|| 2500K @ 5GHz 1 thread, 4.8 2 threads, 4.7 3, 4.6 4 1.284V ||
|| P8P67-M Pro || 8GB @ 2133MHz ||
|| 5850 @ 1000/1225 || XFX 650W || Silverstone FT03B ||
|| 37" LCD TV || CM Hyper 212+ || Samsung 2.1 Soundbar ||
Okay thanx! I'll concentrate on the GPU then
I'm having problems with the GDDR5 ram. It wil barely do 4Ghz stable, allthough this is stockspeed for Palit Sonic 1GB HD4870's.
With Cat 8.12 I have to clock down to 975MHz to get rid of artifacts in 3Dmark06 and Crysis .
With Cat 8.11 clock set to 1000Mhz is artifact free, but 1025MHz wil give minor artifacts and 1050MHz gives massive artifacts.
Chips used are Qimoda GDDR5 and are rated 1.0ns. In all reviews these chips do at least 1100MHz.
Am I missing something here? I have a Swiftech waterblock on the GPU, and Enzotech sinks on chips. I'm having the exact same problem with two cards, so could it be that it's not the cards themselves giving artifacts but something else?
Last edited by DemonEyez; 01-19-2009 at 03:14 AM.
Proc: Q9650 9x496 @ 1.440v batch L844B703
Ram: 2x2GB OCZFlexII PC-9200 4:5 1240MHz 6-6-6-18 @ 2.12v
Mobo: Gigabyte EP45-UD3P rev 1.0 - bios F9b
Video: 2x Palit Dual Sonic HD4870 1GB in CF (840/4400)
PSU: OCZ PowerStream 520W + Thermaltake VGA 450W
HDD: 3x250Gb Barracuda 7.10 SATA2 Raid0 (ICHR10) 2x200Gb Barracuda 7.10 SATA2 Raid1 (ICHR10)
Audio: Audigy 4
Case: Dual Coolermaster Stacker; one for system rig and one for watercooling loop and cooling
[SIGPIC][/SIGPIC]
Hello,
I vMEM and vGPU in my 4870 PowerColor (Gainward BIOS), could not stabilize it in any way, I can go up to 890MHz Core ATiTool without the artifact (30mins) and memory up to 4550MHz (30mins), plus you can not even 840/3600MHz 3D Mark 2006 or any other game, what may be happening?
Last edited by 8800's@over; 01-20-2009 at 05:29 PM.
| Core i5 2500K + NH-U12P | Gigabyte P67A UD4 | 4096MB G.Skill 1600MHz | SLi NGTX 480 M2D15 | WD 500GB Raid 0 | SevenTeam 850 Z-AF |
AtiTool is out of date for testing artifacts. I use 3DMark06 and Crysis for testing overclocks, Canyon flight and Deep Freeze are most sensitive to GDDR5 instability.
I test on the highest settings possible 1920x1200 8xAA/16XX so I know that the full 1024MB on my card is used.
Proc: Q9650 9x496 @ 1.440v batch L844B703
Ram: 2x2GB OCZFlexII PC-9200 4:5 1240MHz 6-6-6-18 @ 2.12v
Mobo: Gigabyte EP45-UD3P rev 1.0 - bios F9b
Video: 2x Palit Dual Sonic HD4870 1GB in CF (840/4400)
PSU: OCZ PowerStream 520W + Thermaltake VGA 450W
HDD: 3x250Gb Barracuda 7.10 SATA2 Raid0 (ICHR10) 2x200Gb Barracuda 7.10 SATA2 Raid1 (ICHR10)
Audio: Audigy 4
Case: Dual Coolermaster Stacker; one for system rig and one for watercooling loop and cooling
[SIGPIC][/SIGPIC]
Speaking of respect, you just lost any respect from my side and supposedly from many others, too.
As ceemic already denoted, you decided to risk killing your card and you think it's the manufacturers duty to replace it, free of charge? The whole blah about how they decide to process the RMA or not is just miserable. If you were honest for one moment, you would acknowledge that you wouldn't even have to send it in if you didn't expect them to replace it.
IMHO, they should charge you a penalty fee for trying to screw them. The problem is that they know that you could still hurt their reputation by posting online how bad the company handled your perfectly justified RMA, i.e. plain lying. That's why most companies process the RMAs anyway, although they know it's not covered by warranty.
If I had sent in any piece of hardware that I killed due to my own stupidity or carelessness, I would have a additional system standing right next to me.
