BFG Tech. GeForce FX5200 [ASLM52256] – full recap info and cooler mod

[momaka]

The more time passes, the more nostalgic I get with old hardware. Video cards and motherboards tend to be my weak spot. As such, I often scavenge eBay for old PC hardware, especially anything with bad caps or in need of similar minor repair.

For today's consideration, I have a popular low-end card from back in the early 2000's: an nVidia GeForce FX5200. This one is the BFG Tech. Asylum take on the card, with 256 MB of DDR RAM running at 400 MHz and 128-bit bus (vs. 64-128 MB and 64-bit bus on many of the cheaper versions.)




I think it's easy to see what's wrong with this card. Bad caps are one of my favorite problems on old hardware, as this is usually all that's wrong. Although FX5200 cards are hardly worth new caps (given their performance), I still found it absolutely worth it for the fun of getting old hardware fixed. At $10 total for the card (including getting it shipped to my door), why not, right?

The card came exactly as pictured… but no mention in the auction that the fan on the cooler was also bad.
https://www.badcaps.net/forum/attach...1&d=1565539312

Not that I'd cry about another dead 40 mm fan, as those things are too loud anyways. What worried me is that the card looked overheated on the back of the GPU area, probably due to the fan seizing. Nevertheless, I crossed my fingers and started working on the recap.

First step (for me, anyways), is identify the various rails on the card. I like doing this and making a little reference sheet, so that I can decide what caps to use (I rarely use identical spec caps on video cards, unless I happen to have them.)



For those who can't log in and view the pictures, here is the text version:

Electrolytic caps in spots EC6, EC7, EC1**, and EC2** ---> 3.3V rail from PSU
Electrolytic caps in spots EC12 and EC13 ---> GPU V_core
Electrolytic caps in spots EC5, EC8**, EC9**, and EC10** ---> RAM V_ddr (2.5V)
Electrolytic caps in spots EC14 and EC4** ---> GPU V_tt

** ---> free / empty cap spots that can be populated, if desired

Note: since most rails are 3.3V or less, 6.3V caps will be absolutely fine to use for the entire card. In fact, even 4V (polymer) caps could be used on some of the rails – though I personally don't recommend using polymers on this card, as there won't be any benefit. If you do, only do the GPU V_core rail, as that is a buck-type regulator. RAM V_ddr and GPU V_tt are linearly generated, so ultra-low ESR caps are not needed at all and, in fact, may be an issue (I didn't try, so I don't know – consider it experimental if you do that.) Unlikely. But I still don't recommend it.

Speaking of caps and what to use… I did a test recap first (again, due to card looking overheated) with a bunch of OST RLP 6.3V, 1000 uF caps I pulled from a board that already had many of these same ones bad (the few I used were tested good on my ESR meter.) Mixed those with a few new Rubycon ZLH 6.3V, 820 uF caps to keep the ESR low (particularly on the GPU V_core.) And just for testing, I used a fan from another video card I bought and modded around the same time (a GeForce 7300 GT AGP version). So here is the temporary recap with the temporarily-fixed cooler:


Ugly – yes! But it worked for a full test just fine. As such, I opted to do a more proper recap. Again, it may not seem worth it on such an old and slow card. Considering that this card survived all these years (including with the seized fan), however, I think it has already proven itself that it deserves a second chance at life.

So… step 2: removed old cooler and recap:


Here is the old HS without the fan:
https://www.badcaps.net/forum/attach...1&d=1565539312

And here is what I came up with (again) to replace that original cooler:
https://www.badcaps.net/forum/attach...1&d=1565539312
It's an Xbox 360 GPU rev1 heatsink, but with a few cuts and trims to make it fit on the board.

The recapped and modded video card:
https://www.badcaps.net/forum/attach...1&d=1565539312

I didn't stop there, of course. Despite the fact that the above modded heatsink can probably cool the FX5200 GPU passively, I still attached a small 50 mm fan.


To make the fan more quiet, I rewired the fan connector to supply only 7V to the fan, I did this by removing Q11 and adding a wire between the fan connector's ground/black (-) lead and a 5V rail point on the board.
https://www.badcaps.net/forum/attach...1&d=1565540075
So now, the fan essentially runs on 7V (difference between PSU's 12V and 5V rails.)

And that is all for this video card recap, really. The whole card runs very cool to the touch, even at ~30°C / 86°F room temperatures.

[momaka]

BFG Tech. GeForce FX5200 [ASLM52256] – GPU-Z and performance

For those interested in the card’s specs and performance, here's a GPU-Z screenshot and a little bit of testing:


Ooops! Looks like the previous owner must have flashed it to a Quadro card. Either that, or GPU-Z is just confused. Whichever the case, it's still an NV34 GPU running at the correct core and memory clocks for the FX5200.

Again, worth noting here is that this is the version with 256 MB, 128-bit bus, and probably the fastest DDR RAM that these FX5200 cards come with. Only the FX5200 Ultra come with faster RAM… and that’s really the only thing they have going for them.

