Printable View
sorry to nitpick but this caught my eye.Quote:
Originally Posted by dejanh
1100mhz in CCC = 4400mhz RAM clock
400mhz is a huge OC, I only ever run my cards at 1125 for suicide bench runs. 24/7 at 1100 will most definitely kill your cards RAM. 1025-1050 is the safe range :up:
Yeah sorry, early morning, I did not pay attention. I originally wrote 10% OC on the memory but then I wrote in 4.1GHZ. :rofl:Quote:
Originally Posted by SNiiPE_DoGG
Sleepy. Corrected now :)
Anyway yeah, I agree with you completely. That's the point I was trying to make. The memory being rated at 4GHZ max is being abused on the XOC stock card at 4.4GHz. Especially since they give it no more juice at all.
By the way, what do you use for your 24/7 settings?
Some people are running with there ram at 1200 and plenty seem stable at 1100 :confused:Quote:
Originally Posted by SNiiPE_DoGG
is that based on anything other than speculation?Quote:
Originally Posted by SNiiPE_DoGG
Not to be a jackass, but who is running their ram at 4800MHz?
(1) Is is stable in FurMark?
(2) Is it artifact free in ATITool?
(3) Can it run lengthly 3DMark06 Batch Tests?
That person needs to step up and tell us all what they are doing exactly...
Oh and for those that want a bit of interesting reading re VRMs and temps, look here...
http://www.xtremesystems.org/Forums/...=194121&page=7
@affiliate13 - I am most curious at your results now, because the approach you took leveling the HR-09 Type 2 seems to really do a good job cooling the VRMs. Like I said before, I tried the HR-09 Type 2 couple of weeks ago with mixed results, but I did not have the ability to make it have level full-contact to VRMs (it was slanted, like yours initially :p:). I really want to see your final setup. Can you post detailed pics please?
Thanks :)
Well, it is not quite speculation Wizz. The ram is rated at max critical 2.00V with 4GHz max speed rating. You do not think that more than 10% is a big OC? I think that with good cooling you should be okay to run about 1100/1125, but anything over that you are probably slowly but surely killing your ram.Quote:
Originally Posted by W1zzard
plenty of ppl don't use atitool to test artifacting. Anyone who does know that they cant run at 1100 without artifacts. the pencil mem voltmod fixes this but if you aren't supplying more volts and adequate cooling (ie. ramsinks not redplate) then you ARE damaging at least some of your RAM at 4.4ghz.Quote:
Originally Posted by Willhub
A card with all 8 ICs binned high enough to run 4.4ghz at stock volts is a golden one for sure
sorry for double post,Quote:
Originally Posted by W1zzard
one of the first cards I had (ive had 3) I ran ATI tool for 60minutes with the ram at 1100mhz , came back the card was artifacting at desktop 2d and at boot, had to RMA
List what you would like close ups of, i will take it all off and photo it as separate parts if you think this will be of use to you. I will post them up tomorrow night.Quote:
Originally Posted by dejanh
Most of it is documented through the thread in my sig.
The most fortunate thing about the fins i cut out of the the AC S1 VRM sink was that they are slightly tapered which given the angle of the contact between the VRMs and HR09 made a nice wedge shape to fill the gap. I would have preferred something less awkward like the Type4 with bolts but this will do.
Guys this issue about how high to run the RAM, i had it running at 1130 for six consecutive runs through 3dmark06, am i wrecking it running it so high?
i had to put it up to 1150 before i saw artifacting so i though it would be OK just below that. I'll try ATI tool too and post what is stable.
I'd mostly like two angle shots of the seated HR-09 showing how well it is in fact seated. How much gap do you still have left if any? It would be nice to get a few shots of the HR-09 when it is not mounted but with those fins used for padding. I am assuming that you have permanently attached those fins now to the HR-09? Also, I'm curious if you are going with a thermal pad now between the VRMs and the HR-09 or if you are using thermal compound.
Finally, I'd like to know what kind of spring you had to use so that the HR-09 stays firmly seated on the VRMs. Do you think the spring loaded clips are enough with your final solution, or is that nut-and-bolt setup you have with what looks like a firmer spring crucial?
