Chadti99 wrote on 2022-03-28, 01:01:

maxtherabbit wrote on 2022-03-28, 00:56:

pshipkov wrote on 2022-03-27, 23:50:

@maxtherabbit
Yeah, prices of everything are up and up.
But retro perf pushing takes what it takes.

Yeah I'm all for pushing stuff to the limit. I'm just not in favor of going ham on a part with a chisel removing shit that doesn't need to be removed - all under completely false pretenses.

I’ve re-read that thread several times and the guy is saying it was working without any of the components. How, I don’t know. I was talking to him a bit last year through PM but he hasn’t been on the forum for awhile. It could of been completely made up but wanted to give it a shot. Apologies for the horrors bestowed upon you today. Look away in the future 🤣.

The way I see it, I saved a part from scrap, and now it’s useable again. Cosmetically was not in great shape to begin with, already missing the heat sink/fan. Took me an hour to bend all the pins “straight” again. The ceramic isn’t likely scratched, it’s just the metal of my tools has rubbed off on it, which can be cleaned/removed.

I'm just griefing you about trusting that guy at all 🤣 If you think about it his claim is obviously ridiculous

But in the end no real harm done. You can replace the regulator with the diode mod if you wish, or just leave it jumpered and use the VRM present on most socket 3 mobos

That 1x50 MHz tricked too when i tried the first time the delided CPU.
For a moment i was like - okay - that was easy. : )

Do you at least get consistent POST screen ?
We entered a new space where POD125 is a story.
Magical CPUs need ... (still good chance yours is one of them).

@maxtherabbit
I think everybody here is respectful to the old junk.
Chad was not consciously going rampage on the CPU.
Thanks for pointing-out anyway. We know your intentions were all good.

It appears to be consistent with POST screen yes, unable to enter bios setup though. I get the blue background but no text.

And the CPU is frozen ?

pshipkov wrote on 2022-03-28, 15:01:

And the CPU is frozen ?

Not yet, just air cooled. Gotta find some time.

Don’t get too excited but I did manage to pull this at 7v to the peltier. Admittedly the surfaces aren’t mated up all that well so there is room for improvement. I can’t run 12v because of this as it dumps heat back into the CPU. Had to set cache to 3-2-3. Cache settings of 2-1-2 wouldn’t start DOS. Took about 20 attempts. Apologies for catching the pic mid monitor refresh. Would of been really cool to of had this as an option on socket 3.

92.28 CPU Score!

3-1-2 Cache settings

Removed the epoxy, re-applied thermal paste. Applied 12v to peltier. Only one side of CPU was freezing however. Would more consistently get to and launch dosbench. Most of the benchmarks would crash before starting. Was able to pull off #2 at 312 cache timing.

great stuff.
the fact that you can boot into dos is already impressive.
so far i cannot complete post no matter what. and even that is once in many attempts.

economics are against us for finding healthy chips. p24t goes for $100 to $160 on ebay these days, which is verry different than the $24 for am5x86.
😖

pshipkov wrote on 2022-03-29, 00:11:

great stuff. the fact that you can boot into dos is already impressive. so far i cannot complete post no matter what. and even t […]
Show full quote

great stuff.
the fact that you can boot into dos is already impressive.
so far i cannot complete post no matter what. and even that is once in many attempts.

economics are against us for finding healthy chips. p24t goes for $100 to $160 on ebay these days, which is verry different than the $24 for am5x86.
😖

Yah pretty cool to see for sure. What motherboard are you testing with? Have you tried with lower cache timings? This is with the stock cooler removed and the peltier right on the ceramic, with thermal paste of course?

Yah the odds of finding one of these guys stable even with a peltier is slim to none.

More benchmarks coming up!

Alright here we go, the paste must of settled in with some time, and cache timings of 313 helped. I was able to get 12v to the peltier before this run but it didn’t help much and my wiring was getting way too hot. Dropped it back to 7v and still got a good layer of ice around most of the CPU.

POD83@125 (50fsb) 5volts
1:1 CPU:PCI
256k cache 313 timing
16MB EDO Fastest timings
Lucky Star LS486e Rev C2
Voodoo3 2000
Peltier at 7v
pshipkov’s dos 6.22 image using Dosbench, tests 2,3,4,5,6

POD83@125 dosbench
Test 2 98.5
Test 3 80.3
Test 4 25
Test 5 28.8
Test 6 12.1

Unfortunately no Doom or Quake benchmarks would run. I’ll get a fresh comparison with the same gear at 100MHz soon. The last notes I have seem a bit low.

POD83@100 dosbench
Test 2 86.8
Test 3 68.8
Test 4 20.4
Test 5 25.2
Test 6 10.2

wow

Very very cool. 😀

I tried with Biostar UUD, Asus VLI, LuckyStar D.
With most conservative BIOS settings and both L1/L2 cache disabled.
In fact L2 cache chips were also removed to make sure there is no interference.
Ark1000VL, MGA and V3 VGAs.
It is the CPUs that don't hold here.