In terms of performance… wait, what performance?? The GeForce FX5200 cards are weak! Although my card is noticeably better than the 64-bit variants (I also have an MSI FX5200 with 128 MB of RAM and 64-bit bus, which I got in a pack of other eBay cards a while back), this doesn’t say much.

Game-wise, the BFG FX5200 struggled to do 1024x768 on most games from its era (NFS Hot Pursuit 2, NFS Underground, Half-Life 2, and Collin McRae Rally 3 & 4) that I threw at it. On the other hand, 800x600 was noticeably better (though still not great by any means). At 800x600, I got “acceptable” (for the time) FPS – about 30 or so. Colin McRae Rally 4, for example, managed to do about 30 FPS average (low 25, high 65) at 800x600 if I turned some settings Off, like Car Shadows, and around 23-25 FPS average at 1024x768. Need For Speed Hot Pursuit 2 also ran very well… for the most part: although it managed around 50 FPS average (and often peaking up to 80), there were also many frame drops and dips in the 20 FPS –range – mostly when there were a lot of cars around on the map. Surprisingly, the FPS didn’t change much for this game at 1024x768. Thus, it’s possible the issue might have been something else too, because I noticed the same behavior even with more powerful nVidia cards (maybe lack of driver optimization for that game?). Meanwhile, NFS Underground worked quite well at 1024x768 – better than 800x600, actually. Reason why is because NFS U has built-in automatic frame-limiter (and V-sync possibly?), which works at either 25, 30, or 60 FPS. At 800x600, the game couldn’t decide whether to settle on 30 or 60 FPS, so I got bad “rubber-banding” effects as FPS jumped between the two thresholds. Once I bumped up the resolution to 1024x768 (and lowered a few visuals, like Reflection Update Rate), the game settled and stayed at 30 FPS, rarely dipping to 25.

The most surprising results on this card were with the Source engine. I tried Half Life 2 EP2 (“no-steam” version to reduce overhead from Steam platform on the old hardware I use for testing). As this game defaulted to DX8.1 rendering (no HDR and dynamic light effects), I was able to pull off around 40 FPS average (low 30, high 80’s) @ 800x600. Only caveat: don’t use the flashlight, as that brought down the FPS to sub-35 FPS. I then tried 1024x768 and the framerate dropped further to 25-30 frames AVG (with no flashlight), and around 20 with flashlight On.

As my cooler mod and recap above was substantially better than what the card came with, I decided to try some overclocking. I had no problems pushing the slider for the GPU core all the way to GeForce FX5200 “Ultra” speeds (325 MHz vs. 250 MHz stock) But over an hour of playing, I found the card would eventually crash in games. So I dropped the GPU clock down to 320 MHz, and it was perfectly stable. Memory-wise, I only tried pushing the onboard Samsung K4d551638d-TC50 chips an extra 25 MHz, which they took with no problems. Probably could do more, but I didn’t try. Reason why is because I noticed that overclocking this card pretty much didn’t yield any significant performance gains. For example, in CMR4, the OC did very little to help in areas of the game where I was getting low framerates (which at 800x600 was typically 25 FPS minimum) – maybe 1 to 2 FPS increase. The OC really only increased the FPS in areas already running at “high” framerates (from ~50 to ~60 in CMR4 and even more wild swings in HL2 EP2)… which I don’t really care about, because after all, it’s the low FPS areas that matter. So overclocking on this particular card is pointless.

Clearly, the FX5200 card is not strong at all – even this 128-bit 256 MB version. Its sweet spot resolution (at least for most games) is 800x600, which is not only laughable these days, but was also considered pretty marginal even for its times. Nevertheless, remember that in the early 2000’s many folks still used CRT monitors. Thus, 800x600 on an average 17” CRT didn’t look that bad actually. On that note, I plan to use this BFG FX5200 with my Pentium 3 PC, which IS hooked to a 17” CRT.

Moreover, that PC doesn’t really have any demanding games on it. At most, it will be asked to run NFS Hot Pursuit 2 and NFS Underground. In reality, I’ll probably be using it more to play Counter-Strike 1.5 and Half-Life Deathmatch (both being original GldSrc engine games! ) and maybe NFS Porsche Unleashed and High Stakes (both of which run OK even on Intel Extreme Graphics 2 / i865 IGP, if that is to tell you anything ) with my nephews. So, really, I’m not disappointed at all with this video card, despite its not-so-strong performance. For what I would like it to do, I think it will suffice. And it’s also one less item from the AGP era / Windows XP days going into the trash.

[Dan81]

Re: BFG Tech. GeForce FX5200 [ASLM52256] – full recap info and cooler mod

I know I have the 5500 variant of it, although mine comes from Chaintech rather than BFG, and it has DVI as well.

[momaka]

Re: BFG Tech. GeForce FX5200 [ASLM52256] – full recap info and cooler mod

I know I have the 5500 variant of it, although mine comes from Chaintech rather than BFG, and it has DVI as well.

