excellent. ty
Printable View
excellent. ty
I flashed my Asus HD4870 with the Asus TOP bios. Runs 850 core 24/7 but no problems up to 890Mhz (highest the ASUS TOP allows) core stock, no vmods done at all.
Cooled with MP-01 and some VRM sinks with a 120mm blowing over them.
Here was a run at 850 mhz,
http://3dxtreme.net/other/ASUS%20HD4...870flashed.jpg
http://3dxtreme.net/other/ASUS%20HD4870/img0353tz1.jpg
dnottis - aren't those memory heatsinks a bit too low-profile? And what are the top clocks that you get on memory with those 850GPU? Just those 1000 or can you play with 850/1100 or something like that?
Also, those black heatsinks on VRMs, are those 3 pieces of these put one by the other -> http://www.swiftnets.com/products/mc21.asp
Seems like a lot of people that do water cooling use these for VRMs, so can you post your VRM temperatures during some heavy load (like Crysis benching, AtiTool, or something similar - I won't say Furmark on purpose :D )
what are you using to test stability besides 3DMark? if that is truly stable, you have a special little GPU.Quote:
Originally Posted by dnottis
I can do 1100, but the extra memory bandwidth does nothing in benchmarks. Why kill DDR5 when there is already more than enough memory bandwidth. choose 1000 to make it a nice even 4gb/sec memory bandwidth.Quote:
Originally Posted by LuxZg
Yea, those are the Swiftech mosfet sinks from the 8800 series. I thermal taped them together and installed them as a whole one piece heat sink over the VRMs. A low speed yate sits over the top.
My VRM temps during Furmark hit tops of 84c and thats at 860mhz. I can game and benchmark at 890mhz including COD4, Crysis, Stalker, Grid, etc, haven't tried Furmark at 890 mhz though yet. I will just to see. With my setup the core doesn't break ~38deg c. Using Cats 8.8 Furmark doesnt push the VRM past 85deg c.
As for stability, I ran 3dmark but as everyone else gaming is the real stability test. Yes, its completely game / bench stable at 890, but I run 850 or lately 860 24/7. I flashed this card (Asus non TOP) with the Asus TOP bios and this thing has proven to be a monster. The first HIS HD4870 I had died at stock settings after about 2 weeks. But the Asus has proven to be a crazy overclocker.
Thanks for the details, it seems I'll have to find me a way to get to those Swiftechs somehow :(
Are you volt modding at all 'cause it does not look like you are...Quote:
Originally Posted by dnottis
nope, stock in that respect.Quote:
Originally Posted by dejanh
Here is a few mins at 900mhz, again no vmod just flashed to Asus TOP bios. That is the max of the CCC with this bios.
http://3dxtreme.net/other/ASUS%20HD4870/900core.jpg
that's awesome. i'm jealous.
crazy!
I can do 850mhz stock volts, stock bios, using the Ati traytools. Temp around 47º.
Powercolor hd4870
dnottis - you've said "Using Cats 8.8 Furmark doesnt push the VRM past 85deg c." Have you renamed furmark.exe to something else? YOu do know that new driver cuts the load on Furmark for roughly 40-50% ?
Doesn't matter anymore. As of yesterday the card will no longer boot, just up and died one morning. SDS (sudden death syndrome) hits another HD4870. These cards are garbage, first one was bad off the bat, this one died 3 weeks later.Quote:
Originally Posted by LuxZg
I'm going to an Nvidia GTX 260 this time.
Quote:
Originally Posted by dnottis
u should go for a 4850 gainward golden sample
dnottis - sorry to hear that another one went dead :( Seems like watercooling is not a good idea on these cards, unless you've got a water block with complete card coverage..
And this is just one more reason for me to stay away from Furmark in general, I'd say it is just not healthy for these cards.
Already have another solution...
http://www.xtremesystems.org/forums/...02&postcount=5
Why..no Furmark? :D
my memio is 15c higher than the other gpu sensors..... The 4870 pcb is so weak that the Ek full cover block never sits right. It bends a lot.
I ran Furmark, VRMs get about 80c on the GTX 260. RTHDRIBL is more on par with the temps I see in games though.Quote:
Originally Posted by LuxZg
Can someone, with Sapphire HD4850, post their max core and mem OC with no hardmod but air cooled? Mine was 760mhz core and 1160mhz memory when i passed Vantage Performance run. Vantage gave graphics driver errors on 770 mhz core speed. I haven't tried out the max memory OC. I have AS1 as a heatsink with 2x120mm@800rpm. Temps dont go over 45C in Vantage.
Quote:
Originally Posted by Viilutaja
30ºC IDLE / 39ºC Load @ Furmark
@840Mhz/2000Mhz
http://img512.imageshack.us/img512/1909/vga1rr1.jpg
http://img294.imageshack.us/img294/9203/vgatempfp3.jpg
http://forums.techpowerup.com/showthread.php?p=879319
This VMOD is for red PCB, but I have blue one. Two pictures of the problem:
http://shrani.si/t/39/10u/kg0SaKK/vddcsolder.jpg http://shrani.si/t/37/1n/4yNWBI53/img5012.jpg
Is there VMOD for blue PCB? Or does anyone know the bios who would raise GPU voltage?
someone how much voltage the 4850 Core can handle ? i got one and in 1.5 she got sic forever :p
I've got my reference sapphire at 820Mhz with 1.44vcore 24/7 for several months without a problem. I have tried up to 1.6 vcore for benches without a problem. I couldn't get it 100% stable at 840Mhz no matter what voltage i used, even at 1.6 it wasn't 100% stable. The only problem that i have found is that at high voltage during furmark, when i'm not stable (for example when i was trying for 840) and i got artifacts, these artifacts wouldn't leave until i shut down my computer. After that everything was normal. With just a reboot the issue wasn't resolving and i don't know why.
Well, I just picked up a brand new Palid HD 4870 Sonic Dual Edition and I am trying to volt mod it. The voltage seems to be determined through a VID table and is controlled by NCP5388 (http://www.onsemi.com/pub_link/Collateral/NCP5388-D.PDF). The phase controller is located on the top side of the card, but it is connected to resistors on the back side arranged in two rows. The rows have 3+4 1k Ohm resistors soldered on in the top and bottom row respectively. These look to be pullup resistors however so I have no idea how VID is determined currently. Nothing seems to match the tables. Anyway, I took the DMM and measured the voltages to be 1.245V idle and 1.263V load. And that's about as far as I got. I managed to map connections on the bottom to the pins of the NCP5388 and they map as follows (I'm using the word bridge here to indicate the small section where two connected or unconnected solder points are where resistors could be added or are already present):
Pin 2 (VID0) maps to 1st set of bridges in both rows.
Pin 3 (VID1) maps to 2nd set of bridges in both rows.
Pin 4 (VID2) maps to 3rd set of bridges in both rows.
Pin 5 (VID3) maps to 4th set of bridges in both rows.
Pin 6 (VID4) maps to 5th set of bridges in both rows.
Pin 7 (VID5) maps to 6th set of bridges in both rows.
Pin 8 (VID6) maps to 7th bridge in the lower row.
Pin 9 (VID7) maps to the complete set of bridges in the lower row.
The pics of the card from the front and from the back are attached to this post (I took these from the web as my camera sucks for closeup shots)...
Help please :)