, especially when you consider the incredible price gouging going on in the UK laptop market.

An M17x in the UK with stock parts; and an 1920x1200 screen, BluRay and GTX280M SLI would cost £2548 as a comparison, mine cost £650 less, including buying my HDDs, CPU upgrade and RAM upgrade.

For performance, I will let a few screenshots speak for themselves.







I haven't even begun looking into re-pasting and overclocking my GPU setup, which I may do in the near future once I've worked out how to safely remove the crossfire cable without damaging it; and I find a programme that can monitor temps for both cards without keeling over. I am using OCZ Freeze atm.

I haven't got round to reinstalling all my original benchmarks again on this Vista installation yet, hence the slightly different mix

 

your temps are way to high...my OC'd qx9300 never got that hot

nice vantage gpu score though!

 

All luck of the draw with QS chips I guess I've cleaned the heatsink assembly and chip thoroughly, and made sure not to go silly with TIM application, not to mention this is IBT Linpack 64bit...pretty much the hottest thing known to man with a current 64bit OS and CPU.

Average CPU idle is mid 40s. Most other things will put the processor into the 60s-70s, which for a 3.3Ghz mobile processor, I can live with.

The NB temp rise seems about average for most people with a pad (10-12 degrees above CPU) so Im not entirely worried on that area either. I think my X9100 is just a hot one!

 

I got my max temps using OCCT, which is just as hot as IBT.

Pretty sure my CPU was never going over 80

 

Dang, it is hella hot n all, prehaps some modifications?

Project freeze showed that ram heatsinks worked well

 

My temps on my Qx9300 at 3.06 are max 78 on the NB with no cooler . I do have a copper shim though as there a decent gap between the NB and the heatsink and paste alone may be not filling it enough and to much can be far worse since its just smothering it . for now i would put the pad back on till you get something else to fill the gap .

 

also making a hole so the cpu cooler actually gets some air

my temps were with a hole
helped a ton

 

@ Scook, depends on if you were setting OCT to 100% or default stress linpack

@Quad, the NB seems to be operating within normal tolerances, unfortunately until I can get some new thermal 1.5mm pad, I cant refit it, to see whether the increased gap is causing some issue with the surface mounting of the X9100, the old pad got damaged removing the cooler (hence I cleaned it off).

Prime95 64 bit seems to get my CPU up to about 80 degrees, so again makes me suspect either some really bad core contact somewhere, or simply a hot x9100.

Im trying to get some fresh pad to see if it makes a difference, but Im under the suspicion my X9100 is just a hot one; considering it idles in the low 40s.

I dont particularly want to drill new holes in the case incase I do have to reassemble and send back to AW incase of any warranty faults in the next year.

 

then just run it with the bottom cover off is even better!

 

Anyone happen to know if 1.5mm is the thickness of thermal pad I should be looking at for putting over the northbridge, or even tried themselves and found the best thickness, looking at the original stuff, Im guessing 1/1.5mm? I don't have a hardware store local that I can get copper from

 

Wondering if this product will actually work on the NB.

http://www.overclockers.co.uk/showproduct.php?prodid=TH-001-AK

 

ask Zfactor to try to get some OCZ whitebook bottom panels he may be able to get a extra one for you ( for a price or coarse )

 

Far too thin unfortunately

I've ordered a couple of 1-2x1-2cm thermal tape strips (1/1.5/2mm) from ebay to see if I can get one that will work spotlessly, and see if that reduces temps, as I'd love this X9100 to run cooler; but am trying to resign myself to the possibility it might not!

 

Hm, couldnt you cut some pieces the same size and put them on top of each other to make it thicker ?

 

Would rather not risk getting air caught in it (its an adhesive sticky pad) and it not working efficiently.

Hopefully I'll get the pad I've ordered by early next week and have a play; OCZ/application shouldn't be giving me gyp, so Im hoping its just a contact issue with the northboard pad gone causing the CPU core to not make full contact, or a hot dual which I may have to resign myself to.

 

Getting the Heatsink to lay perfectly flush can be a bit tricky and if its not the temps will be all over the place as i had the same issue with the Qx9300 . What i did was put the heatsink on then wiggle it round till you hear all four screws lock in there respective holes and at that point crank it down.

