4890 highest core clock

[Wile E]

Those are the vregs on my GPU not mobo.

He knows. They absolutely need to have heatsinks on them, and airflow going over the sinks. Otherwise, you will kill your card.

 

[Spaceman Spiff]

DarkEgo, what kind of space you have between the Vregs and the Musashi? Is it enough for 14mm?
http://www.petrastechshop.com/enbmpufocora.html

I was thinking of getting the Low profile 9mm but my mushashi wont be here till next week.
http://www.petrastechshop.com/enbmpufocora1.html

Any chance you could measure the clearance?

 

[DarkEgo]

I ahve a heatsink on them, it is just the stability test for GPU tool makes the temps go way up, they dont go above 65 in l4d.

 

[_33]

Highest core clock, fine... But, do we need to certify it works with Furmark?

 

[erocker]

I wouldn't use furmark. Unless you like destroying your 4890.

 

[DarkEgo]

Then what is the best stability test? One that won't destroy my card.

 

[mudkip]

No furmark = no stable overclock = shit overclock = useless overclock = noob overclock

 

[_33]

I wouldn't use furmark. Unless you like destroying your 4890.

The worst is, I absolutely believe you in reference to destroying a GPU. My older card the 4850 was destroyed while playing a game! OK it was overclocked, but not by much... But now I need to send it back to Sapphire....

 

[mudkip]

Then what is the best stability test? One that won't destroy my card.

Stability test , Extreme burn more , MSAA : 16

 

[erocker]

No furmark = no stable overclock = shit overclock = useless overclock = noob overclock

Bullshit all the way. I've seen plenty of cards fail furmark or get destroyed by furmark, and not by other stress test means or actual applications. It's like trying to stress test a car by flooring it in neutral for minutes on end. You're going to burn it up. Before furmark was even an idea you know what worked for testing stable overclocks? Games. Oh, and ATi tool.

 

[_33]

OK so let's say that if I can play Fallout 3 for half an hour then I'm fine. LOL J/K

 

[BooStFeD]

Wizzard's new GPUTool has worked great for my 4890. I do not like using FurMark as i feel it is not needed to stress a gpu that much.

I downloaded and installed GPUTool from TPU. Then used the Find Max functions for core + mem. The stability tester was able to run for 10 minutes artifact free.

With GPUTool and default volts i am stable @ 940/1106
GPUTool + 1.38 volts i am stable @ 1005/1106

 

[DarkEgo]

Wizzard's new GPUTool has worked great for my 4890. I do not like using FurMark as i feel it is not needed to stress a gpu that much.

I downloaded and installed GPUTool from TPU. Then used the Find Max functions for core + mem. The stability tester was able to run for 10 minutes artifact free.

With GPUTool and default volts i am stable @ 940/1106
GPUTool + 1.38 volts i am stable @ 1005/1106

What are your VDDC Slave temps at load? Mine go realy high with furmark and gputool.

 

[Wile E]

Bullshit all the way. I've seen plenty of cards fail furmark or get destroyed by furmark, and not by other stress test means or actual applications. It's like trying to stress test a car by flooring it in neutral for minutes on end. You're going to burn it up. Before furmark was even an idea you know what worked for testing stable overclocks? Games. Oh, and ATi tool.

On my system, for some reason, COD4 seems to be the ultimate test of stability for me. I can play Crysis for hours on end, or GTA 4, or run 3Dmark 400x in a row, and not have a single issue with an OC, but COD4 always crashes within 2 minutes if anything in my setup is unstable. lol.

 

[phanbuey]

No furmark = no stable overclock = shit overclock = useless overclock = noob overclock

remember that your VRMS are usually rated for 1000hrs (42 days) at 100C ~ and that number shortens dramatically past that temp... so at 110 or 120 C which is what furmark does to those voltage generators, you're really burning that card up.

Plus - OC's are not set in stone. If you're playing a game and it crashes or you see artifacts, then back off a bit - that's all. L4d, COD 4, Oblivion and a whole bunch of other games are great testers of "stable" OC's.

 

[BooStFeD]

What are your VDDC Slave temps at load? Mine go realy high with furmark and gputool.

Just ran both GPUTool and FurMark @ 940-1106 default volts + 50% fan speed.

Rather large difference in temps. between these two apps. I noticed GPUTool only loading the GPU about 90% while FurMark loads GPU 99%.