I am going to match exactly how you have the CPU setup and see if that makes a difference. I have the VRM bypassed, but you have one of the other components removed. Don't know if that makes a difference.

So i am looking at the metrics captured in [url=Re: 3 (+3 more) retro battle stations]this[/rul] thread and with LuckyStar D + POD100 + all BIOS settings on max, the PC Player Benchmark numbers are 25.2 fps and 10.2 fps. So, even with inflated L2 cache timings there is ~12% increase in perf.
But then looking at Asus VLI + POD100 -> 29.3 fps and 11.6 fps ! : )
Maybe you should try POD125 on VLI ?

I was wondering if that fourth SMD should be there or not, it was missing on this particular CPU.

I’ll def give it a shot in the VLI!

Those are all SMD bypass caps IIRC. If anything removing one would make it less stable not more

A slight deviation from the topic at hand... Do either of you have NT4 drivers for the Promise EIDE2300? I only see drivers for NT 3.5 and they return some error in NT4.

Haven't seen NT4 driver for EIDE2300.
NT 3.5 driver is the latest i have.
Sorry.

Btw, how is that "under fertilized flower" of yours coming along ?

Thanks for asking. My BL3-75 VLB system is coming along quite slowly. I am setting it up such that one 8 GB CF card has Win95c + NT4, while the other 8 GB CF card has DOS 6.22 + WIN 3.11. I am still working on the w95/nt4 CF card and haven't started the DOS/w311 card. I've run into some snags, first with a Magic S23 (not S23A) sound card, which has a Crystal CS4232 and Yamaha OPL3 chip. The drivers are a nightmare. I have to use a hodgepodge combination of different driver versions to get it working in both DOS 7 and Win95. The issues were due to the MPU-401 interface, which I suspect was default disabled. Today I hope to get it working in NT4, which doesn't really matter if midi is working properly.

I have the system setup with a Trio64 (not the 64V+), IBM 10/100 NIC, Magic S23, Promise EIDE2300, and native PS/2 mouse adapter. I am stuck on the issue of no Promise EIDE2300 VLB drivers for NT4 right now. The default NT4 PCI dual channel IDE driver seems to work, but surely wouldn't be as optimal as a Promise IDE driver. The issue with the NT 3.5 driver is that some error preventing installation on NT4,

1An error has occurred.
2External library procedure SetupChangeServiceStart reported the following error: Bad or invalid parameters passed to open service call.

After rebooting and looking in he SCSI adapters tab in Control Panel, I do see the PTIVGAPI.SYS driver from promise listed in SCSI devices, but the driver tab is empty, implying the driver won't load due to the above error. Curiously, if I delete NT4's dual channel PCI IDE adapter/driver from SCSI Adapters, the system will still boot even though the driver field is empty. Odd behaviour.

Another possibility is to run NT3.5 instead of NT4. Perhaps I could clone my NexGen's disk, which runs w95c/NT3.5, but I don't really want to put too much effort into this since the DOS/Win311 CF card will likely be the default OS used.

Right now, I want to bench the Promise EIDE2300 in Win95c and compare it with the results in NT4, both with the NT4 PCI IDE controller driver and the half installed Promise PTIVGAPI.SYS driver installed to see if performance loss is considerable.

Once the CF cards are setup, I hope to experiment with other BL3 modules in the hope I can get a stable 100 MHz or 80 MHz configuration.

Were you able to get the Windows 95 driver for the Promise EIDE2300 Plus working properly? Once I load the driver, reboot, and try to access Windows Explorer, or open a benchmark utility, the system hangs. The Windows default IDE controller didn't have this issue.

Also, I am no longer able to load the DOS driver. I can load the driver, but once the system has booted to the C: prompt, pressing the keyboard keys returns odd results. Like if I hit 'P', I might get an 'x', or a space, or nothing at all. This didn't happen before during my tests. Changes made since then were adding the Acer Magic S23 sound card, adding the IBM 10/100 ISA, and using the PS/2-KBC adapter. I'm not sure which of these is the culprit, but since the DOS driver didn't offer a substantial performance boost (like 4000 KB/s vs. 4400 KB/s), I am OK to let it go.

What parameters do you use in config.sys for EIDE2300.SYS, e.g. /D0:8, or /M0:8 ? They benched the same in Coretest on my system.

Interestingly, the half-installed Promise PTIVGAPI.SYS file for NT4 works and offered some modest gains in HDD performance, like 2.99 MB/s up to 3.38 MB/s.

One of the reasons I switched to the DTC 2278EB was that I got so many problems with the promise drivers.
For me even win3.x drivers hang the system and I couldn't start windows. I have a feeling though that the stock
windows NT & 95 drivers from MS would be fine. I guess you need to bench the system and compare values
from different OS to see if it's setup properly.
Also, I'm not sure you should load the DOS driver with Win95