• Please review our updated Terms and Rules here

SMP Pentium box with AGP? (Was: I have a plan...

Raven

Veteran Member
Joined
Mar 7, 2009
Messages
2,752
Location
DE, USA..
The PPro stuff is interesting. Perhaps it's the micro-op conversion then? I've used a Pentium III extensively under DOS and 98SE, and under 98SE it's fine, but under DOS things seems sluggish. I've done timing comparisons, as I said, and they're at least as fast, but something is different. Maybe I've looked at it the wrong way all along - perhaps it's not the CPU at all, but the video system?

I know about Zeno (designed to speed up video output on PPro+ iirc), but I don't think that fixed the effect I describe. I was convinced it was the CPU, bar none, for a long time, which is why I blamed the PPro and it's descendants - maybe I was wrong. Since I assumed this, I focused all of my research on the 486, Pentium, K5, K6, P2, and PPro architectures, to see what the difference was between the old and the new. The two things I found were that there was a focus on 32-bit optimization on the PPro and above, and the PPro had a RISC core with a translator to take normal x86 instructions and translate them into little micro operations which are then executed out of order, to increase efficiency. My theory was either that the lack of optimization caused it, or the micro-op translation caused it. The K5 and K6 were based on a CPU by NexGen that used the same RISC core concept as the PPro, so they wouldn't be any different, or so I theorized.

I ended up using a Pentium 3 for two reasons - one fell into my lap that week, and I also theorized that if the micro-op translation OR 16-bit optimization was the difference, perhaps just jacking the raw speed up as far as possible would negate the effects and at least feel as snappy as a 486 at 66Mhz. I set up my 1ghz P3 box, but no, it still felt sluggish.

In retrospect, however, after your (everybody who's talked about it's) input, I'm beginning to think that my base assumption was flawed/premature, and perhaps the video card/subsystem is to blame. Perhaps a PCI video card, while faster or more powerful, is not as snappy as one that was designed when DOS was dominant? I'll need to perform some new tests - I'll set up my midrange Pentium box with a PCI card, and see if it feels sluggish. Then I'll set up the newest box I can find that has ISA slots with an older ISA video card, and see what that feels like. Perhaps even a Pentium box would exhibit the same sluggishness to me, because I almost always use 486 boxes to play DOS games and use DOS, and use VLB in those cases. I've not read extensively about the workings of VLB, but perhaps it's closer ties to the CPU provide that snappiness I'm used to, regardless of card power/speed?

Thoughts?

Voodoo 5 5500

You lucky bastard! LOL. I'm sure you know how rare those are and how lucky you are to own one. o_O I envy you for that, heh (not that plenty of things on that list don't make me envy you, but the V55 above all others..).

(As an aside, I found a picture of FOUR 5500s in SLI! :eek: http://www.3dchip.de/Grafikkartenmodds/Grafixmodds/quantummercury.jpg Damn do I wish I owned that - I'd put it in a brand new box and it might even keep up well enough to be usable with modern games - 16 cores after all... o_O)
 
Last edited:

luckybob

Veteran Member
Joined
Feb 3, 2009
Messages
1,344
Location
denver
pentiums were limited to 2 processors and no agp. singles have agp.
pentium 2's & 3's are limited to 2 processors and have agp 2x
pentium 2 & 3 XEONS go up to 8 way processors but I've only seen agp on dualies.
 

Raven

Veteran Member
Joined
Mar 7, 2009
Messages
2,752
Location
DE, USA..
Ah you're right, but I thought the 5500 wasn't particularly common, either.

Depending on the result of my yet-to-be-done investigating, I may end up trying to make a max-speed dual P3 box.
 

Unknown_K

Veteran Member
Joined
Sep 11, 2003
Messages
8,930
Location
Ohio/USA
PCI offered Plug and Play and more bandwith then VLB. The VLB bus offered direct path to the CPU and much faster speed/bandwith then ISA but at a cost to idling the CPU when it lost control to the VLB when it could have been reading cache. ISA just all around sucked because everything ran on it and it had low bandwith and speed.

Now for gaming the bus mattered but the video chipset mattered much more. Look at the S3 line of chips, some rocked on VLB for DOS and others were slow as hell even on the PCI bus. Then you have stuff like the Tseng Et4000 that was faster if you had 2MB of RAM instead of 1MB (interleaving RAM). The ET6000 was realy fast in last generation DOS.

Even in DOS you had 320x200 games, VGA games, and SVGA games that used VESA modes. Some chips were fast for specific modes and sucked for others. In the Windows 3.1 era quite a few chipsets were optimized for Windows and not DOS so they might fly in Excel and CAD but socked for games.

