Suspected inefficiency in several component specs

Showcase your Retro PC / Build logs.

Suspected inefficiency in several component specs

Postby Gahhhrrrlic » 2018-5-15 @ 15:16

In my current Shuttle HOT-557 based system I am noticing some inefficiency or slowness that appears to be abnormal for the hardware. To summarize the rig:

HOT-557 Mobo
Pentium 120 (@133)
32 MB SIMM memory at 70nS
ISA ALWAYS IN-2000 SCSI adapter
ISA SB16 w/ dreamblaster wavetable
PCI ATI 3D RAGE II 2MB Graphics
PCI NIC
ISA MODEM

The items of biggest concern are the: CPU, SCSI, VIDEO, which all underperform. RAM is a question mark.

So going through the list, the motherboard features all appear to work, board has a fresh battery (Thanks Jed) and all slots function. For some reason only 1 of the 2 USBs work but I consider myself lucky that any of them work and this is probably irrelevant.
MOBO OK

The CPU appears slow to me. Benchmarking it in DOS puts it more on part with an AM586 @133 and certainly lower than a P133. DOS games that use software rendering such as System Shock still lag at 320 resolution. Sure it was unplayable on my 386 and now it is playable but it should be perfectly smooth and this is not the case. It still stutters a bit when lots of stuff is in view and is unplayable at 640 resolution. Quake at 320 is bordering on unplayable and stutters heavily. Zipping and unzipping files takes forever too. I just feel like it's either damaged, set incorrectly or being bottlenecked by something else.
CPU SLOW

The RAM is only SIMM RAM at 70 nS so I'm hoping the problem lies here. The mobo supports PC66 EDO DIMMS so I am going to grab 64MB of that at 60nS and see if that helps. I will also lower the timings in the BIOS. One thing to note too is that there is 256kb system cache but the expansion slot is not populated so I don't know if another 256k will help with cpu speed.
RAM UNKNOWN EFFECT

The SCSI card is abysmal. I don't know why. I've determined it is limiting my drive to 2GB - Fine, I'll live with that for now - However the transfer speeds are only 1-2 MB/s depending on whether it's read or write and I feel it should be faster. Is this an ISA bus bandwidth limitation? The card is set to asynchronous mode for improved transfer speeds. Windows 95 has built-in drivers for this card though I'll be damned if I can find newer ones. Windows takes too long to boot in my opinion and while I don't think this is affecting my games, it is theoretically possible. I think even 1MB/sec should be ok for games of that era with very small texture sizes anyway.
SCSI VERY SLOW

The sound blaster card is fantastic, aside from the usual Creative BS like clicking and popping once in a while.
SOUND OK

The video card sucks - but I mean more than it should. There's this game that was bundled with the 3D Expression cards called wipeout, which was supposed to be ATI's game, their tech demo so to speak, and I can barely get it playable at 320 resolution. It looks nothing like the promo videos and I am not even sure it's hardware accelerated on my computer, though I did select D3D HAL. I can't get Quake with the GL driver to work at all, even though the card should support 1.0 most games either look terrible or don't load with DX enabled. I know it's not broken though because DXDIAG passes all tests and my VRMAN TidalForces screensaver works ok at 640 resolution which also uses D3D HAL.
VIDEO SLOW

NIC IRRELEVANT
MODEM IRRELEVANT

So does anyone have any theories about what could be the root-cause or the main contributor to one or all of these slowdowns? I'm just trying different drivers to see if I can fix it that way at the moment.
Gahhhrrrlic
Member
 
Posts: 257
Joined: 2017-12-05 @ 00:39

Re: Suspected inefficiency in several component specs

Postby lvader » 2018-5-15 @ 16:27

I don’t expect changing the ram will have a major effect, it will only be a small incremental upgrade. I’d upgrade rhe processor to a 200 or 233 mmx and get a decent video card.
lvader
Member
 
Posts: 202
Joined: 2015-11-04 @ 14:33

Re: Suspected inefficiency in several component specs

Postby Gahhhrrrlic » 2018-5-15 @ 17:12

To get a proportional increase in speed I would do just that but the thing is, everybody else on youtube with the same hardware is doing things with their computers that I can't do, suggesting that my system specs are not limiting but instead I'm just not getting nominal performance from it.
Gahhhrrrlic
Member
 
Posts: 257
Joined: 2017-12-05 @ 00:39

Re: Suspected inefficiency in several component specs

Postby cyclone3d » 2018-5-15 @ 17:32

You might want to see if anything is sharing IRQs.

