The 4800 series Driver, OCing, Cooling and Tweaking Thread

[sparks.nl]

Yesterday I flashed the asus top bios and it worked fine. I rebooted a few times and so, it still worked fine.
Today I had some bas news when I woke up my pc. My display was black and stayed this way!

After putting my good old 6600 from my 3rd pc into the machine it booted fine, but not the 4870. After flashing it 3 times it now workes. I did see in windows that the speed it was "running" while it didn't work was something like 20000MHz or so.
I hope this doesn't happen again. In a few days I will try a different bios but for now I'll stay default.

 

[LuxZg]

You are using a eXtreme Power 600W with a power efficiency of only 70%. What this means is that your power supply requires 100W AC input to produce 70W DC output. At higher output demands (if your PC setup consumes near 300 watts), efficiency can decrease further as a result of heat.

I too do have low efficiency and low power PSU, though I do get ~500 3dmarks when overclocking as far as Overdrive allows me (still haven't made any modes or flashes), and I don't get any noticable rise of temperatures neither on GPU or PSU.
What is weird for me, is that I get these small hickups during 3dmark runs, like very few seconds I get a drop in the FPS and I notice it as if it would be dropped frames (not fluent animation anyway).
Could this be cos of power as well? I don't get artifactc, or crashes, just those negative spikes..

I'll try to underclock a bit, and lower my CPU clocks, just to check what happens (that should lower power consumption a bit)

EDIT: Nevermind, found out who's the guilty one.. When I have GPU-Z/CPU-Z running together with 3DMark, or probably any graphics app, I get those seemingly "dropped" frames. But looks like my score is virtually unchanged, I don't have"hickups" without them running in background, but score doesn't increase. Well, at least it's not my lower power PSU

 

[largon]

Would someone care to run a Crysis bench with an OC'ed HD4870 with the given settings (see pic)?

 

[EastCoasthandle]

I too do have low efficiency and low power PSU, though I do get ~500 3dmarks when overclocking as far as Overdrive allows me (still haven't made any modes or flashes), and I don't get any noticable rise of temperatures neither on GPU or PSU.
What is weird for me, is that I get these small hickups during 3dmark runs, like very few seconds I get a drop in the FPS and I notice it as if it would be dropped frames (not fluent animation anyway).
Could this be cos of power as well? I don't get artifactc, or crashes, just those negative spikes..

I'll try to underclock a bit, and lower my CPU clocks, just to check what happens (that should lower power consumption a bit)

EDIT: Nevermind, found out who's the guilty one.. When I have GPU-Z/CPU-Z running together with 3DMark, or probably any graphics app, I get those seemingly "dropped" frames. But looks like my score is virtually unchanged, I don't have"hickups" without them running in background, but score doesn't increase. Well, at least it's not my lower power PSU

That's odd that 2 programs cause you to lose performance like that. Tell me something, how do you power your 4870? I looked up your psu and per the homepage your PSU only has 1 vga 6 pin connector.

For anyone wondering AMD does have certified PSUs

 

[EastCoasthandle]

Would someone care to run a Crysis bench with an OC'ed HD4870 with the given settings (see pic)?

4870 @ 790/1050

Was your OC using a G92GTS @ 864/2268/2106?

 

[largon]

Very interesting...
Compared to this OC'ed HD4870 beats GTX 260 (820/1642/2632) by quite a margin. Max FPS for the Radeon is somewhat lower but average FPS is clearly higher - ofcourse, it's the average FPS that matters. I'm kinda surprised to see Radeon wins by such a clear margin.

And yes, my screen is the result my beaten G92GTS @ 864/2268/2106 managed.

...urge to upgrade grows...

 

[mmaker]

4870@830/1000 with Q6600@3300 and ATI settings with max quality

what cpu and clock you had eastcoast?

 

[nafets]

...When I have GPU-Z/CPU-Z running together with 3DMark, or probably any graphics app, I get those seemingly "dropped" frames. But looks like my score is virtually unchanged, I don't have"hickups" without them running in background, but score doesn't increase. Well, at least it's not my lower power PSU

I can also verify that this hickup occurs when running GPU-Z while the HD4870 is under 3D Load.

When running Furmark and with GPU-Z on the 'sensors' tab, there is this "ticking" about every 1 second. The Furmark donut doesn't move smoothly, but rather stutters, every 1 second.

It seems the HD4870 is a bit picky and sensitive to the polling that GPU-Z does of the card's current GPU/MEM speeds and temperatures, during high 3D loads.

Performance isn't affected though, with frame rates staying the same, but rendering of 3D objects seems a bit choppy, while monitoring the HD4870 with GPU-Z open...

 

[EastCoasthandle]

I can also verify that this hickup occurs when running GPU-Z while the HD4870 is under 3D Load.

When running Furmark and with GPU-Z on the 'sensors' tab, there is this "ticking" about every 1 second. The Furmark donut doesn't move smoothly, but rather stutters, every 1 second.

It seems the HD4870 is a bit picky and sensitive to the polling that GPU-Z does of the card's current GPU/MEM speeds and temperatures, during high 3D loads.

