well I'd say I get around 45fps avrege at 2560*1600 all max, no aa with the occasional dip to 30fps and above avrege to 60fps.Quote:
Originally Posted by B.E.E.F.
Printable View
well I'd say I get around 45fps avrege at 2560*1600 all max, no aa with the occasional dip to 30fps and above avrege to 60fps.Quote:
Originally Posted by B.E.E.F.
:D say hello to my new friend
http://img379.imageshack.us/img379/995/111kb6.gif
Seems like cats 8.9 broke GPUz clock readings. The main page shows the weird clocks but sensor tab shows the correct clocks.
someone at techpowerup needs to run the debugger on their code :fact: :DQuote:
Originally Posted by NapalmV5
modded BIOS? i get the same thing.Quote:
Originally Posted by SNiiPE_DoGG
btw, have any good MANUFACTURER BIOSes been released with a better fan profile? i'd really like to go back to an official BIOS instead of an RBE-modded BIOS because i think my RBE BIOSes are causing problems. i don't want to use the Asus Top BIOS because i'm unstable at 815 core :shakes:
Nope, I'm not 100% sure I checked if it worked in cats 8.7 but it sure worked fine with the 8.6s installedQuote:
Originally Posted by xMrBunglex
post vcore mod/stock cooling
http://www.xtremesystems.org/forums/...4&postcount=63
rock stable @ 900/950 vcore mod/stock cooling
http://img156.imageshack.us/img156/6883/900te0.gifhttp://img71.imageshack.us/img71/992/901zf0.gif
edit: soldered 2 x 4.7k fixed @ mvvdc/mvvdq
now memory stable/comfortably @ 950
I am using a 4850, just thought I would mention that gpuz didn't give me those messed up mhz readings until I pencil modded. I think it might have to do with the vmod, not the drivers.
Hey,
ive just fitted an EK full cover block to a powercolor PCS+ 512 card.
all 4 vrms show up in gpu-z.
idle gpu is 31. and idle vrms are 44 to 47.
under gaming load GPU temps are less than 40. vrms under load ( playing crysis ) they are in the early sixties.
the Benchmark_GPU.bat in crysis gets to 44-48 and the vrms 68-76.
gonna get some thermal epoxy and add some extra cooling to the vrm side of the cooler. il post the results if its successful.
P
This is what I've got going for now. This is the first OC I tried right off the bat. More will come later when I receive my XSPC Razer full cover block and I start messing with the bios. ;)
For know I'm cooling my card with the MCW60 and the stock heatsink's metal back plate for memory and VRMs (not using stock fan, though). OCed with AMD GPU Clock Tool, all stock voltages.
Furmark + ATI Tools to check stability. :D
http://i59.photobucket.com/albums/g3...B/bf4e28d2.jpg
You need to rename furmark to be able to run it properly. ATI crippled it since 8.8. As it is now this does not give you a complete picture of stability unfortunately, though ATITool being artifact free bods well :DQuote:
Originally Posted by Makubex_GB
Edit: Apparently the reason for renaming it is because "Catalyst 8.8 have been optimized to detect FurMark and downclock the frequencies of a Radeon HD 4850/4870 to avoid to burn the GPU". But as you can see AMD GPU Clock Tool seems to overwrite that problem since my clocks didn't go down.Quote:
Originally Posted by dejanh
No it does not downclock the card, it just cripples the program. That description is incorrect. Rename it and run at 1280x1024 as a minimum.Quote:
Originally Posted by Makubex_GB
I did renamed it and FPS was the same. And please, describe how does it "cripple" it then.Quote:
Originally Posted by dejanh
It doesn't need to be 1280x1024 minimum. What makes you think that? How is that a requirement?
I'm running two programs stressing the GPU quite a lot, Furmark at 640x480 + ATI Tools is enough to test stability and is more that what most people do to test it. No need for 1280x1024, even if I did change the res the result is the same, just lower fps.
Driver limits the FPS. For specifics ask ATI, not me. One thing it can do is simply force the GPU to render every second frame only and that cuts the FPS in half and the load by 50% :) In terms of resolution not mattering, you have got some major stuff to relearn about graphics it seems. By your logic running games at 640x480 vs. 1280x1024 would be the same. Think about it ;) Anyway, if you got off on the wrong foot this morning don't take it out here or on me. I simply pointed out a test for you that will really push the limits of that OC. If you do not care to try it, don't, I got nothing to lose.Quote:
Originally Posted by Makubex_GB
Also, if you are running Vista that explains the FPS not changing. On my system in Vista 64 FurMark runs bad. FPS are always low so they seemingly stay the same all the time, no matter what res you are running. In XP it is very different, resolution to resolution.
