VOGONS


First post, by ahyeadude

User metadata
Rank Member
Rank
Member

Okay, so I've got a Pentium 233 MMX Socket 7 build (parts from 97-98). I've replaced every single electrolytic capacitor at this point on the motherboard and soundcard (some leaking, some not). I have not yet touched my Riva 128ZX because the output is perfect and the tantalum caps seem fine visually. However, I've seen a lot of posts and videos of tantalum capacitors failing spectacularly and when they do, they typically fry a motherboard trace or something else adding to the damage. This ticking time-bomb is giving me anxiety and I'm finally motivated to address the potential problem.

8rmatNu.jpg
7KCLOEg.jpg

It's been a little difficult to identify the exact specifics of these capacitors, but I'm thinking the 10uF are most likely these...

https://www.mouser.com/ProductDetail/AVX/TAZH … RekIG3yuw%3D%3D

However, I'd like to replace them with aluminum polymer caps to avoid future fireworks. These look like they should physically work, but I'm worried they are too low ESR.

https://www.mouser.com/ProductDetail/Murata-E … CvsCL7b9Q%3D%3D

Tantalum part shows an ESR of 1.4 ohm and the polymer shows an ESR of 0.04 ohm. Ripple on tantalum is 0.3 A and polymer shows 1.6 A.

Will these differences matter for this application? Has anyone gone down this path before? Any suggestions?

Thanks!

Reply 1 of 20, by Byrd

User metadata
Rank Member
Rank
Member

Tantalum caps from a good quality late 90's device are not going to spontaneously combust any time soon - I'm not sure why you are so anxious; caps like these are in many vintage devices you probably own and not needing replacement. In the rare event they do blow, the damage is often just around the tantalum cap footprint itself and easily replaced.

Reply 2 of 20, by matze79

User metadata
Rank l33t
Rank
l33t

Is this really 8Mb ? looks like 4Mb
isnt 8Mb ZX SDRAM only ?

https://www.retrokits.de - blog, retro projects, hdd clicker, diy soundcards etc
https://www.retroianer.de - german retro computer board

Reply 3 of 20, by Garrett W

User metadata
Rank Oldbie
Rank
Oldbie

No, the ZX exists in SGRAM form as well and it's the one to look for as the SDRAM version is only 64bit and thus very slow.

Reply 4 of 20, by ahyeadude

User metadata
Rank Member
Rank
Member

Yea, says 8MB in the bios message at post as well as DirectX config in Windows 95. Datasheet for the memory chips show they are 2MB SGRAM.

https://pdf1.alldatasheet.com/datasheet-pdf/v … 636327Q-10.html

I went ahead and ordered the aluminum poly caps, as well as a new e-cap to replace that tiny 4mm SMD one which is likely dried out. I know I am being paranoid, but these things are 22 years old at this point and I don't want to risk the destruction if one of the tants shorts out. I'll post an update once they are replaced.

Reply 5 of 20, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

Riva 128ZX came in 3 flavors:
4 Mb SGRAM 64-bit bus (only 2 chips on board)
8 Mb SDRAM 64-bit bus
8 Mb SGRAM 128-bit bus

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 6 of 20, by ahyeadude

User metadata
Rank Member
Rank
Member

Replaced the caps, perfect fit-wise. And even though the polymer are much lower ESR, it didn't seem to effect anything. Output is still great.

Reply 7 of 20, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

I've been looking to put tantalums into any retro video card I can, and have so far seen only the Viper 330 come with what look like derated tantalums. The 10 uF 25 V caps would be 16 V in Rivas that use wet alus, and other tantalum-using variants I've seen have 16 V as well. I assume these parts see about 12 volts in use, so 25 V agrees with the modern paranoid recommendation of 50% derating, not likely to fail in a lifetime. The other tantalums might expect to see some 5 volts I'm guessing, so their 16 V is plenty headroom as well.

Are there other retro graphics cards besides the Viper 33o where the tantalums were derated by more than the alternative wet alus?

Reply 8 of 20, by Dorunkāku

User metadata
Rank Member
Rank
Member
vvbee wrote on 2025-03-24, 10:24:

I assume these parts see about 12 volts in use, so 25 V agrees with the modern paranoid recommendation of 50% derating, not likely to fail in a lifetime.

No, the 12 Volt fingers on the PCI connector are not populated. The highest voltage the caps on this card will see is 5 Volt.

Reply 9 of 20, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

True, with a multimeter my Viper 330 PCI measures 5 volts across the 47 uF next to the VGA connector and 3 volts on the smaller ones, on the DOS bootup screen anyway. C52 I didn't measure because of a larger heatsink in the way. This card is a later revision than the OP's and the 10 uF tantalums are rated for 35 V, not sure why Diamond chose to seemingly over-derate the 10 uF like this. Could just be availability but I'd imagine 35 V tantalums were much more expensive than the 16 V others seem to have used. Also raises the question why even 16 V if the baseline expectation in use is 3 volts.