As for the ram, I'd put it to 1100 to be on the safe side. It gives you 10% OC but it is much less likely to cause problems in the long run. The ram is not your bottleneck anyway :)
I started to type up answers but i need the photos to correctly illustrate the little awkward spots and how i eventually got around them. its far from perfect and id call it luck that i had that spare sink to butcher up.Quote:
Originally Posted by dejanh
I will supplement my thread over on clunk to answer all of the above, i been playing COD4 for the last couple of hours and need to go to bed.
did some testing today with my VRM. My card even though watercooled and the VRM's are heatsinked (What a :banana::banana::banana::banana::banana:!) the temps were still high. Upper 80's while in 3dMark Vantage. Furmark however no matter what crashes immediately, dunno why. Someone really needs to come up with a good aftermarket sink that goes on those VRM's cuz ramsinks on those things are the biggest pain in the ass ever.
AAAAAAAAAAHHHHHHHHHHHHHHHH these ramsinks wont stick on these little :banana::banana::banana::banana: VRM's! Got so sick of them falling off I just pulled em off and left the fan blowing right on them. Temps didnt change so I guess ramsinks dont do squat for them. Still only gets in the 80's during my gaming or Vantage.
The Accelero S1 VRM's are useful as they got screw holes so you can screw it into one hole.Quote:
Originally Posted by Budwise
I have AS5 on the atm, going to change that tommorow or monday and put some thermal pads on and re do the core with MX-2.
I was told that I'd get better cooling by putting the red plate back on to coll the VRM's and ram, would that actually be better than the little square heatsink?
VRM hits 105 degrees here max in furmark, game about 80-90. TDU over an hour it hit about 56 the core did.
Quite frankly, I think there are some engineering issues with the reference 4870 design, particularly relating to the VRMs. Everyone gets excessive temperatures with them and they are a b**** to cool. I will try mounting an HR-09 as per my original ideas and affiliate's ideas and attempts. He seems to have been able to shave off 17C to 70C+ off the VRM temperatures with this mod with a Type 2. I will try to mount a Type 4 and see what happens. Got to get my hands on a spare 4870 though first :p:
Just a couple questions for the 4870 mod gurus:
1. What are the best heatsinks to cool the VRMs?
2. What are the best heatsinks to cool the memory modules?
3. Should every memory module / VRM be cooled individually or is it okay to put an oversized sink over multiple chips?
4. What is the best compound / tape / any other method to hold the sinks onto the memory / VRMs?
5. Does the stock cooler get in the way of the heatsinks? If so, is it a good idea to personally modify the heatsink?
6. What are the highest safe clocks obtainable on the stock cooler? How about aftermarket coolers such as the Accelero S1?
7. What are safe voltages to run on the vgddr and vgpu ?
8. Safe temps on the 4870? In GPU-Z what is the difference between GPU temperature, and GPU temperatures #1, #2, and #3? What are the respective maximums?
9. Thank you!!!
1. There are no "best" heatsinks to cool the VRM. It really depends. Coincidentally, the only good cooling so far I have seen for the 4870 VRMs is a modified HR-09 Type 2.Quote:
Originally Posted by junkmonk
2. Any heatsinks will do really, but copper is always better than aluminum. Then again, it costs more :p:
3. You can have an oversize block for all at once.
4. If you ever want to remove the sinks, the only way to go is thermal tape. Otherwise, if you want permanent installation and better conductivity you can use thermal adhesive like Arctic Silver Epoxy. There is also ways to do a combination between AS5 and ASE that allows a semi-permanent install with thermal compound but I would not recommend this as it is pretty tricky to do.
5. Stock cooler and after market sinks will not work together unless you are using only the block from the stock cooler and then mounting more sinks on top of it.
6. If you crank up your stock cooler to the max then you will sooner max out the clocks then you will be able to max out the heat. This is unless you do a voltmod in which case stock is maybe not such a good idea.