Video or sound or hard drive controller sharing an IRQ with anything else can lead to slowdowns and other issues.

I would get rid of that ISA SCSI controller. Theoretical ISA shared bandwidth when running the ISA bus at 8Mhz is 16MB/s. The PCI bus theoretical shared bandwidth, when running at 33Mhz, is 133MB/s.

What do you have the ISA bus speed set to in the BIOS?
User avatar
cyclone3d
Oldbie
 
Posts: 1820
Joined: 2015-4-08 @ 06:06
Location: Huntsville, AL USA

Re: Suspected inefficiency in several component specs

Postby firage » 2018-5-15 @ 17:43

Wow, ISA I/O in a Pentium build. Get rid of that 386 level controller. The onboard IDE might be fast enough to live with, but a separate PCI card for either Ultra DMA or SCSI probably wouldn't hurt.

Is the CPU recognized correctly and does it run at 120 MHz? Faster memory might make some difference, but shouldn't be a night and day difference, if what you already have is properly configured.
User avatar
firage
Oldbie
 
Posts: 1050
Joined: 2013-1-06 @ 21:43
Location: Finland

Re: Suspected inefficiency in several component specs

Postby cyclone3d » 2018-5-15 @ 18:44

A Promise Sata I - 150 based PCI SATA controller would work great as well. SSD on an old system is super sweet.
User avatar
cyclone3d
Oldbie
 
Posts: 1820
Joined: 2015-4-08 @ 06:06
Location: Huntsville, AL USA

Re: Suspected inefficiency in several component specs

Postby Gahhhrrrlic » 2018-5-15 @ 18:49

IRQ Sharing: I was under the belief that only PCI devices could share IRQs? I believe there is some sharing going on with the PCI devices (NIC/USB/2nd IDE) but the ISA SCSI is set to 10 and the SB16 is on 5 so unless one of the PCI devices tries to use 10... I will check when I get home though. Also, video cards always say N/A for IRQ. I assumed this meant they didn't require one.

ISA Speed: I don't recall my BIOS even having the option to set the CLK/X if that's what you're referring to. Again, I will check.

SCSI: I should probably just swap this card with my 386, which is an Adaptec and that one gets 5 MB/s. I mean, that computer doesn't need more than 2GB anyway and 1MB/s probably won't make a difference with anything I am running on it.

CPU: It's definitely running at 133 (not 120, I overclocked it a long time ago). Everything I check reports back that speed. I never had any sort of blue screen or crashing that was directly implicated to the CPU. It can run for hours without a hitch so I doubt it's actually faulty. If it is running slow, it's not the MHz value itself but something loading it down perhaps.

Here's a youtube video of someone playing quake on 320 with a 586 133MHz AMD.

https://youtu.be/KbcFvUOGA44?t=1m11s

This is almost exactly what I am getting with my machine, but my CPU should be faster at the same speed. I think we can rule out hard drive here because once a scene has loaded its textures into memory, it shouldn't lag from that anymore. At 320, 2MB should be enough of a frame buffer for the resolution so I think the game is CPU limited in my case - or RAM. On my 386, I did CPU tests that varied by 5-10% depending on my RAM timing.
Gahhhrrrlic
Member
 
Posts: 257
Joined: 2017-12-05 @ 00:39

Re: Suspected inefficiency in several component specs

Postby mrau » 2018-5-15 @ 21:30

probably a little basic but - remove nic and modem and reset to defaults bios and try, then reset to optimized defaults and retry?
mrau
Oldbie
 
Posts: 950
Joined: 2015-11-28 @ 12:43

Re: Suspected inefficiency in several component specs

Postby gdjacobs » 2018-5-16 @ 01:59

Aside from IRQ handling as noted above, are there any RAM timing parameters that can be tuned in BIOS?
User avatar
gdjacobs
l33t++
 
Posts: 5339
Joined: 2015-11-03 @ 05:51
Location: The Great White North

Re: Suspected inefficiency in several component specs

Postby Gahhhrrrlic » 2018-5-16 @ 03:52

Yes, all the RAM timing can be adjusted and I have tried. While it didn't crash, it seemed a bad idea as there was no performance gain and some weird stuff started to happen sporadically so I switched it back. I am happy to buy new ram though to be able to lower these timings. I am currently looking at 32MB EDO 60ns 16-chip 4x4 4K nonECC 168-pin DIMMs (2 of them) for this machine so that should remove any and all memory related bottlenecks.