Performance isn't affected though, with frame rates staying the same, but rendering of 3D objects seems a bit choppy, while monitoring the HD4870 with GPU-Z open...

I don't believe your situation is the same as LuxZg as he is claiming lost performance

 

[EastCoasthandle]

4870@830/1000 with Q6600@3300 and ATI settings with max quality

what cpu and clock you had eastcoast?

E8400 at 3.60GHZ, watercooled
What do you have your fan speed set to?
When overclocking the GPU beyond CCC a voltage increase maybe needed. Create a profile and edit this:

<Feature name="CoreVoltageTarget_0">
<Property name="Want_0" value="1263" />
<Property name="Want_1" value="1263" />
<Property name="Want_2" value="1263" />

(or whatever your voltage is set to)
to this

<Feature name="CoreVoltageTarget_0">
<Property name="Want_0" value="1263" />
<Property name="Want_1" value="1263" />
<Property name="Want_2" value="1273" />

(or .10 higher then it was before)
see if there is an improvement in Crysis.

Caution:
Make sure you can properly cool both the vregs, gpu, etc (as you haven't filled out your system specs yet).

 

[mmaker]

hmm tried with 1273, 1283, 1300 and 1400 without any difference.. guess your higher CPU mhz gives you the fps increase over me then

i use the asus top bios atm, 30% idle, 39-45% under full load (furmark), the bios basically do what it can do have you gpu at max 70c

 

[EastCoasthandle]

hmm tried with 1273, 1283, 1300 and 1400 without any difference.. guess your higher CPU mhz gives you the fps increase over me then

i use the asus top bios atm, 30% idle, 39-45% under full load (furmark), the bios basically do what it can do have you gpu at max 70c

Interesting, it worked on my 2900 (although an earlier cat), I guess this has been "locked out" now with current Cats. What is your fan speed?

 

[H82LUZ73]

hmm tried with 1273, 1283, 1300 and 1400 without any difference.. guess your higher CPU mhz gives you the fps increase over me then

i use the asus top bios atm, 30% idle, 39-45% under full load (furmark), the bios basically do what it can do have you gpu at max 70c

thats because i see a new version number here is my original

ATOMBIOSBK-ATI VER011.003.000.001.029254

Here is the ASUS one maybe a sign that the manufactures are getting new bios file for the fan control and no lock on Over Drive?

ATOMBIOSBK-ATI VER011.003.000.001.029329<< this number is higher.

 

[Wshlist]

Hah, I'd wait to see actual results from several sources before I abandoned the assumption that the 'power control' they advertise with isn't broken in hardware.

 

[LuxZg]

That's odd that 2 programs cause you to lose performance like that. Tell me something, how do you power your 4870? I looked up your psu and per the homepage your PSU only has 1 vga 6 pin connector.

Just to clarify. Situation is real weird, cos I get stuttering that lowers fps as observed during benchmark run, but final scores are about the same, as if there's no stuttering. So I could say that 3DMark doesn't exactly measure those stutters and it's not visible in scores.. But when running, I get like 25fps, and when stutter appears it's for a moment on 13-15 or something like that. I've tried with and without GPUZ, and stutters appear, but score remains pretty much the same (only few points difference, which is never the same anyway). I do agree with nafets, it looks like polling of the sensors, as it's pretty much every second.. I think you have to check the option in GPUZ saying something like "poll even when not in focus"

As for my PSU, it has only one PEG, but I've used one 2x 4-pin to one 6-pin PEG adapter, and it works. Here are a few pics, perhaps it will make things clearer First one is adapter cable, second one is a closeup with me holding cables, clearly you can see 2x peripheral connectors on "Y" adapter going to one 6-pin PEG, and second PEG is directly from PSU, last pic is same as one before, but you can see it's the "right" card, soooo..it's not me having fantasies
; ;

On the other hand, I had huge troubles with getting a new PSU for my wifes brother cos these adapters weren't packed with cards, are pretty much impossible to buy somewhere, and I just had one.. so in the end we had to buy a new PSU for him (550W version of Chieftec that I have, with 2 PEGs).
There is one guy on another forum powering his system with 4870 inside, with just Fortron ATX-400PNF which should have only 4 4-pin peripheral connectors, not even one PEG, so I'm waiting for his reply how he did THAT magic When I get an answer I'll make sure that I post it here as well, cos that PSU costs like 60-65$ where I live, probably more like 45$ in US by the usual "european-to-american-prices-conversion-rates" EDIT: Updated info over here!

The ASUS gives 1200 max in RBE for me.

HEy man! I've just noticed you've said it gives you 1200 IN RBE. It says 1200 even for my stock BIOS (Club 3d as well). Could you check what Catalyst says? I think RBE is reading it wrong.. Other guy confirmed his Club3D (with ASUS bios) has 1100 clock in Overdrive (Catalyst)

 

[sinister_steve]

E8400 at 3.60GHZ, watercooled
What do you have your fan speed set to?
When overclocking the GPU beyond CCC a voltage increase maybe needed. Create a profile and edit this:

(or whatever your voltage is set to)
to this

(or .10 higher then it was before)
see if there is an improvement in Crysis.

