Jump to content
HWBOT Community Forums

Radeon R9100 recap


Recommended Posts

A long time ago I get this old card. It was, even back in his day, low end card. But it was capable of running fanless and as backup AGP card it was usefull. And exactly that is the reason, why I ended up recapping the card completely. That is, because of the problems I having with the MSI PM8M3-V mainboard stability in gaming, I need a rock-stable and 100% reliable AGP graphic card.


However this card was unlucky from the day one. First it was unlucky, because it got known bad Licon caps. These 470uF 10V suxxkas show good capacity, yet terrible ESR. 60mOhms is way over what is usable:


Radeon_R9100_Licon_cap.jpg R9100_Licon_cap_fail.jpg R9100_100uf_cap_fail.jpg


...not to mention the little 100uF 16V SMD caps having lost chunk of their capacity and ESR starting to pump up to whooping 890mOhms...


Therefore it is no surprise, that the card crash during SoF2 gaming. Almost always the game crash when level loading...


Also the card was unlucky, because when I decided to upgrade it's cooling with half of the Zalman ZM80A-HP vga heatpipe cooler. But there are no holes near the chip. So I decided to glue it to the chip. And during this the weights slided a bit and the result is, well, not optimal and ughly...




So I decided to give the Radeon R9100 much better caps (as you can see already the caps are replaced), completely polymer (and ceramic) ones and add few and bump the capacity a little too:




All these caps are Nichicon polymers, that have quite much better specs that the original crap caps. Even the SMD caps are Nichicon polymers now and their capacity was, with the 470uF exception, bumped from 100 to 330uF (220uF in the 6.3V case, as the bottom cap run at 5V, not 3.3V as the rest):




That is the bottom cap - look at the shocking number of places, where caps might be! If there is cap everywhere, then the card will be probably designed to take far much powerfull graphic chip that the R9100 is:




Near the ram chips I added 4 22uF 6.3V SMD ceramics to four empty places near the ram chips and I also added another two 330uF 4V Nichicon CK caps for better ram stability (having overclocking also in mind):




Two of these ceramics are added to the bottom of the card:




When the heatsink are screwed in, then the whole fault is revealed and... it did not looks nice at all. But at least it fit into the AGP slot w/o any issues...




So, to recap this unknown Radeon R9100 card, you need:


Radeon 9100


4x 470uF 10V d10 Licon - 1x 1200uF 4V Nichicon F5 (d10x13) RR50G122MDN1 - 3,3V (top one, near coil)

- 3x 1000uF 2.5V Nichicon LF (d10x13) PLF0E102MDO1 - under 2.5V

4x 100uF 16V d6.3 SMD - 1x 220uF 6.3V Nichicon PCK0J221MCO1GS - 5V (bottom one)

- 3x 330uF 4V Nichicon CK PCK0G331MCO1GS - (3,3V)

1x 470uF 6V d8 SMD (3,3V) - 470uF 6.3V Nichicon HA (d8) SMD RHA0J471MCN1GS

2x 22uF 16V d4 SMD (2,5V) - 22uF 6.3V Murata X5R (1206) JMK316AB7226MLHT

- added 2x 330uF 4V Nichicon CK PCK0G331MCO1GS (near rams)

- added 4x 22uF 6.3V Murata X5R (1206) JMK316AB7226MLHT (near rams)



The result?

Well, when I put it into my MSI PM8M3-V mobo, it at first show ram errors (heartbreaking, after so much work and so many polymers and added caps...) on the screen, so I have to pull it off and check.

Fortunately, I managed to find the bit of tin, that ended up on one of the bottom ram chips and after cleaning - removed and - hoooray, all works like a charm!


Immediatelly the card go into AGP 4x mode (!) and despite being TAD slower that PNY 6800GT, the system feels notably faster. That is, because, PCI mode DO SUXX BADLY. Therefore the card is now used now to test, what is wrong with the mainboard :)

Link to comment
Share on other sites

Guys, guys... easy, easy on me, pls.


As I stated above, it was mainly attempt to get my AGP working - finally. It failed, but with the 100% good and recapped with very high quality caps R9100 I'm now confident, that the problem must be on the mainboard, with the voltage regulating mosfets. That make sense, because it works, till it heat up and crash. 4h gaming from cold state, about 40min when heated up to the crash.

