n3xu5 wrote:

I'm seriously leaning towards that route and placing a nice low profile copper heatsink on the K6-2/3 and overclocking it to 600Mhz. I'm thinking of a way to come up with a 3 way toggle switch for L1, L2 Cache enable disable. How did you use your CF cards for DOS? IDE > CF adapter? How well do they run or how slow are they? 😊

Yes you can buy CF > IDE adapter for next to nothing on eBay. They come as single card or double card version (so you can use 2 CF cards per IDE channel). Performance is totally fine for MS-DOS. I doubt you will complain about the performance.

Having said that I didn't do any benchmarks. You can look at my YouTube videos (the ones recorded on my SS7 machine) to get an idea, but MS-DOS isn't very taxing to be honest...

The problem with CF cards is that they dont have the wear leveling or other features that SSD's have in order to preserve performance and life. They are good for machines that load and never do any significant writes.

I think this may be worth testing. I have an ide raid card, (several actually), several scsi cards, cached and regular ide, and everything.

Maybe this weekend I'll get off my duff and do it. Naturally i'd make a new thread. and I would need to find some way to accurately benchmark the drive without windows.

But that is exactly what you'll be doing in DOS. I'll install the stuff and then only read from it. The few safegames are not going to kill it. Aside from that, just push it into your USB card reader and back up to your 2 TB HDD in case of CF failure.

Or you grab one of the earlier 30 GB SSDs and get a SATA converter. The SSDs are too small for win7 so they are sold of frequently. A SATA to IDE converter is about as cheap as the afore mentioned IDE-CF one.

I have all the parts, but wanted to use the SSD with my "soon to be" Ivy SSD-caching Mobo.

CF cards used as IDE disks are very popular in the Amiga world. I've been running one for a while in my A1200 without issues. However some users have reported their cards working perfectly then next minute, suddenly dying. Your mileage may vary, keep a backup of anything important 😀

Speaking of which, do CF-to-IDE adapters have a significant advantage over SD-to-IDE adapters? I would have thought high-capacity SD cards would be cheaper and easier to find at this point.

Jorpho wrote:

Speaking of which, do CF-to-IDE adapters have a significant advantage over SD-to-IDE adapters? I would have thought high-capacity SD cards would be cheaper and easier to find at this point.

Well, the ones you buy on ebay are cheap Chinese junk. If you need four you should buy six. Just to cover crappy shipping. As far as SD vs CF its not going to matter. Personally I weould use CF because they are so very close to IDE.

ux-3 wrote:

AFAIK, there is software to do L1 for pentiums. It even works for some P3 boards.

Right you are, but think of how convenient a toggle switch on the front of a case would be to switch between normal operating speed, L1 disabled, and L1 + L2 disabled. I think I've seen this done here before, I know it can be managed with boards that use jumpers to control all of these settings. The Gigabyte GA-5AX is a dip switch board though 😒

Does anyone know if adjusting the CPU FSB & Multiplier have any value in slow down projects?

Gigabyte GA-5AX Socket 7 Motherboard

Concerning this board & it's chipset, I've seen a thread on the forum here which report a memory caching bug on cpu's L2 cache. It can be viewed here:

question on socket7 boards onboard cache

My concern is if anyone knows whether or not it was discovered which GA-5AX revisions had this issue and which did not. Mau1wurf1977, I see you were part of that thread, did you find a solution to the 128mb problem?

Does anyone know how the Vibra 16 Sound Blaster CT4180 sounds with a decent daughterboard? I'm thinking about using this card instead of trying to find an MT-32 online.

n3xu5 wrote:

Does anyone know if adjusting the CPU FSB & Multiplier have any value in slow down projects?

The multiplier has little impact, but the FSB and also the memory settings (timing or delay). Especially when you have L1 disabled and L2 enabled you can quite significantly control performance through the FSB.

Take a look here:

L1 is disabled for all tests. When L2 on you can see quite some scaling through FSB. With L2 off not so much, but still a tiny bit. This table also shows how the choice of CPU predetermines your results.

Mau1wurf1977, I see you were part of that thread, did you find a solution to the 128mb problem?

If you go with the K6-2+ it's not an issue!

Just to clarify, when you use the K6-2+ or K6-3+ you will actually have L1, L2 and L3 cache (this is the mainboard cache). The BIOS however still refers to the mainboard cache as L2.

Now L1 and L2 are on the chip and SUPER fast compared to L3. Putting the L2 cache on the CPU is one of the main reasons these chips have good performance. All chips afterwards have the L2 on the chip and mainboard cache disappeared (which is why P3 systems are very limited for cache slow-down projects).

So yes the L3 cache won't cache memory over 128MB, but it doesn't matter as the much much faster L2 cache will do this anyway.