 

So you basically laid the heatsink flush on top of the processor; then wiggled slightly til the processor holding screws clicked slightly and stood straight? Sounds like the fun I had with an old Zalman heatsink I was using a few years back ><

How much did it affect the temps with your QX9300 out of interest, if I wasn't for work tomorrow I'd probably be trying that now

 

Yep exactly .

It was a 10-15 degree difference . The real way to tell is to just open up HW monitor first thing after boot and just load up something like 3d06 , you dont even have to run the program but what your looking for is a massive temp spike and if you see that then you know its not flush .If say it gos from 40-60 or higher instantly you know somethings up.

 

Seems there could be something in what you say from just trying that.
From 41-42 degrees idle, both cores jump to 51 loading 3dmark06 (the system detection process before going to please register and finally the actual program) with the first core spiking to 60 degree.

Im gonna hope that means my X9100 hasn't seated properly with the HS and might still be saved (cooled further), perhaps I focused on getting the NB well covered this time round :/
What I'll probably do is try taking the CPU out of its slot tomorrow night and then replacing it, incase it isn't sitting exactly flat either, gives me a decent chance to clean off the TIM from the parts as well, Im not even using that much (bit larger than a pinhead) to ensure I dont apply too much; which tends to cause issues quite quickly.

 

Yeah its more then likely just seated badly. Also redo the thermal application everytime dont just pull it off and set it down again . I reseated mine so many times i cant even begin to tell you while working with that copper mod and i tried all sorts of things so im just giving you advice based off the experiences i had .

 

Hopefully! Shame I dont have a thermal pad to lob onto the NB atm; would probably make this a lot easier as I wouldnt have to worry.
How thick is the gap between the NB/Heatsink in your estimation? Im trying to work out which of my incoming pieces would likely fit, having ordered 1/1.5 and 2mm thick pieces

As I've said before Im more used to working with desktops, and in my experience they seem a little easier to mount haha
I never reuse thermal paste; fresh application every time.

 

TBH i dont know the thickness of the piece i have since it came with my machine . Its very thin though like probably as thick as a US dime maybe even thinner . i would have liked to have tried out some thicker ones though just too see if there were differences but i havent yet ,i got the stuttering pretty much nailed down so i hate to even mess with it at this point.


Im glad to hear you redo paste everytime as many do not and i never ever have had good results from it is all i can say.

 

Hmm, seems this chip is gonna be an a** to cool properly; just tried reseating on my lunch; exactly the same temperatures, and I've now got a lovely nick in my processor where the mirror finish has been marred slightly by taking the heatsink on and off

Something tells me my cooling is either not completely flat, isn't making proper contact; I have a duff/hot chip (which is believable as it idles at 34-45 then rockets to 80s under load, surely it'd idle hotter if there was no contact) or its the northbridge causing some issues somewhere without that thermal pad.

Sods law or what? Im not even using much thermal paste incase thats the issue! I get the feeling my cpu mirror finish is gonna have lost that finish by the time I'm finished

 

read project freeze for more info about cooling. i will update the first post later this week. it's not organized.

 

Well had one hell of a reseating session, and it seems to have paid off. My maximum NB temps have dropped into the low-mid 80s, from low 100s, and my CPUs at full load are floating just under 70.

Not a major improvement, but Im reasonably happy with it; especially as this was at 3.32Ghz not stock 3.06

 

look at the CPU fan and then the placement of the CPU fan vents ....

 

I know, made me laugh a little; that said my temperatures were similar with the lid off, so given how many times I've reseated this damn thing, I dont really think Im gonna get much lower temps out of this; unless the botched thermal pad I made for the northbridge and the OCZ freeze settle in a little; which Im not expecting.

Still 72 ish tops CPU and 83ish tops NB is a far cry from the original 88 CPU and 103 max NB, I really just want to enjoy this chip now instead of fretting about it haha...Im half of a mind to test again at stock clocks, but I think I'd rather save myself from another day of getting frustrated with trying to get contact absolutely perfect, which I dont think is going to happen as my CPU heatsink screws dont even seem to want to go in 100% straight, and seem to angle slightly every time!