Radeon X800 Non-Pro/X800 XL/X850 Voltmods

[nightbase]

Hi guys; I bought a new sapphire x850xt pci-e ,try some o/c, but the cards lock only gpu at 540 I was very disappointed ,How can I reach gpu 570-580 or more,, and my psu is "ASUS 550 WATT pci-e supported" and so I'm not a expert for vmod, I only do pencil mod, but when tried this mod nothing happened still lockup gpu=540

 

[Turrican]

Hi everbody

I've modded my Connect3D X850 XT PCI-E, with the detailed article in this forum

GPU:
before mod 550@ 1.41V
after mod 610@ 1.70V@ watercooling (not the max, i'm still testing


but with the ram i've probs.

before mod 600@ 2.10V
after mod 600@ 2.41V @ big cooper freezers+3x 80mm fans blowing on them.(the ram only get about ~35°C hot.)

I've done both ram mods (VDD+VDDQ (Hardmod with VRs)) but i can't oc the ram higher than before the mod
when i set the ram to 605 and i bench or play the screen suddenly goes black and i've to reset

Voltages:

before/after
GPU 1.40/170
VDD 2.10/2.41
VDDQ 2.12/2.45

what's here wrong??


thanks in advance
Turrican

 

[funkflix]

Just normal to be honest, most of the times, a higher Vdd give u no gain in overclocking, sometimes it makes it even worse, then before.

2,4V is to much, try 2,2 - 2,25V.

 

[Turrican]

Just normal to be honest, most of the times, a higher Vdd give u no gain in overclocking, sometimes it makes it even worse, then before.

2,4V is to much, try 2,2 - 2,25V.

ok, i gonna try it with less voltage thx

VDDQ should be a bit higher than VDD, right? e.g. VDDQ 2.25V and VDD 2.20V

c u

 

[funkflix]

VDDQ should be a bit higher than VDD, right? e.g. VDDQ 2.25V and VDD 2.20V

c u

Yes! Good luck

 

[ViperJohn]

ok, i gonna try it with less voltage thx

VDDQ should be a bit higher than VDD, right? e.g. VDDQ 2.25V and VDD 2.20V

c u

Not on x850's. On R420 cards you ran Vddq hotter than Vdd by .075 (which is decidedly against Samsung specs, due to a PWA layout quirk. Vdd on x850 of 2.30 - 2.350 has been pretty much the best range with the Vddq at stock. That is for Samsung chips. If your card got the crap Infineons you will generally go no where with higher memory voltages.

With Samsungs then probably either your ramsinks are not getting the job done or the cores memory controllers are shutting your memory OC off early.

Viper

 

[Turrican]

Not on x850's. On R420 cards you ran Vddq hotter than Vdd by .075 (which is decidedly against Samsung specs, due to a PWA layout quirk. Vdd on x850 of 2.30 - 2.350 has been pretty much the best range with the Vddq at stock. That is for Samsung chips. If your card got the crap Infineons you will generally go no where with higher memory voltages.

With Samsungs then probably either your ramsinks are not getting the job done or the cores memory controllers are shutting your memory OC off early.

Viper

my x850 xt has got the 1.6ns samsung chips.
the ram cooling is for sure no prob ( big cooper freezers+3x 80mm fans blowing on them.(the ram only get about ~35°C warm.))

/edit:
i've tested 2.26V (VDD) & 2.28V (VDDQ) @ 610. 3dmark01 passed once. 3dmark 03 got crazy artifacts in the 2nd test+ BSOD.

now, i will try vipers advice (vddq @ stock & vdd ~2.33V

 

[ViperJohn]

my x850 xt has got the 1.6ns samsung chips.
the ram cooling is for sure no prob ( big cooper freezers+3x 80mm fans blowing on them.(the ram only get about ~35°C warm.))