Therefore the main and only one purpose of this recap was the testing of the MSI mainboard and these cursed Nikos mosfets!


I recapped the mainboard completely long time ago ( http://forum.hwbot.org/showthread.php?t=86667 ) and recently around the AGP, I even exchanged the good Nichicon 1000uF 6.3V electrolyte caps for polymers (except for the two that are for 5V) - 2200uF 2.5V and 1200uF 4V. Still absolutely NO help. Absolutely NO improve. Therefore it cannot be in the graphic card (it fact, it last longer during gaming that with the PNY 6800GT) or in the caps - so the next one in the line are the mosfets.


That is why no overclock (yet).



Antinomy - indeed that suxx. Overclock or die? I know, I know. But there is NO point of overclocking, when your AGP 4x card is running, after the first crash, in PCI mode!!!




Then you have NO PERFORMANCE to speak off. It absolutely and completely SUXX. (from my measuring, it give about half the performance it should do, so no 7850 3DMark01 marks, but about 14k it should yield: http://hwbot.org/submission/2455634_trodas_3dmark2001_se_radeon_9100_7855_marks )



ObscureParadox -

Would have been nice to know some overclocking results with a cooler on it thats practically the same size as it now xD Do you plan to do any overclocking with it and possibly volt mods in the future?


Yes, but only IF I manage the AGP working on the MSI PM8M3-V board :( There is nothing worser that PCI mode, believe me. There is absolutely no point of any overclocking, when the hardware is unstable as it is. It cannot reliably provide more watts for the card, because it is stuggling as it is already.

Volt mods will be probably unnecassary, as the WR for R9100 is only 13700 3DMark 2001 marks ( http://hwbot.org/submission/833462_qwerty84_3dmark2001_se_radeon_9100_13700_marks ), witch should be easy to reach into AGP mode of operation (if the scaling of speed go as with the 6800GT) with almost no overclocking... Of course the rams are crappy (witch is why Sapphire run them at 200MHz, witch is deviation from the standard 250/250MHz clocks for GPU/RAMs for R9100:


So the card is somewhat disadvantaged from the start.


And, even after will work, then it will need a serious CPU and cooling to be able to produce some good scores. Most benches are very CPU driven and P4 is anything but fast.

Even if I get 3.8GHz P4, it will not stand a chance against some overclocked C2D CPU's in good boards, that still had the AGP slot on them. If I manage to source such board (even if I have to recap it or slightly repair it), then it will be quite interesting how far the GPU can be pushed.


Normal are 250/250MHz, just like the Radeon 8500:

ATI Radeon 8500 | techPowerUp GPU Database

...it should be just 250/250MHz...

(XT version was 300/300MHz, tough: ATI Radeon 8500 XT | techPowerUp GPU Database )


And surprisingly, look, the same PCB are used on Radeon 8500 cards - just with all the caps!!!




...damn them :) (but it give me ideas, where to add another caps... but that is probably not worth it, as the A-Data rams are slow and the card is not intended for overclocking, but for stability testing.


Of course I would love to overclock it, but the dying mosfets on my mobo will prevent any serious overclocking :(


It would be great, if I can just stick there temperature probes on these mosfets and read the temps, as I play SoF2 game, till the point of crash. That would be proof, that overheating one of these mosfets cause this crashing... possibly saving me the hassle to replace all there? Dunno. I'm, basicaly, out of ideas about what to do more, if this (replacing the AGP mosfets) fail.

Link to comment
Share on other sites

  • 2 weeks later...

So, since my machine is not crashing anymore during gaming, then I decided to try look out, what overclocking is possible with this recapped Sapphire Radeon R9100 card. And it is not that bad:




It can do 293 for the GPU, but I backed it down to around 290, same for the rams, around 248MHz it can do. That make it clear, why these rams had to run at 200MHz, because they cannot take the normal 250MHz reference design clock, not with any caps.

Hence I cannot match the XT version (300/300 vs 290/248), but at lest some overclocking is possible. It would be interesting to know, if adding more caps on the empty spaces (there is space for DAMN much caps!) could help to reach faster speeds. I was kinda hoped for the 300MHz mark, but I got burned shy 7MHz of it...