Maybe my GPUTool is not working correctly

http://i137.photobucket.com/albums/q210/BooStFeD248/GPUToolLoadTEST940-1106.jpg
http://i137.photobucket.com/albums/q210/BooStFeD248/FurMarkLoadTEST940-1106.jpg

 

[AsRock]

On my system, for some reason, COD4 seems to be the ultimate test of stability for me. I can play Crysis for hours on end, or GTA 4, or run 3Dmark 400x in a row, and not have a single issue with an OC, but COD4 always crashes within 2 minutes if anything in my setup is unstable. lol.

Would it not be a game that makes your system the most power ?.. I cannot say with COD 4 as i do not own it.

But the 2 games i've notice with the higher power usage has been NWN2 and Crysis were there has been around a 30-50w difference. With COH and Arma being on the lower end on total power usage.

I'l add a custom cooler on mine if i find a good way to screw a heatsink on the regs lol.

 

[largon]

HT4U.net did some power consumption measurements in different games using HD4890, and some other cards too.

Also, there's a test out there that goes ~50W higher than Furmark on HD4890... But it's not sensible to run it even on a stock HD4870 as the card will crash systematically, likely due to OCP. Only way to get it to work is underclocking the card.

 

[amit_talkin]

man, how can u guys go so high at stock vcore? I even cant cross 900MHz on my sapphire HD4890. it gives me artifact @ 900/1050 setttings .

 

[Josh81]

No furmark = no stable overclock = shit overclock = useless overclock = noob overclock

sure


or you could play crysis or another GPU heavy game for a few hours

 

[muon]

A word of warning. You must use something better than ramsinks for the vrms. I used to have an arctic cooler s1 installed on my hd 4890 with a 120mm and 80mm fan. The 80mm fan was positioned firectly over the VRMs.

I reinstalled the stock cooler. VRMs with furmark even with the supplied arctic cooler sinks would send the VRMs over 120C (I always stop stressing the card when it gets to 120C to prevent it getting damaged, it would hit 130C+ no problem). The gpu core itself would stay ~50C, but I can't stress it properly due to VRM temps (idle core temps of 35C). The stock cooler lets the core hit upto 80C even at upto 50% fan speed. But does keep the VRMs under 100C.

Its a shame because the stock cooler becomes too loud for my system when it hits 35%.

My former setup:

Components in question

Back to the stock cooler with mx2 applied

 

[Josh81]

A word of warning. You must use something better than ramsinks for the vrms. I used to have an arctic cooler s1 installed on my hd 4890 with a 120mm and 80mm fan. The 80mm fan was positioned firectly over the VRMs.

I reinstalled the stock cooler. VRMs with furmark even with the supplied arctic cooler sinks would send the VRMs over 120C (I always stop stressing the card when it gets to 120C to prevent it getting damaged, it would hit 130C+ no problem). The gpu core itself would stay ~50C, but I can't stress it properly due to VRM temps (idle core temps of 35C). The stock cooler lets the core hit upto 80C even at upto 50% fan speed. But does keep the VRMs under 100C.

Its a shame because the stock cooler becomes too loud for my system when it hits 35%.

My former setup:

http://dl.getdropbox.com/u/20407/22052009337.jpg
http://dl.getdropbox.com/u/20407/22052009338.jpg
http://dl.getdropbox.com/u/20407/22052009340.jpg

Components in question

http://dl.getdropbox.com/u/20407/22052009342.jpg

Back to the stock cooler with mx2 applied

http://dl.getdropbox.com/u/20407/23052009347.jpg

some guy used accelero and the plate from the stock cooler, you could try that

 

[muon]

some guy used accelero and the plate from the stock cooler, you could try that

Theres no plate. The stock cooler is one piece with the copper heatpipes and copper face contained within. The main reason its so much better than the stock cooler is that the core heats up the copper component, heatpipes and fins, while the RAM and VRMs are effectively touching a massive piece of aluminium(maybe steel/iron?).

Look at TPUs review

http://www.techpowerup.com/reviews/Sapphire/HD_4890/4.html

 

[Josh81]

Theres no plate. The stock cooler is one piece with the copper heatpipes and copper face contained within. The main reason its so much better than the stock cooler is that the core heats up the copper component, heatpipes and fins, while the RAM and VRMs are effectively touching a massive piece of aluminium(maybe steel/iron?).

you can unscrew the red plate that is on the stock cooler

 

[muon]

Oh thats brilliant. Thanks, might work a lot better. Hopefully won't get in the way of the S1.