Getting more outta a 4200....

[Highland3r]

Since me main rig is at uni, its Ti4200 overclocking time!
Cards been vmodded (1.98vGPU, 2.96vMEM) via VR. P4 sink added to GPU and chopped up K7 on the ram, with 120mm on front and 53 cfm on the rear.

This gets it to 344/702 prime stable, have done 10-15 runs, no prbs, this gives 16.2k 01 onions.

HOWEVER, increasing the clocks via rivatuner produces GPU lockups, wether it be increasing GPU or the Memory. (344 -> 350 core, 702-715 mem as rivatuner doesnt allow smaller hops than that, dammit) dropping the GPU to 340 stops the core lockups and gets the ram to 715 where it artifacts.

Card has 12/5v supply via molex adding 3.3 didnt improve anything so the mod was removed. Any ideas on getting the card up any further? (have considered removing the IHS after discussion with Stevil earlier today, but need to practice on a 4800SE first to make sure the card doesnt get trashed)

Voltage wise, 2.0v is the max for GPU and 3 for Mem, considering chaging the resistor to allow 3.5v on the mem but with the problems with increasing mem leading to core lockups is there any point?
Is there a need to VTT mod the card? VTT is currently reading at ~ 1.27 which seems a tad low considering it should be 1/2 the ram voltage (or does this only hold for motherboards and Standard DDR?)

Thanks :D

[-=TriX=-]

I'm sure GeForce4Ti4200 (http://www.xtremesystems.org/forums/member.php?s=&action=getinfo&userid=4427) has some tips ;)

[Geforce4ti4200]

at this point maybe youd just be better off buying or using that ti4800se to mod? I did 330/690 without mods and 340/740 with mods and artifacts to the ram. I can however tell you that you can oc by 1MHz using rivatuner, click on the "more or advanced" button and set overclocking increments to lowest safety, this will give ya 1MHz. if no go, use powerstrip

[Highland3r]

Ideal thanks. 1mhz increases will be ideal. Will drop the gpu back to 340 and increase the mem a little more, should give some headroom on the voltages as well if the ram requires some GPU voltage to go further. 1mhz bumps is what's needed :) Not sure how much of a gain removing the IHS will give, any ideas?

[Zeus]

Lock-ups are caused by the core clocked too high, clocking the mem too high just causes artefacts.

The only way to clock the card higher is improve on cooling.
Since it only does 344MHz with that CPU heatsink and 2V on the GPU don't expect too much, i don't think it's a golden sample so to speak. :(

My ti4200 did 325/730 stock, 340/750 with Vmods (2,0Vgpu-3,0Vmem) and thermaltake GF4 copper cooler, 355/750 with CPU heatsink+60mm delta and 370/758 with -10 coolant through a maze4 with 2.15Vgpu.

So, the only way to improve your card's clocks is better cooling.

Did you lap the core?
You can sand it all the way till the copper shines through, or even till the IHS is completely copper.

Did you try more voltage?
With good cooling these cards can even stand 2.35V, i wouldn't recommand that high though.

What's your set up? 16K2 doesn't sound too high with your card clocks. :confused:

edit: never mind, i saw you're on a 2,19Gig. XP

[Highland3r]

Core isnt lapped basically becuase I've read (on XS somewhere IIRC) that theres an air gap bewteen the IHS and the core itself, so seemed kinda pointless as the sink is making good contact. If anything I'd remove the IHS and lap the sides down to form a kinda shim.

Here's a link to the score (save me typing out the specs)

http://service.futuremark.com/compare?2k1=8032556

[Zeus]

Well, lapping the core shows how concave it is really.

Mine had a dip in the middle, it took me over half an hour of sanding down to get it somewhere close to being flat.

Don't know about an airgap but i do know the contact between the sink and chip can be improved a lot by just lapping it.

[Highland3r]

K will have a practice on a 4800SE tomorrow and then try on the 4200 :S

[Zeus]

Originally posted by Highland3r
K will have a practice on a 4800SE tomorrow and then try on the 4200 :S

Isn't this 4800SE much better an overclocker?
Or is it broken?

[Highland3r]

Its buggered lol! not even detected on boot now... :(

[gabbax]

Originally posted by Zeus
Well, lapping the core shows how concave it is really.

Mine had a dip in the middle, it took me over half an hour of sanding down to get it somewhere close to being flat.

Don't know about an airgap but i do know the contact between the sink and chip can be improved a lot by just lapping it. Mine to, i say maybe a 0.5mm of gap in the middle:mad: ..and that take me about a 1 hour to do just like u did Mr:Z...but ofcourse it did better ive gained a few mhz more on the core:)

[Zeus]

Originally posted by gabbax
Mine to, i say maybe a 0.5mm of gap in the middle:mad: ..and that take me about a 1 hour to do just like u did Mr:Z...but ofcourse it did better ive gained a few mhz more on the core:)

G, did u sand it down till the IHS was completely copper?

I'm planning to do so, as of now i have the edges being copper but the centre is still silver.

As soon as temps drop, i will run some more benches with this card, hoping to see some 375ish on the core. :)

[r3b0rN]

my fx5900xt has 1 of those heat spreader thingy's ontop of the core... if i sand it back/remove it... will the stock heatsink still make good contact? or should i screw it in? (instead of using push pins)

[Highland3r]

Originally posted by r3b0rN
my fx5900xt has 1 of those heat spreader thingy's ontop of the core... if i sand it back/remove it... will the stock heatsink still make good contact? or should i screw it in? (instead of using push pins)

Depends on the thickness and it the push pins are sprung, but it'd probably be a good idea to bolt the cooler on to be safe.

[r3b0rN]

what i dont understand is why they put these things on them in the first place...

[Highland3r]

Suppose its to help prevent crushage of the cores, though the ATI cards dont seem to have any problems.