7. You cannot run any higher than 1.35V - 1.37V on the core and 1.6V - 1.7V on the memory. The card will be very unstable at higher voltages.
8. Get GPU-Z 0.2.7 and it is self explanatory. The maximum temps should never exceed 100C.
9. You're welcome :)
wow, thank you very much! didn't expect someone to answer every question! :D:D:D
As for the stock cooler not working with aftermarket sinks: does the accelero s1 work with aftermarket memory/vrm sinks?
finally, is there a cooler better than the s1 for the same price range?
My memory has degraded from 1100 to 1075, becareful leaving ur memory at 1100 24/7 as Qimonda sucks, its the cause of all video card deaths in the past year but they keep using this garbage.
I believe they don't have much choice at this moment. :(Quote:
Originally Posted by GAR
What proof do you have that running 1100MHz will kill your memory? Unless you've been volt modding your memory, whatever you achieve on stock cooling, whether it be 1100MHz or not, isn't going to kill your memory. Mine ran at 1100Mhz 24/7 for just over 2 weeks before I sold the card to buy a 4870x2. Never once had any issues with those speeds on stock cooling, and the only adjustment I made was setting the fan to 30% in the profiles, which is hardly extreme.Quote:
Originally Posted by SNiiPE_DoGG
Your GDDR5 didn't degrade and we've already discussed this in another thread. And for what it's worth, I fail to see how you can call it garbage as long as it's running the default clock speed without any issues. It wasn't selected by the ATI engineers so that you could overclock it to 1100MHz, that's overhead for the stock speed of 900MHz.Quote:
Originally Posted by GAR
did you miss the part where i said it killed the ram on one card? apparently you didQuote:
Originally Posted by Blkout
Snipe man I have some respect for you but even you should know in the pc world NOTHING is manufactured perfect. Thats like saying well I oced my northbridge and it blew up this board must suck. Its luck of the draw you know this. The only reason your memory should be killing itself is if you are getting artifacts. 1025-1050 is usually a safe bet and above the rated speed of these 40x chips so if your running at 1100 your pushing it and should expect something to break. The qimonda chips arent bad and with a lucky card you can take them past 1100 and with a unlucky card you might not reach 950 they are specced for 900.
My card starts to artifact at 1073 and when I mean artifact I can run ati furrycube for hours without it erroring but if you watch closely there are random yellow dots that pop up but its not bad enough for it to toss the error up so I am happy at 1050 or I need to volt-mod.
As for the poster above with all the questions you can mod the stock heatsink for the backside of the card. I run all day at 825/1050 and the vrm doesnt break 100c even with furmark. The HR mod is great but not needed unless you want too.
I got some stuff here ready to take the heatsink off my 4870 again.
Just wondering.
Which should I do?
Use the stock fan base and put some new thermal tape on where the vrm is as the stuff for the ram is still fine. Or should I re use the stuff on the VRM thats allready on?
Or should I use the VRM sink that comes with the accelero 2 and put some of this Akasa thermal tape on the VRM sink? or use MX2 for the VRM sink also?
Yes, it's complete speculation and FUD. Memory doesn't work that way. If it can't handle a certain speed, you will see artifacts, or worse yet, lock-ups. The memory itself won't degrade at the stock voltage. No proof by anyone has ever been given that supports this statement of memory degradation.Quote:
Originally Posted by dejanh
And you have proof of this? Overlocking the memory did this? Can you explain how this happened to one card and hasn't happened to millions of other people that are overclocking their memory? If this was a widespread problem I might agree with you, but a single case here and there leads me to believe it has more to do with the quality of the card itself, not overclocking the memory. There are bound to be a few defective cards out there somewhere, but memory degradation without voltage increases simply isn't likely. Anyone who's been around PC's long enough will know that. It's no different for SDR, DDR, DDR2, DDR3, DDR4, and now DDR5. Memory is memory whether it's on a video card or not, it all works pretty much the same. Memory degradation without voltage increases just doesn't happen unless you're talking about very long periods of of use, such as 7-10 years.Quote:
Originally Posted by SNiiPE_DoGG