I'm trying to upgrade the SIMMs in my 8580 with Reply Turbo Board with 16MB ones, but every single stick I have gives me random NMI (parity) errors. All my sticks are 9-chip 60ns Kingston branded. I've tried both parity and ECC (separately of course, both are supported according to the manual). There's no way I have more than a dozen dead sticks, so there must be some some sort of compatibility issue here.

As far as I can tell, the fastest this motherboard takes is 70ns. Is it possible the ram is somehow unsuitable because it's too fast? Or is it possible that the motherboard doesn't like the chip configuration that these sims use (9-chip vs. 12-chip, etc)? The motherboard works totally fine with the original 12-chip 70ns 4MB sticks, so I don't think the planar is broken.

Anyone have any idea what could be going on?

On Wednesday, June 28, 2023 at 10:04:33 PM UTC-4, William Murray wrote:
> I'm trying to upgrade the SIMMs in my 8580 with Reply Turbo Board with 16MB ones, but every single stick I have gives me random NMI (parity) errors. All my sticks are 9-chip 60ns Kingston branded. I've tried both parity and ECC (separately of course, both are supported according to the manual). There's no way I have more than a dozen dead sticks, so there must be some some sort of compatibility issue here.
>
> As far as I can tell, the fastest this motherboard takes is 70ns. Is it possible the ram is somehow unsuitable because it's too fast? Or is it possible that the motherboard doesn't like the chip configuration that these sims use (9-chip vs. 12-chip, etc)? The motherboard works totally fine with the original 12-chip 70ns 4MB sticks, so I don't think the planar is broken.
>
> Anyone have any idea what could be going on?

I have a Model 80 with a Reply TurboProcessor also, I actually picked it up via eBay from Greg Frantz who's posted on here in the past. It sits under the same desk that has my main gaming rig on it and I use it often. As far as the 128MB ram battle, I have fought this battle for a long time... and continuously lost as well. It has been an expensive one...... I've even tried a bunch of 16MB SIMMs, FRU 43G1796, 60G2950, even tried EOS RAM, FRU's 11H0622 and 11H0647... all the same thing. Instability. I worked hard to find cheap EOS RAM because I've seen pictures of this board populated with it and was super disappointed.

I did a lot of searching of this news group and I'd be surprised it if it was ever working right with 128MB. I'd have to find it it somewhere, but I swear I found a comment somewhere that mentioned received a 128mb kit from Reply directly and it still didn't work. Right now I have it reliably working with 8MB SIMMs at 64MB and found a 128kb L2 cache for it on the internet. I also found that there is a Reply BIOS and reference disk and an IBM branded BIOS and Reference disk, but the IBM one is much more stripped down. Not sure if the RAM support is more stable and didn't want to chance it. I'm currently running the 1.38 Beta that correctly identifies the POVD83 as 84Mhz. I have also tried these SIMMs with a Kingston Turbochip 133 and an Intel 486 Overdrive (DX4OPDR100) with no improvement in stability.

What CPU are you running with yours? I'm run the Pentium Overdrive 83Mhz, but it's about 30% slower than the same Pentium Overdrive running on a PS/2 Valuepoint. Speedsys is like 41 vs 60 on the Valuepoint. Other bench mark utilities show the same differences... even against Youtubers randomly benchmarking the POVD 83 on whatever rando machine they are benchmarking. It really makes me think this board is slow which... it has the Syncrostream controller that the 95's had. So what gives? Louis? David? How can the PS/2 Valuepoint 6382 K51 (http://ps-2.kev009.com/pcpartnerinfo/ctstips/3aa2.htm), with the same Pentium Overdrive, or any other generic ISA board whoop it's ass, by a third?

I'm still experimenting with the onboard IDE, but it seems like the Spock and Adaptec AHA-1640 are the best SCSI cards for this board. I haven't been able to find a fairly priced BusLogic card.

Sorry I went on and ranted a bit hahaha

That’s a bummer. I guess maybe 128MB is impossible. I was beginning to think it was because I used 9-chip SIMMs with quad-CAS parity chips, but it seems like you tried 12-chip parity SIMMs and yours still didn’t work. It’s strange; I feel as though stability decreases over time, but if you give the RAM sticks a rest, they go back to being more reliable. Kinda like CPUs when they overheat/get overclocked. I wonder if something like a memory controller could be overheating (IDK if these boards have such a thing)?

I’m just using the board how I got it for now (486DX2-66, no cache).. For storage I have a 2MB Spock with an IBM DPES-31080 HDD. I’ve tried other CPUs in it though, with no change in reliability (Overdrive ODP DX2, 5x86 AMD & Cyrix). Maybe I should also test 25MHz vs 33MHz bus speeds though. I also bought a Pentium Overdrive and 128kb cache module and they’re on their way, so if you’re curious about comparing benchmarks I can do that too.

On Wednesday, July 5, 2023 at 9:29:08 AM UTC-4, William Murray wrote:
> That’s a bummer. I guess maybe 128MB is impossible. I was beginning to think it was because I used 9-chip SIMMs with quad-CAS parity chips, but it seems like you tried 12-chip parity SIMMs and yours still didn’t work. It’s strange; I feel as though stability decreases over time, but if you give the RAM sticks a rest, they go back to being more reliable. Kinda like CPUs when they overheat/get overclocked. I wonder if something like a memory controller could be overheating (IDK if these boards have such a thing)?
>
> I’m just using the board how I got it for now (486DX2-66, no cache). For storage I have a 2MB Spock with an IBM DPES-31080 HDD. I’ve tried other CPUs in it though, with no change in reliability (Overdrive ODP DX2, 5x86 AMD & Cyrix). Maybe I should also test 25MHz vs 33MHz bus speeds though. I also bought a Pentium Overdrive and 128kb cache module and they’re on their way, so if you’re curious about comparing benchmarks I can do that too.