Again, try it or don't, I really do not care. I was simply trying to make a suggestion on how to test.
I'm not saying that in games running a lower res is the same as a higher res, I'm saying that in this case, I am already running two GPU intensive programs at the same time, hence the low res is fine. I'm not taking anything out on you, I'm just asking where do you come up with stuff like "1280x1024 minimum", as if there was some sort of rule. If I was running Furmark alone, then I would understand, but you still don't seem to get that I'm running Furmark and ATI Tools at the same time. I don't need to run it at at high res because I'm already getting a high enough load on the GPU. If I go by what you're saying then what the hell I could just come off and say, no it's 1600x1200 minimum, where did I get that? It's a rule I just made up. Furmark at that 640x480 + ATI Tools is good enough, no game is going to stress it constantly so much and like I said before, most people don't even stress test their GPUz that much and call it stable. I'm running XP and I admit that FPS went up, at first I thought it didn't because it looked the same, but then I discovered it was because I was opening GPUZ and while GPUZ opens, it makes the FPS go down. Still even after renaming furmark and upping the res, the load on the GPU shows to be the same on GPUZ, and results are the same.Quote:
Originally Posted by dejanh
Just because I didn't set the highest possible stress test you can think off on the card doesn't mean it's not good enough to test stability. I could set Furmark at 1600x1200 and run ATI Tools and then someone could come and say, hey but why not run a second instance of Furmark to really stress stability, but that doesn't mean that the way it was previously being tested wasn't good enough. That's all I'm trying to say.
Dejanh was just trying to help. ATI made a profile for Furmark starting with the 8.8 driver to slow it down because people were overheating their 4870's.
Yes I understand, I thanked him for letting me know about Furmark, what I am questioning mainly is the "1280x1024 minimum".Quote:
Originally Posted by xMrBunglex
The way I came up with the 1280x1024 minimum is that I found on my 4870 that it did not overheat or crash if I ran FurMark at anything less than 1280x1024 in the stability test. 1280x1024 or higher would cause the temps on the VRM to skyrocket and the card to crash. I guess it is a personal experience only...I just went by it anyways. I do agree though that if your GPU load while running ATITool and FurMark in your setup is at or very near 100% for the whole duration that is then good enough.
Thanks for explaining. :)Quote:
Originally Posted by dejanh
I just put the FC Koolance block on my 4870 over the weekend, and my VDDC temps are in the 80s while gaming and in the mid 90s while run GPU folding client. The GPU itself is in the mid 40s though. I put a dual slot fan right under the card too running about 80%. I'm wondering if running them all day in the mid 90s will kill it or not. Looking at temps in gpuz 0.2.7
I did run furmark and stopped it when they went over 105C
Thermal maximum for the DVRMs used on the reference 4870 is 125C. Therefore, you are likely fine but if your temps are getting over 110C I recommend you look at some additional cooling.Quote:
Originally Posted by skindogg
If Im right in considering a Full Cover block covers all thermal point such as ram, voltage regulators and gpu core. He must have mounted it improperly if he's reaching over 100c on the vreg's with it or the block is rubbish.
The thing is, on the stock cooler, there's that slot next to the right edge where air is getting sucked over the vregs. However on the FC block they are just covered up, but no contact to the block.
I got a slot fan with dual fans blowing up onto the card from right below it, but I think the stock cooler probably cools vregs better than this.
I can't believe the mount was bad since the GPU temps are all in the 40s while gaming.
I guess I was curious as to whether it's ok if the vregs operate in the mid 90s all the time. I'm fine with not running furmark ever. I don't plan on volt modding either.
Just because the core is being cooled well doesn't mean it's not bent or something toward the vreg area making not the best contact.
Also I would think the coolness provided by the water and lower core temp would radiate through the block material thereby cooling the vregs better.
You'll be fine in the 90's though as someone else said there spec'ed for 125 degrees celcius.
http://img504.imageshack.us/img504/2...saisruniy6.jpg
Just first fast run , stock coolin , no mods. :up:
anyone try upping volts in the profile with positive results? my card is good for 875 on stock volters.
no artifacts in Ati tool?Quote:
Originally Posted by Rattle
Can run 3dmark @ 870 / 4500 , but artifact free core is only @ core 850
i dont use atitool, but its fur, rthdrbl, 3dmark06 and hours of cod4 and crysis stable without artifacts
06 mark is benchable only @ 850 / 4400
http://img205.imageshack.us/img205/8017/3dmark06rw2.jpg
06 at 850/1000, i can run 1100 on the mem but I dont see any difference.