Does anyone know how the Vibra 16 Sound Blaster CT4180 sounds with a decent daughterboard? I'm thinking about using this card instead of trying to find an MT-32 online.

MT-32 is MT-32 but the daughter-boards are General MIDI. So this will also pre-determine what games you can use. Then there are the hanging note bugs and noise issues. So it's a very complex topic. Many use non-Creative cards, some use two Creative cards (solves the hanging note bug) and others use Roland cards (MPU401 interface).

So it really depends on what games you want to play. If all your games use MIDI then I wouldn't worry about the card having a real Yamaha OPL3.

Maybe tackle this part once the machine is up and running?

Interesting table. It confirms my own conclusion, that P1 and 486 remain much more responsive when you start to peel away the cache. With a P2 or P3 you basically drop back to a fast 286 once you turn off L1. And nothing else really matters there. Still, there is some good software to do to the slowing, starting from a low multiplier and fsb. But the P1/486 way seems preferable.

P2/P3 would be much better IF they had mainboard cache. But no such thing. SS7 are the last boards with mainboard cache.

Unless you can get PCI cache cards or something like that? Haven't come across anything yet though.

someone is offering to donate two Dell Power Edge 4100s to my retro project, are these older server boards practical for retro gaming?

2 Socket 8 Pentium Pro 180 MHz or 200 MHz
1 GiB Max 8 slots, 168-pin DIMM

Edit: ahh well the 4100s had the 200Mhz 512 cache cpu's a lot of memory and no hard drives.

n3xu5 wrote:

someone is offering to donate two Dell Power Edge 4100s to my retro project, are these older server boards practical for retro g […]
Show full quote

someone is offering to donate two Dell Power Edge 4100s to my retro project, are these older server boards practical for retro gaming?

2 Socket 8 Pentium Pro 180 MHz or 200 MHz
1 GiB Max 8 slots, 168-pin DIMM

Edit: ahh well the 4100s had the 200Mhz 512 cache cpu's a lot of memory and no hard drives.

200mhz 512cache chips are actually kinda uncommon. every one I see is 256kb and is scrapped for gold. I'd hold onto a 200/512 to be honest.

luckybob wrote:

200mhz 512cache chips are actually kinda uncommon. every one I see is 256kb and is scrapped for gold. I'd hold onto a 200/512 to be honest.

I was thinking the same, keep the cpu and memory, and toss the 75lb cube cases. If I got these 6 years ago I'd be all about case modding them, but lately I've been building smaller matx - itx systems to save desk space.

I obtained a MT-32 on ebay just recently! excuse my ignorance but it should connect to the Vibra 16 Sound Blaster CT4180?

Does anyone know how well the Vibra 16 OPL3 and MT-32 might provide compatibility with most games?

Many games will work but not all of them. The Creative cards don't have the intelligent mode compatibility which stops some games from working correctly.

But for many Sierra games at least there are patches to get around this 😀

Then you have the issue of hanging note bugs as well.

ux-3 wrote:

A K6-2 at 500 MHz is too slow to cause the RT200 bug in Turbo Pascal games while a P2-266 is fast enough already.

Wait, what? You're saying a P2-266 will trigger the bug but a K6-2 at 500 MHz won't?

(Indeed, I used to be able to trigger the bug inconsistently on a P2-333; I'm just surprised that the K6-2 is that much slower in that regard.)

Just tested this: K6-3 @ 500 MHz won't trigger the bug, but it does at 550 MHz.

5u3 wrote:

Just tested this: K6-3 @ 500 MHz won't trigger the bug, but it does at 550 MHz.

this shouldn't matter if you're lowering the multi, tweaking FSB, and disabling onboard cache right?

example: low multi, 66 or 100FSB, L1 & L2off ~ 386 33 equivalent.

5u3 wrote:

Just tested this: K6-3 @ 500 MHz won't trigger the bug, but it does at 550 MHz.

Did you give it a couple of tries? Like I said, it might not show up every time.

n3xu5 wrote:

this shouldn't matter if you're lowering the multi, tweaking FSB, and disabling onboard cache right?

As far as RT200 is concerned, it makes much more sense just to patch the executable and be done with it. (The methods for doing so are well-established at this point and should not have any effect on the core programming.)

Jorpho wrote:

Did you give it a couple of tries? Like I said, it might not show up every time.

Yeah, the RT200 bug shows up consistently at 550 MHz, at least with the two programs I tried. But apart from that, I've been running the K6-3 for years (mostly at 500 MHz), and I've never encountered the bug. I was only reminded of its existence because I've built myself an Athlon machine recently.

Whenever I want to play a game that gives that RT200 bug, I just underclock my 350MHz Deschutes to 66MHz FSB ( giving me a PII 233), works every time. Does not respond to multi configuration (DIP or soft).