BTW, I determined the ISA speed was CLK/4. I don't understand how 133/4 (which is 33) gets down to 8MHz... somebody will have to explain the math to me. Anyway I changed it to CLK/3 and bad things happened - namely my sound card stopped working and it took several reboots to (kinda) get it back. My joystick port still doesn't work at the moment but that might be a registry issue. Close call. Won't do that again.
Gahhhrrrlic
Member
 
Posts: 257
Joined: 2017-12-05 @ 00:39

Re: Suspected inefficiency in several component specs

Postby cyclone3d » 2018-5-16 @ 06:34

Gahhhrrrlic wrote:Yes, all the RAM timing can be adjusted and I have tried. While it didn't crash, it seemed a bad idea as there was no performance gain and some weird stuff started to happen sporadically so I switched it back. I am happy to buy new ram though to be able to lower these timings. I am currently looking at 32MB EDO 60ns 16-chip 4x4 4K nonECC 168-pin DIMMs (2 of them) for this machine so that should remove any and all memory related bottlenecks.

BTW, I determined the ISA speed was CLK/4. I don't understand how 133/4 (which is 33) gets down to 8MHz... somebody will have to explain the math to me. Anyway I changed it to CLK/3 and bad things happened - namely my sound card stopped working and it took several reboots to (kinda) get it back. My joystick port still doesn't work at the moment but that might be a registry issue. Close call. Won't do that again.


33/4 = 8.25
User avatar
cyclone3d
Oldbie
 
Posts: 1820
Joined: 2015-4-08 @ 06:06
Location: Huntsville, AL USA

Re: Suspected inefficiency in several component specs

Postby Gahhhrrrlic » 2018-5-16 @ 14:27

Oh, I thought CLK meant CPU clock (133). So it must be either FSB clock or PCI clock or something right?
Gahhhrrrlic
Member
 
Posts: 257
Joined: 2017-12-05 @ 00:39

Re: Suspected inefficiency in several component specs

Postby gdjacobs » 2018-5-18 @ 01:12

FSB.
User avatar
gdjacobs
l33t++
 
Posts: 5339
Joined: 2015-11-03 @ 05:51
Location: The Great White North

Re: Suspected inefficiency in several component specs

Postby PTherapist » 2018-5-19 @ 10:49

Gahhhrrrlic wrote: One thing to note too is that there is 256kb system cache but the expansion slot is not populated so I don't know if another 256k will help with cpu speed.




You've probably already tested this, but just in case - have you ran something like CACHECHK in DOS to see if the 256K is actually present?

I don't know that particular motherboard, but generally I've found that most boards of that era with a COAST slot often don't have real onboard cache and some even have the BIOS modified to pretend it exists when it doesn't.
PTherapist
Member
 
Posts: 220
Joined: 2017-3-24 @ 02:00

Re: Suspected inefficiency in several component specs

Postby Gahhhrrrlic » 2018-5-24 @ 15:37

Dear God! Why would anyone do that?
Gahhhrrrlic
Member
 
Posts: 257
Joined: 2017-12-05 @ 00:39

Re: Suspected inefficiency in several component specs

Postby lvader » 2018-5-24 @ 15:45

I have one of those! The funny thing is that it’s faster than many motherboards with real cache. You really wouldn’t know from performance tests that the cache is missing.
lvader
Member
 
Posts: 202
Joined: 2015-11-04 @ 14:33

Re: Suspected inefficiency in several component specs

Postby Gahhhrrrlic » 2018-5-24 @ 16:37

lvader wrote:I have one of those! The funny thing is that it’s faster than many motherboards with real cache. You really wouldn’t know from performance tests that the cache is missing.


All the more reason to add more I figure. Whether the first half of the cache is there or not, I can only see a benefit in adding another 256kb.

I used to run Unreal Tournament on this machine but it was choppy. With fast RAM and more cache, I'm hoping to see an improvement. V2 graphics helps too.
Gahhhrrrlic
Member
 
Posts: 257
Joined: 2017-12-05 @ 00:39


Return to System Specs

Who is online

Users browsing this forum: No registered users and 2 guests