Caution:
Make sure you can properly cool both the vregs, gpu, etc (as you haven't filled out your system specs yet).

hey there guys..sooo whats the best crysis ? the newest one or the first one or what? cuzzzz ive been stuck on ET QUAKE WARS for long time and have always wanted to get crysis and so now im gonna and want to know which one to get?..thanks,,,steve

 

[erocker]

hey there guys..sooo whats the best crysis ? the newest one or the first one or what? cuzzzz ive been stuck on ET QUAKE WARS for long time and have always wanted to get crysis and so now im gonna and want to know which one to get?..thanks,,,steve

Huh? There is only one Crysis.

 

[LuxZg]

E8400 at 3.60GHZ, watercooled

guess your higher CPU mhz gives you the fps increase over me then

Hmm... So E8400@3.6 is THAT better in Crysis? I get 45fps average in same above conditions (except I'm running in Vista, but I've run it in DX9) with my E4400@3.2GHz I mean, that's straight 50% better fps EastCoast got over my score!

Could it all be just 'cos of CPU? Or do you think that Vista lowers my score anyway, much like 3dmark..? Btw, when running same test in dx10 with 64bit I get ~35fps.
Oh, and I've used unpatched Crysis, does that matter? What have you used?

 

[Scrizz]

Hey LuxZg, something's not right with your setup.
I have the same cpu@ same clock speed and I score higher in 3dmarks06/vantage.
btw I only have a 4850. Test was done @stock

http://service.futuremark.com/compare?3dm06=7532709

just ignore what it says for cpu. sometimes it sees it correctly as e4400 and others as e6700

 

[EastCoasthandle]

Hmm... So E8400@3.6 is THAT better in Crysis? I get 45fps average in same above conditions (except I'm running in Vista, but I've run it in DX9) with my E4400@3.2GHz I mean, that's straight 50% better fps EastCoast got over my score!

Could it all be just 'cos of CPU? Or do you think that Vista lowers my score anyway, much like 3dmark..? Btw, when running same test in dx10 with 64bit I get ~35fps.
Oh, and I've used unpatched Crysis, does that matter? What have you used?

Possibilities are:
-PSU without the proper 2, 6-pin PCIe 2.0 connectors
compound by
-p35 chipset which only support PCIe 1.1 slots
compound by
-Vista
compound by
-CPU OC
-etc

Take it as only my "guess" to why there is such a difference.

 

[sinister_steve]

Huh? There is only one Crysis.

Oh ok hahahahah...I guess its just different releases ive been seeing like special release and such...hey heres my 3d mark with little OC on cards and cpu...http://service.futuremark.com/resultComparison.action

 

[sinister_steve]

Hey LuxZg, something's not right with your setup.
I have the same cpu@ same clock speed and I score higher in 3dmarks06/vantage.
btw I only have a 4850. Test was done @stock

http://service.futuremark.com/compare?3dm06=7532709

just ignore what it says for cpu. sometimes it sees it correctly as e4400 and others as e6700

what the hell I got the same 3d mark as you to the T..HEres mine ...
http://service.futuremark.com/resultComparison.action

 

[LuxZg]

Possibilities are:
-PSU without the proper 2, 6-pin PCIe 2.0 connectors
compound by
-p35 chipset which only support PCIe 1.1 slots
compound by
-Vista
compound by
-CPU OC
-etc

Take it as only my "guess" to why there is such a difference.

Ummm.. well, something is not right here. I think it's Vista mostly, I'll have to reinstall anyway one of these days so I'll install all tools & games once more in both Vista and XP just to see what's the difference on a clean system. I can perhaps borrow better PSU as well, but I doubt that this would be reason, I've just checked with Everest, and my 12V line seems stable (it should be dropping under 12V if it was overloaded, right? It hovers over 12.15 all the time, 12.15-12.19V) --> EDIT: It's 90% Vista's fault; see few post below for an update

Oh, and as for that polling->stuttering thing. I've just noticed it happens with Everest as well. Since it polls every 5 seconds (as much as I've noticed) stutters appear exactly every 5 secs. This was in 3dmark demo animation, and in crysis bench as well. Ain't that real stupid, that polling would cause stutters?! I think ATi made some kind of mistake here, or they just haven't provided right ways to poll the sensors to the software devs..
EDIT: LOL! This is annoying and hilarious at the same time Even when only in Vista, right now, while reading this forum, I get that "sandwatch" by the mouse pointer every 5 seconds, for just a quick moment That's doing nothing, just sitting in IE on this very page doing literaly NOTHING I'll reboot to XP just to see what happens.. -> EDIT: Seems like there's no stuttering with GPUZ in WinXP.

 

[H82LUZ73]

here is a theory on the High end Top And Black overclocked cards ,

see the V-Tec chip why oh why does that look like some resistors missing? I wonder if you pulled the Top`s HS off you will see some changes? Does this V-Tec not regulate the volts ?

 

[Devilduck]

Using a Sapphire HD 4870. Have fan profile set @ 65%.

What is the best air cooler to use for fan mod? The Zalman FC-ZV9 worked great on X1950 card, but doesn't this card need a stronger fan for mod? Would the Zalman FC-ZV9 even work on 4870 form factor?