Not sure but yes I've noticed that with the EOS RAM. It works fine and then dies after a while. Wonder if a thermal camera would help? Haha.

Really surprised you found a 128kb cache module! I searched a long time and ended up buying a Honeywell PLC board, that wasn't listed for $14,000...... on a make an offer... I still really over payed for it... Did you happen to win Betty's auction with the NCR processor board? That had the cache card and I was considering bidding to have an extra or to hold onto for someone who had a reply board. I would definitely be interested in comparing benchmarks. I've also been having a difficult time finding compatable 1mb SIMMs for the Spock. I use a ZuluSCSI RP2040 with the Spock using a 3..5 bay mounting adapter: https://ebay.com/itm/304399284311

Yes I just bought the NCR board. I’m very happy with the price I paid for it. The last time that seller sold the same item, I gave up on the auction because it got to $350+. For my Spock, I made my own 1MB modules by modifying normal ones. I’m not sure how much of a difference cache makes though, and I bet you get even less of an impact with solid state storage.

I just did some preliminary testing, and I also got 41 via speedsys for the 8580 TurboProcessor with Pentium Overdrive. My Eduquest 55 with the same Pentium Overdrive scores 61.

Other benchmarks show unexpectedly low performance as well, which is weird since most artificial CPU benchmarks are not sensitive to other stuff on the motherboard. For example, in Norton Sysinfo 8, my 8580 TurboProcessor only scores 100. HOWEVER, my IBM XT with Cyrix 5x86-120 scores ~250! It's a very strange hacked together system that uses an Intel Inboard 386 and a Transcomputer 386-486 adapter, so it has mediocre RAM speeds with a cripplingly slow 4.77MHz bus, so obviously the test only cares about the CPU; I have no idea why the TurboProcessor should score so low. The POD-equipped Eduquest with Sysinfo scores ~260. I can't run speedsys though on my XT, so IDK what it would be like.

I am using the 1.31 BIOS. As far as I can tell, 1.31 works with the Pentium just fine. It identifies the processor as P24T, but erroneously marks it as 66MHz (chkcpu identifies it as operating at the intended 83MHz though).

Also, the Eduquest does WB L1 cache instead of WT like the 8580, so that could at least partially explain things. Also, I don't know if L2 Cache is used for speedsys, but I have 256KB in the Eduquest and 0KB in the 8580.

I wonder if it could have to do with some Pentium-specific registers not being set? It almost feels like the cache is not being used, or some enhancement has not been properly activated. I also might want to try out a Cyrix 5x86-133 and see how it goes, since messing with the performance-enhancing registers on those CPUs is easy, though I need to get a heatsink for that first.

On Wednesday, July 5, 2023 at 10:17:46 PM UTC-4, William Murray wrote:
> I just did some preliminary testing, and I also got 41 via speedsys for the 8580 TurboProcessor with Pentium Overdrive. My Eduquest 55 with the same Pentium Overdrive scores 61.
>
> Other benchmarks show unexpectedly low performance as well, which is weird since most artificial CPU benchmarks are not sensitive to other stuff on the motherboard. For example, in Norton Sysinfo 8, my 8580 TurboProcessor only scores 100. HOWEVER, my IBM XT with Cyrix 5x86-120 scores ~250! It's a very strange hacked together system that uses an Intel Inboard 386 and a Transcomputer 386-486 adapter, so it has mediocre RAM speeds with a cripplingly slow 4.77MHz bus, so obviously the test only cares about the CPU; I have no idea why the TurboProcessor should score so low. The POD-equipped Eduquest with Sysinfo scores ~260. I can't run speedsys though on my XT, so IDK what it would be like.
>
> I am using the 1.31 BIOS. As far as I can tell, 1.31 works with the Pentium just fine. It identifies the processor as P24T, but erroneously marks it as 66MHz (chkcpu identifies it as operating at the intended 83MHz though).
>
> Also, the Eduquest does WB L1 cache instead of WT like the 8580, so that could at least partially explain things. Also, I don't know if L2 Cache is used for speedsys, but I have 256KB in the Eduquest and 0KB in the 8580.
>
> I wonder if it could have to do with some Pentium-specific registers not being set? It almost feels like the cache is not being used, or some enhancement has not been properly activated. I also might want to try out a Cyrix 5x86-133 and see how it goes, since messing with the performance-enhancing registers on those CPUs is easy, though I need to get a heatsink for that first.

It's odd isn't it? Your results are identical and I found the L2 cache didn't do much, but I found it didn't do much for CPU tests on the Valuepoint either. I figure that the boost from the cache will be application or game specific. I will say this, the Kingston TurboChip and the 100Mhz 486 Overdrive also score much less than in other systems so I didn't feel that it was Pentium Specific. There is a jumper on the bottom left of the board, J14, that is undocumented and I've even tried flipping that jumper with no help either. I've never been able to tell what it does.