VOGONS


Bitchin' dual p-pro setup

Topic actions

Reply 20 of 84, by dirkmirk

User metadata
Rank Oldbie
Rank
Oldbie

Quake 1.06
Nosound
two screen sizes below maximum?
640X480

Intel PR440FX
P2 OD @350mhz (5X70mhz)
256meg of ram
ATI 9100 PCI

Only 23.3fps in real dos mode for me, am I doing something wrong or your machine that awesome? Surely the i960 can't make that difference and I'm running a faster cpu as well, pretty sure different video cards didn't make a difference when I tested in dos.

Reply 22 of 84, by dirkmirk

User metadata
Rank Oldbie
Rank
Oldbie

Guess I'm hard out of luck as the bios has no options for ram timing or anything else, looks like I'll be better off finding a board with more advanced settings, is their a way to know which boards would allow those settings? I thought overclocking the CPU would give the best performance the thought never occurred about bios settings which the PR440FX lacks.

Reply 24 of 84, by Arctic

User metadata
Rank Oldbie
Rank
Oldbie

The mother of all boards!
Congratulations! I wish I had that board too 😁
That would be my PPro setup

It's amazing that you got all the parts. Even a UP8!!
The UP5 is imho already cool enough! 🤣

Reply 25 of 84, by feipoa

User metadata
Rank l33t++
Rank
l33t++

dirkmirk, did you check the BIOS with MODBIN to see if there are hidden memory timings which you can unhide?

Plan your life wisely, you'll be dead before you know it.

Reply 26 of 84, by dirkmirk

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote:

dirkmirk, did you check the BIOS with MODBIN to see if there are hidden memory timings which you can unhide?

No I haven't, where can I find more information about "MODBIN"?

edit: watch this space I forgot to run VIDSPEED!

Reply 28 of 84, by mr_bigmouth_502

User metadata
Rank Oldbie
Rank
Oldbie
luckybob wrote:
30FPS @ 640x480 in dos quake? YES PLEASE! […]
Show full quote

30FPS @ 640x480 in dos quake? YES PLEASE!

lq9l8Pmm.jpg

Is that with FASTVID or without?

EDIT: Never mind, I found my answer. 😉 But what exactly does FASTVID do to influence DOS video performance on P6 systems?

Reply 29 of 84, by luckybob

User metadata
Rank l33t
Rank
l33t

Without. I might give fastvid a test, just for giggles, but I don't expect much. If memory serves it was designed for AGP not PCI.

In other news, I got pissed off at winblows NT. Long story short, the games I want to play are going to want a later version of direct X than I can provide with NT. There were many other annoyances that led to the decision to go to win2k.

I pulled the CF cards I was using to boot from, and installed 2k on a 36gb 10krpm scsi drive. So far everything is quick and responsive. I ran a few benchmarks from AIDA, for the hard drive. The scsi drives come close to saturating the scsi channel, as expected.
aEnZJokm.png

Also, thanks to win2k, I can use a sata card. I dropped in a drive from a DVR and it worked flawless. In fact the drive EASILY saturates the PCI bus.
3gftvSYm.png

More testing to come. I installed an IDE cd drive from lite-on to give my 40x scsi drive something to fear. I want to see any changes with scsi in this situation.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 30 of 84, by feipoa

User metadata
Rank l33t++
Rank
l33t++

For a dual PPro system, going to IDE looses some cool points, even if it is just a CD-ROM.

Many SATA host controller cards have NT4 drivers. Did you dry DirectX5 for NT4?

Back in the day when I was using a dual PII-400 (late '98, early '99), I still found NT4 to be quite a bit faster than W2K. I un-upgraded and went back to NT4 shortly after installing W2K. I kept NT4 going for few more years before giving in to W2K.

Part of the fun these days of using NT4 is getting it running well.

Plan your life wisely, you'll be dead before you know it.

Reply 31 of 84, by rgart

User metadata
Rank Oldbie
Rank
Oldbie

Is that DOS Quake using the onboard S3 Trio64?

Do you have plans to install a different video card?

Did you end up sticking with Windows NT 4.0 ?

=My Cyrix 5x86 systems : 120MHz vs 133MHz=. =My 486DX2-66MHz=

Reply 32 of 84, by luckybob

User metadata
Rank l33t
Rank
l33t
feipoa wrote:
For a dual PPro system, going to IDE looses some cool points, even if it is just a CD-ROM. […]
Show full quote

For a dual PPro system, going to IDE looses some cool points, even if it is just a CD-ROM.

Many SATA host controller cards have NT4 drivers. Did you dry DirectX5 for NT4?

Back in the day when I was using a dual PII-400 (late '98, early '99), I still found NT4 to be quite a bit faster than W2K. I un-upgraded and went back to NT4 shortly after installing W2K. I kept NT4 going for few more years before giving in to W2K.