Yes, I saw that same PCB I posted above on a number of FX 5200 and FX 5500 cards from different manufacturers.

According to GPUreview.com pictures, the following cards use this same exact PCB (or nearly identical):
Albatron FX5200EQ
Albatron FX5200U 2.0
Chaintech A-FX20-N-128
Chaintech A-FX20-N-256
Chaintech SA5200-128
Chaintech SA5200-256
Chaintech GeForce FX 5500 128/256 MB
BFG GeForce FX 5500 OC 128/256 MB

The only difference you might see, of course, is the number of memory chips. The 128 MB cards (and also likely 128-bit bus cards) will likely only have 4 out of the 8 memory spots populated. And like you noted, yours came with a DVI port. My FX 5200 above has the space for it, but it just wasn't installed (which I don't mind, as I don't care much for DVI anyways.)

This isn't really surprising to me anymore, as I learned over the years that BFG, Chaitech, PNY, Sparkle, Albatron (and probably a bunch of others I didn't list here) are basically re-branding companies - that is, they buy already made cards from some manufacturer and simply just put their logo/brand on it. The BFG GeForce 7300 GT I mentioned above in my first post, for example, shares the same PCB as a PNY GeForce 7600 GS that I have. And I think Sparkle also used that PCB for their 7300/7600 AGP cards.

[Dan81]

Re: BFG Tech. GeForce FX5200 [ASLM52256] – full recap info and cooler mod

Chaintech SA5200-256

Mine reads SA5500T4-256D-MC85. Has 5ns chips, although I don't know how much that would matter, since if I wanted something great, I already have a 3D Prophet Radeon 9700 Pro 128MB.

[momaka]

Re: BFG Tech. GeForce FX5200 [ASLM52256] – full recap info and cooler mod

Yes, the FX5200 in this thread also has 5 ns Samsung chips. I tried overclocking them by 25 MHz (50 MHz DDR) and... didn't really see any significant difference in performance. Probably could overclock even more, but I don't see why bother. The FX5200 and FX5500 cards are low-end cards from that era, and they will always perform as such.

The 9700 Pro (and even non-pro) are miles ahead of them. In fact, even the FX5600, which isn't that great, is also miles ahead of the FX5200/5500. And what I find even more funny is that even the GeForce 4 Ti 4200 is faster.

Now, if I wanted something good (though not "the best") for AGP, I just use my Sapphire Radeon X800 GTO that has all of its pixel and vertex shaders unlocked (16p/6v).

[Dan81]

Re: BFG Tech. GeForce FX5200 [ASLM52256] – full recap info and cooler mod

Yes, the FX5200 in this thread also has 5 ns Samsung chips. I tried overclocking them by 25 MHz (50 MHz DDR) and... didn't really see any significant difference in performance. Probably could overclock even more, but I don't see why bother. The FX5200 and FX5500 cards are low-end cards from that era, and they will always perform as such.

The 9700 Pro (and even non-pro) are miles ahead of them. In fact, even the FX5600, which isn't that great, is also miles ahead of the FX5200/5500. And what I find even more funny is that even the GeForce 4 Ti 4200 is faster.

Now, if I wanted something good (though not "the best") for AGP, I just use my Sapphire Radeon X800 GTO that has all of its pixel and vertex shaders unlocked (16p/6v).

What about the X1650 Pro? I have one of these in working condition that just needed a new cooler as the old one was bad.

[momaka]

Re: BFG Tech. GeForce FX5200 [ASLM52256] – full recap info and cooler mod

What about the X1650 Pro? I have one of these in working condition that just needed a new cooler as the old one was bad.

x1650 is pretty nice too, but the x800/850 still beat it in terms of "raw" performance. Mostly has to do with the fact that the x1650 has only 4 ROP and 4 TMUs, whereas the x800/850 have 12 to 16 (depending on variant.) So this allows them to run higher resolutions without choking. At lower resolutions (if you don't mind them), the difference between x1650 and x800/850 isn't much.

That said, worth noting is that the x1650 is DX 9c compliant, whereas the x800/850 are only DX 9b. So in some games, like Half-Life 2, you will have HDR available with the x1650.

In terms of overall rank, I think it goes similar to this:
Radeon 9200 < Radeon 7500 < Radeon 9000 < FX 5200 < Radeon 8500 < GeForce 4 TI 4200 < GeForce 4 TI 4400/4600 and FX 5600/5700 < Radeon 9600 < Radeon 9500 / 9700 < FX 5800/5900 < GeForce 7300 GT < Radeon 9800 < 6600 GT < x1600 / x1650 < GeForce 7600 GS < HD2600 & HD3650 Pro AGP < x800 (Pro, GTO, XL) and GeForce 6800 (GS, XT, vanilla, GT, Ultra) < 7800 GS < HD4650 AGP < HD3850 AGP.

Let me know if I missed one or if not accurate (waiting for ChaosLegionnaire to chime in... about now? )