I was heavily into gaming from the 286 era (late 1980's) to about 2000 and had all kinds of cool hardware and in the last decade purchased even more to experiment with. The DOS era is like the 3D era, there were major improvements along the way and hardware tended to shift every so many years. Some hardware was just better for specific games it was optimized for. Drivers and API were also a factor which kept companies like 3DFX ahead because they pushed GLIDE and ran those games the fastest while Nvidia was better under DirectX.

The 3DFX 5500 AGP wasn't that rare (I have mine stored in the original box which I saved), the Mac PCI version was rare and the 4500 (1 GPU instead of 2) is also rarer.
 

deadcrickets

Experienced Member
Joined
May 17, 2010
Messages
127
Location
pennington gap, va
The PPro stuff is interesting. Perhaps it's the micro-op conversion then? I've used a Pentium III extensively under DOS and 98SE, and under 98SE it's fine, but under DOS things seems sluggish. I've done timing comparisons, as I said, and they're at least as fast, but something is different. Maybe I've looked at it the wrong way all along - perhaps it's not the CPU at all, but the video system?

I know about Zeno (designed to speed up video output on PPro+ iirc), but I don't think that fixed the effect I describe. I was convinced it was the CPU, bar none, for a long time, which is why I blamed the PPro and it's descendants - maybe I was wrong. Since I assumed this, I focused all of my research on the 486, Pentium, K5, K6, P2, and PPro architectures, to see what the difference was between the old and the new. The two things I found were that there was a focus on 32-bit optimization on the PPro and above, and the PPro had a RISC core with a translator to take normal x86 instructions and translate them into little micro operations which are then executed out of order, to increase efficiency. My theory was either that the lack of optimization caused it, or the micro-op translation caused it. The K5 and K6 were based on a CPU by NexGen that used the same RISC core concept as the PPro, so they wouldn't be any different, or so I theorized.

I ended up using a Pentium 3 for two reasons - one fell into my lap that week, and I also theorized that if the micro-op translation OR 16-bit optimization was the difference, perhaps just jacking the raw speed up as far as possible would negate the effects and at least feel as snappy as a 486 at 66Mhz. I set up my 1ghz P3 box, but no, it still felt sluggish.

In retrospect, however, after your (everybody who's talked about it's) input, I'm beginning to think that my base assumption was flawed/premature, and perhaps the video card/subsystem is to blame. Perhaps a PCI video card, while faster or more powerful, is not as snappy as one that was designed when DOS was dominant? I'll need to perform some new tests - I'll set up my midrange Pentium box with a PCI card, and see if it feels sluggish. Then I'll set up the newest box I can find that has ISA slots with an older ISA video card, and see what that feels like. Perhaps even a Pentium box would exhibit the same sluggishness to me, because I almost always use 486 boxes to play DOS games and use DOS, and use VLB in those cases. I've not read extensively about the workings of VLB, but perhaps it's closer ties to the CPU provide that snappiness I'm used to, regardless of card power/speed?

Thoughts?



You lucky bastard! LOL. I'm sure you know how rare those are and how lucky you are to own one. o_O I envy you for that, heh (not that plenty of things on that list don't make me envy you, but the V55 above all others..).

(As an aside, I found a picture of FOUR 5500s in SLI! :eek: http://www.3dchip.de/Grafikkartenmodds/Grafixmodds/quantummercury.jpg Damn do I wish I owned that - I'd put it in a brand new box and it might even keep up well enough to be usable with modern games - 16 cores after all... o_O)

I forgot what year it was but I remember that many of the video cards in the Windows 98 era began dropping true DOS support. So what you would see is sluggish performance. Now we are beginning to lose the 2D blitting and other acceleration introduced around the Windows 3.1 days.
 

luckybob

Veteran Member
Joined
Feb 3, 2009
Messages
1,344
Location
denver
max speed p3?

If you dont care about agp... Look for an Asus TRL-DLS or an INTEL SDS2.

At one time, I wanted to get three pci voodoo 5's and make a triple monitor setup. I found that the mac versions of the voodoo 5, have dvi and will run @ 66mhz on the pci bus. However when I found that the cards are relatively scarce and run upwards of 150 each, that idea died.

The nice thing about those boards is that they run dual channel pc133 ram. The biggest bottleneck for the tualatin was memory bandwidth. When I had my SDS2 as a server I was always surprised at just how fast it was.

If you want AGP, the tyan S2507 comes to mind with agp 4x. but I havent used one personally.
 
Top