The thing is just that it's not the way you should act and you as a member carrying the title "champion" should know better. It's pathetic.
I mean if you can't afford to lose a certain piece of hardware, or you don't know what you are doing when performing vmods, then just leave it be.
It's one thing if a customer doesn't realize that he himself killed the piece of hardware, but a totally different if he, just like you, knows exactly that it was his fault.
Quote from one of our professors:
"Reality is hiding in the imaginary part."
I've got my previous 4850 killed (though I'm not sure, it was memory chip corruption and I've only VModded and OC'ed graphics chip, it can still be my fault) and I've RMA'ed it.
I'm not proud of it, but i didn't have money to buy a new card instead(even the cheaper one). It's actually cheating on manufacturer, and I'd rather buy myself new card and don't even worry about it
if I earn more money. Thare's nothing pleasant in waiting a month while they process my card and thinking if they will replace it at all, or I should go and by myself HD3450.
But you can also consider this as a service. How many cards do they replace a month all over the world? 300? 400? Don't you think the cost of replaced cards is included into the cost they sell cards to dealers at? So you just use the service you had paid for, just like in case with insuarance, when you pay for what may or may not happen in the future.
P.S. All above IMO. And sorry for my English. Just couldn't resist posting.
Please,
Who knows what the max Safe Voltage for GDDR5 24 / 7?
4800MHz day can kill you?
| Core i5 2500K + NH-U12P | Gigabyte P67A UD4 | 4096MB G.Skill 1600MHz | SLi NGTX 480 M2D15 | WD 500GB Raid 0 | SevenTeam 850 Z-AF |
Hey guys, just want to make sure. This is what I need for the VGPU mod right?
http://www.westfloridacomponents.com...tor+3106Y.html
EDIT: If I used a 33 Ohm fixed resistor, would the voltage be too much for 24/7 use? This is the resistor I would like to use. I don't like VRs.
http://www.radioshack.com/product/in...odsInSession=1
My card is a reference 4870 512Mb btw.
Last edited by Gir92; 01-21-2009 at 08:54 AM.
I hope you can find out what's causing it.
My 512mb card has no problem with 1125 on the RAM (825 gpu, 835 ran crysis (no lockup) and 3dmark without lockups but i don't trust it, since i can't loop 2006 without registration). But from many of the posts I've seen, most of the 512mb cards can usually hit 1100 or higher while the 1 GB cards start failing at "spec" memory (1000+?). I thought only the hyundai chips failed at 1000?
But yours has qimonda....
You keep asking for people to help you, and you're using a risky trimpot, and wondering why you aren't stable, and they said get a multimeter and measure the voltage, but you said you don't have one...
That's why no one is helping you...we don't know what voltage your card is getting!
So go buy a digital MM please....
Ceemic, here's the pic:
http://i42.tinypic.com/23t3txg.jpg
DuncanOc, what's the marking on that IC? NCP....?
Vmods - what can possibly go wrong? ©
^ 33 ohm resistor perfectly fine for 24/7
Thank you napalm.
Also, I've only done one volt mod before, so I'm not an expert.
Can I use any side of the fixed resistor for ground, or does it have to be a certain side?
DuncanOc, You have to look where pin17 heads to. Mark the trace on Your picture. As I said before, You have to find a SMD component, that is connected to GND.
Vmods - what can possibly go wrong? ©
Also, the 33Ohm resistor from radioshack is only rated for 1/2"
Reduce the flow of electricity in your project with this 5-pack of 33 ohm carbon-film resistors. 1/2-watt, 5% tolerance.
Max Working Voltage: 350V
Lead Length: 28 ±3mm
Temp Range: -67°F to 158°F
What's in the box
33 Ohm 1/2W 5% Carbon Film Resistor (5)
Is this bad? Will I only need to use one resistor, or more in series? Does a certain side have to be ground, or can it be either? Thanks again.
^ either side works
just one resistor
and get the 1/4w 33 ohm
1/4w is more efficient.. better resistance
and your 4870 will perform better
sure your welcome
good luck!
Alright one more question. :p
Is it possible to overvolt and kill the memory on a 4870 from JUST doing the MVDDC and MVDDQ Pencil Mod here? I'm only wanting a slight increase in voltage.
Thanks again.
EDIT: Also, if I can get to 4.4Ghz effective stable on the memory, is there any chance of killing it just from the speed? I've heard stories of people's cards dying at 4.4Ghz.
Last edited by Gir92; 01-22-2009 at 01:31 PM.
Bookmarks