/edit:
i've tested 2.26V (VDD) & 2.28V (VDDQ) @ 610. 3dmark01 passed once. 3dmark 03 got crazy artifacts in the 2nd test+ BSOD.

now, i will try vipers advice (vddq @ stock & vdd ~2.33V

Have never heard of Big Cooper Freezers.

I just went over the tech notes from all the x850's I have done (it's been a while lol). Per card the Vdd they wanted (every card is different and there is no "one size fits all" settings) was from 2.225 to 2.350 with the Vddq running from .075 to .125 lower than Vdd on PCIe cards. In all cases running the Vddq to hot KO'ed the memory OC

The memory OC ranged from 633 to 663 (with slightly tighter memory timings). One card only hit 609 memory, which is a real dud for an x850 PCIe card, and was due to the cores bad memory controller. It sounds like your card core may have a poor memory controller too.

For x850 AGP cards, which use the R420 AGP card PWA, run the Vddq .075 higher than VDD.

When you test the memory OC temporarily back the core OC off a little just to be sure the core OC is not tripping the card up and fooling you into thinking it is the memory.

Viper

 

[Turrican]

ok, i gonna try. i have got the PCI-E Version of the Card

oh btw. my bad, i meant Revoltec BGA RAM Freezer Copper Edition
didn't know the exactly name first

here's a pic

 

[ViperJohn]

ok, i gonna try. i have got the PCI-E Version of the Card

oh btw. my bad, i meant Revoltec BGA RAM Freezer Copper Edition
didn't know the exactly name first

here's a pic

Awh okay. They are a decent sink if forced air cooled, which is what micro fin spacing is designed for, and you do not use cheap frag tape to mount em.

Be sure to keep em clean as dust bunnys build rapidly in the narrow fin spacing fan cooled. As with any all CU sink tanish will build over time hurting their efficiency.

Viper

 

[infrared]

Hey viperjohn, what's the best thermal past for the GPU and tape for the ramsinks? I'm getting the AC ATI silencer 5r2, which i've just ordered, but i wander if there's any better paste or tape i can use to aid efficiency.

Thanks for any input.

 

[Turrican]

hm, no matter what voltage i adjust i can't get past 600mhz, even when the core ist untouched and running @ standard(520). looks like i'm hitting a "wall" @ 600

the strange thing is, that i can adjust the mem timings of the card (@ 600) without getting artifacts or lockups, so there must some more room to OC.


i can set these timings without getting lockups,..

standard--->OC

TRCDRD: 7--->6
TRCDWR: 4--->3
TRP: 5 --->untouched
TRAS: 14 --->13
TRRD: 5--->4
TWR: 7--->6
TR2W: CL+3clock--->CL+2clock
TW2R: 3 --->untouched
TR2R: USE TWR RULE--->untouched
TR2R: 2--->untouched
WR_LATENCY: 1.5--->untouched
CAS_LATENCY: 7--->untouched
CMD_LATENCY: 0--->untouched
STR_LATENCY: WR LATENCY--->untouched
REFRESH_RATE: 0x41-->0x75
TRFC: 27--->23

c u

 

[funkflix]

The wall should be the Memory Controller, as Viper sad before. :-/

 

[ViperJohn]

Hey viperjohn, what's the best thermal past for the GPU and tape for the ramsinks? I'm getting the AC ATI silencer 5r2, which i've just ordered, but i wander if there's any better paste or tape i can use to aid efficiency.

Thanks for any input.

I use AS5 on the cores. For the memory I use my NSTIM. Do not know who actually makes it but it is damn expensive stuff lol!

Viper

 

[ViperJohn]

the strange thing is, that i can adjust the mem timings of the card

You just put the final nail into the core memory controller's coffin. That is what is killing your memory OC and is why the chips are not responding to voltage increases and timing changes are not putting a dent in the memory OC.

Your timings look good but increase the TRCD-Read to 7 and increase TW2R to 4 for starters.

You can also D/L the memory chips datasheet from Samsung and try the tighter timings use on the slower speed rated chips too.