I can only do 825MHz anything else doesn't work - did I get a bad card?
Also I can do 1175MHz on the memory :D
My experiences with GS are similar... GPU doesn't clock well!Quote:
Originally Posted by fatguy1992
Thx so is not just meQuote:
Originally Posted by M-XXXX
I have my second 4870 GS waiting to go in, but i didn't get a CF bridge with either card. Therefore i ordered one off ebay and it will take a while 2 be here :(.
Anyway i'm going to put it in too see what it can do :)
Probably because of only 2 power phases for GPU... I've tested one, and the upper limit was 780:mad:, and my sapphire can do 860-870:D
I'm ordering
Thermalright T-Rad 2 soon and getting vcore mod
I hope getting 900 mhz core 24/7 stable
My HIS 4870 1 GB sucks : it is cooled by a phase change cooler, at -38°C (core temperature) and in 3DMArk06, it is stable at 951 Mhz max (1010 Mhz for the RAM). And the games are stable from 895 (Condemned) to 970 Mhz (hellgate). Most of them are stable at 920 Mhz. Pretty bad since I expected a 40 % overclock like my previous 9800 GTX. Here, I obtain around a 20 % overclock. Since the memory IO is far hotter than the core (between 10°C and 15°C more), checked with Riva Tuner 2.11, I assume that the memory controller is limiting my overclock ability. Since it is a 1 GB card, the memory controller might be stressed more than with 512 MB ?
I will sell it and buy a 4870X2 instead. I don't want to buy a second 4870 1 GB and obtain such a crappy overclock again.
What kind of temps are you getting with the stock cooler? Core 750@800MHz and 67-68 at load and the VDDC temps are 112-113. The fan is at 100% speed.
edit. Ah, nvm already found one...
Well you can't compare percentage overclocks 1:1, Nvidia architecture tends to scale to higher frequencies with less gain. I think if you did a performance scale comparison you would find that the cards overclock about the same. Its a shame you cant reach for 1ghz @ benches. I think the Issue is OVP @ 1.5v on these cards although we don't want to start killing them. ;)Quote:
Originally Posted by overclock
Looks like I got a crappy oc'er again. Can't even do 950MHz on the mem without furmark crapping out and drivers giving a vpu recover. Core seems to do 820MHz but that's with the stock cooler 100% speed. Somehow I'm getting the feeling that the overclocking is much easier with the Nvidia cards...
With a 2900 Pro 1 GB, watercooled by a waterchiller (water temperature = 15°C), I had reached 900 Mhz at 1.15 V and 990 Mhz at 1.35 V.Quote:
Originally Posted by SNiiPE_DoGG
Here, I reached 900 Mhz under phase : even compared to the 2900 Pro card, it is disappointing. And 2900 cards are famous for heating much.
I've had my geCube 4870 512MB for a few months now, and just installed it in a new vista system.
I had been running the Asus top bios (gpu=815 mem=925) which got me around 14.5k in 3dMark06, default settings.
Today i used RBE1.13 to mod the bios, and flashed it succesfully to gpu=850, mem=1000. The flash worked, and after a reboot GPU-Z reported everything correctly, along with the increased bandwidth (128GB/s).
However, i ran 3dMark06 again, and now the score has dropped to 11k.
Nothing else was changed, the tests were run back-to-back (before and after flashing).
CPU is E8600@4GHz, the rest in my sig.
Anyone got any ideas as to why my score dropped so much??? :confused:
You could decrease the vRAM clock and if that doesn't work decrease the GPU clock. GDDR5 has a very strong mechanism for correcting errors so it can clock very high but clocked too high it will be slower because packages have to be re-requested (:p:) and resend.Quote:
Originally Posted by twwen2
Right!Quote:
Originally Posted by alexio
:eek: I just learnt something, thanks!
I couldn't comprehend how an increase in clocks could possibly lead to less performance, but that makes total sense.:yepp:
It is slightly dissapointing, seeing as the ram is rated at 1000MHz.:(
Hmm, well i downclocked to 850/950, and the score dropped another ~150 points. Then i tried 820/950, still the same. These were all modified versions of the Asus "Top" bios, using RBE1.13.