Part of the fun these days of using NT4 is getting it running well.

Indeed. The IDE is just for testing at the moment. Also, I may have to use it. Turns out the IBM scsi cd drive I have does not play well with CD-R disks. I may just have to keep the IDE around, at least for the time being.

So far as I can tell. There is no difference in performance between 2K and NT4. I was just having too many issues with NT4 and drivers. Also, I HATE dealing with software, but I will faff about with hardware for hours on end. True story; I "upgraded" to windows 10 on my main machine months ago. I still have the default background and everything.

rgart wrote:

Is that DOS Quake using the onboard S3 Trio64?

Do you have plans to install a different video card?

Did you end up sticking with Windows NT 4.0 ?

No, the S3 caps at ~16fps. The matrox has other tricks that allow FPS to be what it is, that include bios tricks. look at the ultimate 686 benchmark thread for more info on this.

yes I do. I have several canidates. the matrox 200, a Hercules 3D prophet 4000xt pci (info: http://www.amoretro.de/2011/10/hercules-3d-pr … wervr-kyro.html ), and somewhere I have an ATI 9250 on pci. I'm leaning towards #3 to be honest. Come to think of it, I do have a voodoo banshee. If I decide to dual-boot with dos it might be a better option.

No I will not. 2K is going to be a better fit for me.

moving on. I've done a bit more benchmarking with different drives. The results are dissapointing, but I am not exactly surprised. I've come to realize that devices that cache data suffer a stereo type that they always improve performance. It turns out, this is not one of those. I get a performance boost by putting the I960 into dumb bridge mode. In fact the 70mb/s I was getting went up to just over 90mb/s. Considering I'm using a pci video card, this is the top of performance I will see. I see the same performace boost in writing speeds as well. I tested this with three drives. The before mentioned 10k scsi and 5400rpm sata. I dug through my parts and located a 15k U320 drive and adaptor. This drive completely saturated the scsi channel at 40mbs. total flatline. I might move to this drive for booting, as its access times are going to be as close to 0 as possible without ssd.

The issue is, the cache/buffer is for SLOW devices. The cpu can just send all the data to the I960 and go do something else. The I960 will do its job while the cpu is doing its thing. The drives I have are just TOO FAST. The cache is always empty. So it just becomes a performance hinderance. Now, if I was using period hard drives, the situation might change. I have some old 40 and 80mb drives, I might just test them out doday for giggles to see if I can do anything with them.

Also, I was hoping that the I960 would enable the onboard scsi chip to be raid enabled. This so far does NOT seem to be the case. I had hoped to use a pair of 15k scsi drives in an simple stripe array as boot, but I could find no option to do so in the bios (scsi bios included)

It is a mistake to think you can solve any major problems just with potatoes.

Reply 33 of 84, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Reminds me of a recent thread about VLB caching cards - they only seemed to have benefit with period correct (naturally slow) hardware.

Plan your life wisely, you'll be dead before you know it.

Reply 34 of 84, by luckybob

User metadata
Rank l33t
Rank
l33t

The only slow device I have left is a 10/100 nic. And that isn't going to benefit me here as 99% of the time I will be reading data, not writing. I suppose if the cpu is busy doing something else the cache might help, but with having 2cpus the likelihood is practically nil.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 36 of 84, by luckybob

User metadata
Rank l33t
Rank
l33t

I would venture to say the P65UP5 is uncommon. There is one on ebay now if you're willing to overpay for it. I did have a spare, and it did sell for $150 on ebay. This is the first P65UP8 board I have laid my eyes on since I started looking about 10 years ago. Hell, there wasn't even a good image of the board online until now.

In other news, I got out some old hard drives. Slowest one is 160mb and maxes out at 600kb/s! Buffered or not, made no discernable difference.

I also did some testing with cd drives. I have an old Lite-on 40x drive and these things were amazing. But the cpu usage on IDE is HORRID. I switch over to scsi, and it is a night and day difference. Not in speed, but cpu usage goes for 40% average to ~2%. I even tested a new dvd drive, same results. ( i own a few atapi to scsi converters ) Scsi is amazing in these older systems.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 37 of 84, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Where did the i960 come into play in these tests? I'm guessing it did not. Are your ATAPI to SCSI converters only for CD-ROMs? Which units are they?

Plan your life wisely, you'll be dead before you know it.

Reply 39 of 84, by luckybob

User metadata
Rank l33t
Rank
l33t
feipoa wrote:

Where did the i960 come into play in these tests? I'm guessing it did not. Are your ATAPI to SCSI converters only for CD-ROMs? Which units are they?

I tested first with the I960. reading and writing. I did the same tests in bridged mode and they were faster.

@mrau

As for testing on a database (I assume thats that you mean by db). I don't know how to do this. I have benchmark presets that claim things like web server and such. Would that be sufficient?

It is a mistake to think you can solve any major problems just with potatoes.