Viper

 

[infrared]

I use AS5 on the cores. For the memory I use my NSTIM. Do not know who actually makes it but it is damn expensive stuff lol!

Viper

Thanks

Where do you get it from? lol, and what is it?

Sorry, lots of questions!

 

[ViperJohn]

Thanks

Where do you get it from? lol, and what is it?

Sorry, lots of questions!

Send me an email.

Viper

 

[Turrican]

You just put the final nail into the core memory controller's coffin. That is what is killing your memory OC and is why the chips are not responding to voltage increases and timing changes are not putting a dent in the memory OC.

Your timings look good but increase the TRCD-Read to 7 and increase TW2R to 4 for starters.

You can also D/L the memory chips datasheet from Samsung and try the tighter timings use on the slower speed rated chips too.

Viper

thanks viper, i will try it and downlaod the data sheets

c u

 

[lelek]

Hi!

I'm a noob to this page, to vmodding but I think I have an useful method and I want to share it with the Others(Lost ownz ). So as I wrote to Urlyin in my e-mail(btw he's the one who encouraged me to write this down here, so thanks for that ).

I decided to do the pencil vmod.

The main problem with the original pencil vmod technique that graphite only can be applied in a very thin layer, therefore it's conductivity is quite poor. So I thought somehow graphite powder should be applied on the place of the missing resistor.
So I made some powder using a sharp cutter on my mechanical pencils 0.5mm lead (I think it's HB graphite, not 2B, or I tried this with 4B pencils but HB was better), then I put only 1 or 2 drops of water in a small pot (actually it was the lid of a little bottle with a very short rim on it, so it was perfect in size to contain so little amount of water). After that the powder was dispersed into the water and I mixed the whole thing with the lead of my mechanical pencil.
The HB graphite's powder is so refined that it only floats on the water, but it's possible to get some water and graphite powder aswell on the point of the pencil and after this, it can be literally painted with the point of the pencil between the two solder pads. After the water evaporates there's a perfect graphite bridge (and it sticks well so the air movement can't blow it off, but it's easy to remove it - for example with a little rubber) between the pads and it's thick enough to lower the resistance even to about 380KOhm(this is what I measured) from the original 430KOhm.
Finetuning the resistance is easy too. Again with the point of the mechanical pencil the width of the bridge can be reduced, so the resistance will increase.
I myself tuned the resistance to about 400KOhm, the core voltage is about 1.51V now. The original was 1.41V at 430KOhm.
I have a PrimeCooler PC-VGAHG1 CU on the card - VF700 is out of stock everywhere here (Hungary, Budapest) so I bought this.
Btw this 0.1V means, the core is stable on 575 mhz instead of about 545mhz.
With the stock Cooler on 500mhz core the temperature was about 80C°.
Under load the core temperature climbs to 64C°(~560mhz,62C° with fullspeed fan) now. It was about 60-61C°(on lower core frequency~500mhz) before the Vmod so I think it will be OK on the long run.

Now the next step to get some air movement on the upper side of the card, because ram heatsinks are running a little hot on that side.

I hope this graphite painting thing will be useful for others who have problems because of the missing R1597.

 

[infrared]

oh nice job!! Great thoughts!

The only prob is i can't get my ati silencer 5 off!

Thanks for sharing, i'll give it a go asap.

 

[12k]

I finally got the mod fixed, one issue though, I use heatkiller watercooling, and my "water"cooling is chilled at -28 to -20 deg C so theres a bit insulation, that gives a problem... The thing is.. The measuring point for the VGPU is badly located :'( - So I havnt put a reading spot there.. - Can someone help me out? - I use a 9.83K ohm VR... 15 rounds to the right = 0 ohm... - I would like to hit around 1.775-1.85v on the vgpu, how many "rounds" would ya advise??

I need a table for ohm = vgpu

Etc. 5K ohm = x-Vgpu, 7.6K ohm = x-Vgpu

- The VDD I got a measuring point on

 

[Urlyin]

Hi!