After flashing back to the origional asus top bios (815/925), my score has jumped up to 15,171!:eek: That's even higher than before!!!!??? There was nocitably more FPS in every test (i'm talking an extra 25-30fps), even the two CPU tests were faster.
I'm at a loss! :shrug:
Downclocking FTW!? :confused::p:
What happens when you overclock with the standard BIOS? It could be that the performance profiles were messed-up in the TOP BIOS or that the driver had problems with that BIOS somehow. Just remember that it's not a good idea to use a BIOS with different vRAM frequencies in different profiles. You'll get a weird switching effect when running 3D benches.Quote:
Originally Posted by twwen2
Yeah i always kept the ram the same frequency in all profiles.Quote:
Originally Posted by alexio
I've never tried overclocking the factory (geCube) bios, maybe that's worth a look. Is 15k in 3dMark06 about right for a system with my specs?
I still need to do more testing, but anything over 1000MHz on my Sapphire 4870 1GB results in instant artifacts. But at 1000MHz there are none in ATi tool for hours.
I need to edit the BIOS though to get above 790 on the core, as it handles that with ease.
Using a MCW60-4870.
Just got my 4870 1GB in... :)
I modded the stock BIOS and I figgured 1200 would be enough on the RAM, but aparently not! Its stable at 865/1200 right now, i need to step the BIOS up a notch to get the ram at its limit! But the core cant go above 865, its stable for hours at 865, but locks up at 870 :(.
http://www.techpowerup.com/gpuz/9ewwv/
VMOD TIME!
I was thinking of messing with a softmod, but its probably just easier to hardmod.
Whats a good voltage for the stock cooler? Fan idles around 45%, load at 100%, i think the temps are 46 idle, 54 load. I changed the thermal paste and lapped the heatsink of course.
Your ram can do 1200? :eek:
On air?
It starts getting nasty at 1250, but it does 1245 for 10 loops on crysis. I just backed it down to 1200 since there is no noticeable gains whatsoever from 1200 to 1245, and i like 1200 better anyways. :)
Now i need to vmod that core and hope for 900+ mhz!
And yes, this is all on air. My rig is shoved in my closet, and it gets quite toasty in there after a while. About 26-28ºC ambients generally.
This card is doing well, but it still gets poor 3DM06 scores. I was doing 18k+ easy on my G92 in the sig, now im at 16kish... No cheats or tweaks, i broke 20k cheating on the G92.
Mine is rock solid stable in gaming at 850/1100..
Setting the memory at 1200 results in artifacting on the desktop, and putting the core over 850 results in freezing during game play.
Be careful that you don't kill your card, or shorten it's life expectancy.Quote:
And yes, this is all on air. My rig is shoved in my closet, and it gets quite toasty in there after a while. About 26-28ºC ambients generally.
My card runs very cool. It never passes 53c on load, which I believe is due to the open air heat pipe cooling system. The closed cooling system which is standard on the 4870 gives higher temps, but it funnels the hot air outside the case.
Open air coolers like mine push a lot of hot air into the case, but if you have good air flow, it's not a problem. My Coolermaster HAF 932 has 6 120mm fans, 1 140mm exhaust fan, and one big 230mm intake in the front :cool:
3DM is overrated.. I can't even tell you the last time I ran that benchmark :ROTF:Quote:
This card is doing well, but it still gets poor 3DM06 scores.
The best way to test for system stability is to play some strenuous games, which coincidentally, is much more enjoyable than running benchmarks :D
BTW, what brand is your card?
This shows just that one takes synthetic benchmark test better than the other does, it highlights the uselessness of synth benchmarks; except in specific version:version synth test comparisons... even then it struggles to find use.Quote:
Originally Posted by ZOMGVTEK
Ultimately the 4870 1GB overclocked has an order of magnitude more power potential and memory bandwidth, and also a later version of DX10 (???)
i had an open air 4850 heavily overclocked before and my MB temps after long gaming sessions were max 36c. now with this 4870 card i idle at that temp and will load up to like 45c. I tried to manage the cables as best so as to not block airflow when connecting the extra power cables, but do you think this card is really just generating that much more heat?:confused:
I just ran 3d06 at 850/1100 for 21k on cf asus DK 4870 1b a e8600 never going over 50*c on cards. My secret is to run with a single 120mm fan on the outside of the cards - It has lowered my load about 25*c for about 9.99$ usd. I will crank this up a bit more when i get home from work tonight. I honestly think that these cards are fairly cpu bound at this point, and i'm already running at 4.3ghz+