dimm(sdram) are everywhere in the world, simm(fpm/edo) are getting rarer and harder to find nowadays.
however if you want to use them on a socket7 board, simm is slower than sdram and can take a huge 10% performance hit, so sdram is the better choice.
and if you want to keep some simms, i suggest that only 32mb and 64mb simm modules are worth keeping, 16mb and 8mb simms are too small for even a 486 rig.

Both the 8mb simms and the 32mb dimms aren't really anything special as both are almost the smallest ones of their kind.
Personally I'd keep any simm that's 16mb or larger (I don't even have any 64mb simms, except some of those composite modules) but the truth is, smaller modules aren't really worth selling either (unless theres something special about those modules).

Thanks, makes sense, mostly. Except:

noshutdown wrote:

16mb and 8mb simms are too small for even a 486 rig.

What would you want to run on a 486 that would require more than 8 or 16 MB of RAM? (Yes, you could run Windows 3.x or Windows 95, but you really wouldn't want to do that on a 486 with so many other more suitable options relatively readily available, I'm thinking.)

How much memory is a 486 capable of caching anyway? It's probably not a good idea to install more than that or the system will crawl.

sliderider wrote:

How much memory is a 486 capable of caching anyway? It's probably not a good idea to install more than that or the system will crawl.

Iirc it was 64mb with 256kb cache set to write through (setting it to write back will cut the cacheable area in half).

Correct me if I'm wrong 🤣

Jorpho wrote:

Thanks, makes sense, mostly. Except:

noshutdown wrote:

16mb and 8mb simms are too small for even a 486 rig.

What would you want to run on a 486 that would require more than 8 or 16 MB of RAM? (Yes, you could run Windows 3.x or Windows 95, but you really wouldn't want to do that on a 486 with so many other more suitable options relatively readily available, I'm thinking.)

Personally I don't find 16mb simms to be too small. Usually Socket 7 boards will have 4 simm slots and a cacheable area of 64mb, so you could put 4x16mb in there and have the exact amount of cached memory, so to say.

Depends on what you're doing with the box. On a 486, there's not much you'd want to do that would max out the ram. A pentium is fast enough to make, e.g. web browsing with Firefox 2.0 feasible. That would max out 64mb pretty quickly, and it's better to have 256 of non-cachable ram than to be swapping all the time. But if you're just using period software, 64mb is pretty good. That's what I run in mine.

Personally I suggest that you just keep all of it provided it worked and even if it is worthless right now that won't be the case years from now. As for sdram I got a huge sack of it (17gb worth of 64 and 128mb sticks) and a box of rdram for hardly nothing.

nforce4max wrote:

Personally I suggest that you just keep all of it provided it worked and even if it is worthless right now that won't be the case years from now. As for sdram I got a huge sack of it (17gb worth of 64 and 128mb sticks) and a box of rdram for hardly nothing.

I got some leftover sticks of rdram also...just no boards to put them in! 😵

It is very interesting type of ram but I only have one board that uses it, a huge dual socket 603 server board. The backplane just for the ram is bragging rights alone considering it has 8 slots rather than the usual 4 and I can upgrade all the way to 4GB minus a little (3.2gb). I did find a rdram 478 board that is likely from a Dell sitting at a shop right now in a parts bin but it is banged up a little.

I don't think the "10%" performance hit is really as important as the marketing propaganda suggests. I believe the 10% hit only affects memory performance, not the overall system performance (which might increase by 1 or 2% at best). Has anyone on VOGONS actually tested SIMM vs SDRAM in various configurations and benchmarks to check the validity of the claims? I would guess that the 10% performance increase was only a best case scenario. I ran both EDO and SDRAM in Pentium systems and could never tell the difference.
In my opinion the only real benefit of using SDRAM was that you could install them one at a time rather than in pairs. This was kind of offset by compatibility issues however. I found Socket7s are pretty picky about which DIMMs they will accept. It's kind of like how it's hard to get certain 72 pin SIMMs going in 486 boards. I found 72-pin SIMMs in P1 boards were ideal because they seemed to take just about combination you threw at them.

Has anyone on VOGONS actually tested SIMM vs SDRAM in various configurations and benchmarks to check the validity of the claims?

I haven't done a side-by-side comparison myself, but you may find this to be of interest.

It took me quite awhile to dig that page up, and I still can't quite believe I remembered its existence in the first place, but there it is. It's an old MSI page (from back when they actually still included useful information on their site) showing an apples-to-apples comparison of EDO and SDRAM across a bunch of different benchmarks in one of their 430TX-based boards.