Reply 10 of 20, by zuldan

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote on 2020-08-07, 15:09:
Riva 128ZX came in 3 flavors: 4 Mb SGRAM 64-bit bus (only 2 chips on board) 8 Mb SDRAM 64-bit bus 8 Mb SGRAM 128-bit bus […]
Show full quote

Riva 128ZX came in 3 flavors:
4 Mb SGRAM 64-bit bus (only 2 chips on board)
8 Mb SDRAM 64-bit bus
8 Mb SGRAM 128-bit bus

I’m guessing this is a Riva 128ZX with slow SDRAM 🙁 ?

The attachment 3B04E120-6C59-477E-8DB1-A9376581BEE3.jpeg is no longer available

Reply 11 of 20, by Garrett W

User metadata
Rank Oldbie
Rank
Oldbie

No, that is quite clearly SGRAM.

Reply 12 of 20, by zuldan

User metadata
Rank Oldbie
Rank
Oldbie
Garrett W wrote on 2025-03-25, 10:43:

No, that is quite clearly SGRAM.

Wow I’m shocked. I got the card for a good price and it’s not a fancy Diamond one. I was convinced it had the slow memory. I’m a memory noob. How do you tell if it’s SGRAM or SDRAM?

Reply 13 of 20, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

Usually by memory package. This is clearly SGRAM package. SDRAMs are normal rectangular chips, same as motherboard RAM.

Wow I’m shocked. I got the card for a good price and it’s not a fancy Diamond one.

Riva 128ZX was never a premium card to begin with. It was a budget rehash.

Last edited by The Serpent Rider on 2025-03-25, 11:56. Edited 1 time in total.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 14 of 20, by zuldan

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote on 2025-03-25, 11:51:

Usually by memory package. This is clearly SGRAM package. SDRAMs are normal rectangular chips you can see on motherboard RAM sticks.

Ok I’m confused, the chips on this card are rectangular so that means it’s SDRAM?

Reply 15 of 20, by paradigital

User metadata
Rank Oldbie
Rank
Oldbie
zuldan wrote on 2025-03-25, 11:55:
The Serpent Rider wrote on 2025-03-25, 11:51:

Usually by memory package. This is clearly SGRAM package. SDRAMs are normal rectangular chips you can see on motherboard RAM sticks.

Ok I’m confused, the chips on this card are rectangular so that means it’s SDRAM?

I think he was referring to the package type. SDRAM of this vintage is SOP/TSOP package (pins on two opposing edges of the package). SGRAM is QFP typically (pins on all 4 edges).

Reply 16 of 20, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote on 2025-03-25, 11:51:

Riva 128ZX was never a premium card to begin with. It was a budget rehash.

Anything with tantalums instead of cheaper options is premium by definition.

Reply 17 of 20, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

*laughs in old motherboards going in flames with tantalum caps*

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 18 of 20, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote on 2025-03-25, 20:19:

*laughs in old motherboards going in flames with tantalum caps*

Far more likely to see degraded alus corroding the board and repeated cycles of thermal stress to keep replacing them.

mno2-vs-polymer_img1.jpg

Reply 19 of 20, by shamino

User metadata
Rank l33t
Rank
l33t

Tantalums really just hate being left idle for a long time. If they get routine use then they'll probably last forever. But now that these devices have graduated to a sleepy lifestyle, we do see how angry they can be about waking up again.

vvbee wrote on 2025-03-24, 19:47:

True, with a multimeter my Viper 330 PCI measures 5 volts across the 47 uF next to the VGA connector and 3 volts on the smaller ones, on the DOS bootup screen anyway. C52 I didn't measure because of a larger heatsink in the way. This card is a later revision than the OP's and the 10 uF tantalums are rated for 35 V, not sure why Diamond chose to seemingly over-derate the 10 uF like this. Could just be availability but I'd imagine 35 V tantalums were much more expensive than the 16 V others seem to have used. Also raises the question why even 16 V if the baseline expectation in use is 3 volts.

The parts were ordered in bulk for any card Diamond might make, so they probably wanted a cap they could use safely on 12V if some card needed that. It would make sense for them to order a bulk supply of 10uF 25V, not sure why they went up to 35V though - maybe just the way pricing and availability was at that moment.
If 10V or 16V was appreciably cheaper then they could have stocked multiple voltages of course. But there's also an advantage to reducing the number of different parts that the machines have to handle separately.

I don't know if Diamond did their own manufacturing or just contracted it out. If it was outsourced then who knows what else that manufacturer was building, so maybe for them 35V was handy and Diamond would have listed them as acceptable.