Link to comment
Share on other sites

  • Crew

trodas, I like what you're doing because you're doing something more than "crank the voltage, pour LN2 and click the soft-clock utility" what most do. I'm not pushing on you and if I really thought this suxx, I wouldn't spend my time to comment.


I comment because I see efforts that are very likely to be worthless. When you spend a decent amount of time on hardware that was a bad choice from the beginning and the results are moderate at best. It's like trying to hit top results achieved on modded DFI NF2 Ultra B with an highly modded ECS/MSI board and not being able to hit even a non-modded one.


Just want to keep you from early disappointment.

BTW, the Radeon 9100 is in fact a re-branded 8500 so no wonder the share the same PCB and all the stuff ;)

Link to comment
Share on other sites

Antinomy - well, this project was more about getting a 100% reliable GFX card to determine, how to fix the MSI PM8M3-V mobo that is failing in the AGP / gaming only. Therefore I picked up the card that has the less draw that I get. There are a FX 5600XT too (well overclockable from 235/200 to 375/267), but it will require considerably mode power, so my main machine will become unstable... :(


Therefore yes, the efforts to get back into working state such card is worthless, but since I did not have any 100 reliable AGP card except this one now, then I need these results :D


(and with overclocking - even stable, tested for hour on ATItool - the board crash in gaming, so I did not fixed it yet... :( time to look out for another mosfet(s) to exchange to get the stability... )


It's like trying to hit top results achieved on modded DFI NF2 Ultra B with an highly modded ECS/MSI board and not being able to hit even a non-modded one.


You are right. But I did not attempting to get any high scores with the card. It is too slow for that to begin with :)

However I have saved previous max. stable O/C results and they are:

250/200 - to 265/248 ;)

Now it is:

250/200 - to 290/248 ;)


That means that the rams O/C are not affected at all by the quality of voltage filtering (eg. caps), yet the GPU overclock is affected and when new, the card was actually slower that todays, after the recap :)


Might be fun to add more caps and see, if the GPU clock can reach 300MHz or not :P But sure as hell I need to fix the MSI mobo problem first, because lack of good power on the AGP slot is crucial :(



Turrican - sure he have the point - the card is too slow to begin with, but as AGP testing card it will serve it's purpose well ;) Not much else is expected anyway.

Except 300MHz GPU clock, lol.

Link to comment
Share on other sites

  • Crew

Well, you've misunderstood me a bit.

But I did not attempting to get any high scores with the card. It is too slow for that to begin with
That is dead wrong. It's HWBot - we don't care how fast it is by itself. We card about how it compares to other submissions with this card. I've been recently benchmarking Socket 7 and believe me - that was slow! Nevertheless, I got some gold cups and points too. And a pound of pleasure.

So the wrong point was not about using Radeon 9100 but about recapping a card that has weak memory chips. You can push the GPU in most cases. But if the memory is weak there is not too much you can do about it. Your test prove it. But anyway, your 9100 journey can help you get points.


My main rants were about the MSI board :D



The only bold advice is as follows - "Don't stop and keep pushing".

Edited by Antinomy
Link to comment
Share on other sites

Oh, that. Then yes, the Radeon R9100 is not that slow when compared to others, even in PCI mode (and slow P4) I was not the last on the table w/o even any overclocking, so yes. That makes sense even in HWbot scores.


I mean the overal todays usability of R9100 is doubtfull. For 2D it is fast enought, but that it is. There is no way that it can be used to play any more recent/decent games, that was my point.


And you are right, the MSI PM8M3-V board is nightmare from start to end. Yesterday it again crashed on me when loading a level, so definitively the board is up to the blame, not the graphic card. Therefore next step - replace more mosfets in hope that this will fix it, somehow. The idea is, that during combined load (HDD, ram, GFX, sound) the powering of the northbridge get so weak to the point of crashing - and I cannot come up with anything better ATM.

Sadly I lack money to get anything more usable ATM, even the caps was bought by friend in Digikey, so... No way to escape the MSI "monster" ...


As for the R9100 - I will measure where the GPU voltage is and add caps there in hope for more MHz and we see :)

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Create New...