Anonymous Coward wrote:

I don't think the "10%" performance hit is really as important as the marketing propaganda suggests. I believe the 10% hit only affects memory performance, not the overall system performance (which might increase by 1 or 2% at best). Has anyone on VOGONS actually tested SIMM vs SDRAM in various configurations and benchmarks to check the validity of the claims? I would guess that the 10% performance increase was only a best case scenario. I ran both EDO and SDRAM in Pentium systems and could never tell the difference.
In my opinion the only real benefit of using SDRAM was that you could install them one at a time rather than in pairs. This was kind of offset by compatibility issues however. I found Socket7s are pretty picky about which DIMMs they will accept. It's kind of like how it's hard to get certain 72 pin SIMMs going in 486 boards. I found 72-pin SIMMs in P1 boards were ideal because they seemed to take just about combination you threw at them.

even k6-2 running 66*6 is much slower than 100*4(both using sdram)... so what do you think of edo vs pc100?

noshutdown wrote:

Anonymous Coward wrote:

I don't think the "10%" performance hit is really as important as the marketing propaganda suggests. I believe the 10% hit only affects memory performance, not the overall system performance (which might increase by 1 or 2% at best). Has anyone on VOGONS actually tested SIMM vs SDRAM in various configurations and benchmarks to check the validity of the claims? I would guess that the 10% performance increase was only a best case scenario. I ran both EDO and SDRAM in Pentium systems and could never tell the difference.
In my opinion the only real benefit of using SDRAM was that you could install them one at a time rather than in pairs. This was kind of offset by compatibility issues however. I found Socket7s are pretty picky about which DIMMs they will accept. It's kind of like how it's hard to get certain 72 pin SIMMs going in 486 boards. I found 72-pin SIMMs in P1 boards were ideal because they seemed to take just about combination you threw at them.

even k6-2 running 66*6 is much slower than 100*4(both using sdram)... so what do you think of edo vs pc100?

I'd say that at 100Mhz, sdram would've been much faster. But comparing edo and pc-100 is a bit like comparing apples and bananas.
pc-100 was introduced with Super Socket 7, the original Socket 7 ran sdram at 66Mhz.

Yeah. I didn't mean to imply that PC100 SDRAM wasn't faster than 66MHz EDO. I didn't realise some were unaware of the existence of 66MHz SDRAM.

Tetrium wrote:

I'd say that at 100Mhz, sdram would've been much faster. But comparing edo and pc-100 is a bit like comparing apples and bananas.
pc-100 was introduced with Super Socket 7, the original Socket 7 ran sdram at 66Mhz.

even at 66mhz, we can think of a k6-2 running at 6x which would demand a lot more bandwidth than a pentium, and the small gap between sdram and edo can result in a larger difference of actual performance than what you see with a pentium.

No comment on that. I have never seen a K6-2 @ 400MHz running with SIMMs. I think a lot of boards in the K6-2 era were DIMM only, so it would have been a rare combination anyway. It would still be interesting to see an SDRAM/EDO comparison on such a system though.

Anonymous Coward wrote:

I don't think the "10%" performance hit is really as important as the marketing propaganda suggests. I believe the 10% hit only affects memory performance, not the overall system performance (which might increase by 1 or 2% at best). Has anyone on VOGONS actually tested SIMM vs SDRAM in various configurations and benchmarks to check the validity of the claims? I would guess that the 10% performance increase was only a best case scenario. I ran both EDO and SDRAM in Pentium systems and could never tell the difference.
In my opinion the only real benefit of using SDRAM was that you could install them one at a time rather than in pairs. This was kind of offset by compatibility issues however. I found Socket7s are pretty picky about which DIMMs they will accept. It's kind of like how it's hard to get certain 72 pin SIMMs going in 486 boards. I found 72-pin SIMMs in P1 boards were ideal because they seemed to take just about combination you threw at them.

I have but that was years ago on a old VX board that I sadly don't have anymore. There wasn't much of a difference that anyone would have noticed and in the benches showed that it was the memory controller that really determined the performance. 10ns sdram and 60ns edo, both were 128mb kits so I didn't bench within cacheable limits at that time. 110mb/s was the average and don't remember the final latency.

Anonymous Coward wrote:

No comment on that. I have never seen a K6-2 @ 400MHz running with SIMMs.

I have a K6-2 500 in this board at the moment. 😁 I don't know how stable it is, but it definitely boots...

If you can suggest a quick and dirty diagnostic program, I can give it a whirl with both the SIMMs and the DIMMs. For Science.