I'm a noob to this page, to vmodding but I think I have an useful method and I want to share it with the Others(Lost ownz ). So as I wrote to Urlyin in my e-mail(btw he's the one who encouraged me to write this down here, so thanks for that ).

I decided to do the pencil vmod.

The main problem with the original pencil vmod technique that graphite only can be applied in a very thin layer, therefore it's conductivity is quite poor. So I thought somehow graphite powder should be applied on the place of the missing resistor.
So I made some powder using a sharp cutter on my mechanical pencils 0.5mm lead (I think it's HB graphite, not 2B, or I tried this with 4B pencils but HB was better), then I put only 1 or 2 drops of water in a small pot (actually it was the lid of a little bottle with a very short rim on it, so it was perfect in size to contain so little amount of water). After that the powder was dispersed into the water and I mixed the whole thing with the lead of my mechanical pencil.
The HB graphite's powder is so refined that it only floats on the water, but it's possible to get some water and graphite powder aswell on the point of the pencil and after this, it can be literally painted with the point of the pencil between the two solder pads. After the water evaporates there's a perfect graphite bridge (and it sticks well so the air movement can't blow it off, but it's easy to remove it - for example with a little rubber) between the pads and it's thick enough to lower the resistance even to about 380KOhm(this is what I measured) from the original 430KOhm.
Finetuning the resistance is easy too. Again with the point of the mechanical pencil the width of the bridge can be reduced, so the resistance will increase.
I myself tuned the resistance to about 400KOhm, the core voltage is about 1.51V now. The original was 1.41V at 430KOhm.
I have a PrimeCooler PC-VGAHG1 CU on the card - VF700 is out of stock everywhere here (Hungary, Budapest) so I bought this.
Btw this 0.1V means, the core is stable on 575 mhz instead of about 545mhz.
With the stock Cooler on 500mhz core the temperature was about 80C°.
Under load the core temperature climbs to 64C°(~560mhz,62C° with fullspeed fan) now. It was about 60-61C°(on lower core frequency~500mhz) before the Vmod so I think it will be OK on the long run.

Now the next step to get some air movement on the upper side of the card, because ram heatsinks are running a little hot on that side.

I hope this graphite painting thing will be useful for others who have problems because of the missing R1597.

Thanks Lelek

 

[XibaD]

Hi!

First of all I must say that your guide to vmod X800 is great! So much work is written there

I tried to vmod de GPU of my ASUS X800XL with no results for me. Maybe I didn't paint well the solder pads, maybe yes but it does not work. I have no measuring stuff but ASUS has a sofware that measures VGPU and VDDq so the only thing I have to do is pencil mod carefully and watch if there is any changes in vgpu in windows. I "penciled" amongst the 2 solder pads, but it was of no use, the vgpu didn't change nor the oc capability of my X800XL AGP.

What I did wrong? Does the pencil line have to be much thick? How many times should I draw over?

Sorry for the questions, I'm noob in vmodding vga, and I don't want to kill this one

 

[Urlyin]

Hi!

First of all I must say that your guide to vmod X800 is great! So much work is written there

I tried to vmod de GPU of my ASUS X800XL with no results for me. Maybe I didn't paint well the solder pads, maybe yes but it does not work. I have no measuring stuff but ASUS has a sofware that measures VGPU and VDDq so the only thing I have to do is pencil mod carefully and watch if there is any changes in vgpu in windows. I "penciled" amongst the 2 solder pads, but it was of no use, the vgpu didn't change nor the oc capability of my X800XL AGP.

What I did wrong? Does the pencil line have to be much thick? How many times should I draw over?

Sorry for the questions, I'm noob in vmodding vga, and I don't want to kill this one

Do the Vid vmod .. much easier

 

[XibaD]

Ok, I read it, seems easier. But I don't know exactly where to get a conductive pen. Could pencil work with this mod too?

Thanks a lot.