VOGONS


First post, by Moogle!

User metadata
Rank Member
Rank
Member

Just tried this today. Nothing fancy, just informal testing of the Phil's Dos benchmark programs on two different machines. One was a Pentium 66, Batman's Revenge with 256K cache and 32MB 60 ram and Phoenix Bios, the other a Micronics branded Intel HX with an Intel 233Mhz cpu, 64MB 60ns ram, 512K cache, also with a Phoenix BIOS. Both have their latest known BIOS. Both were booted from the same CF card with DOS 6.22, no drivers or memory managers, or anything. Just F5 to skip Confi.sys and Autoexec.bat. The card is an ET4000/W32p from an unknown brand that I largely bought on a whim.

The results were, for me, surprising.
Intel 233
1MB Dram
Quake 320x240 51.8
Quake 360x480 20.8
Doom max details 896 / 83.359
3dbench 1.0c 146.6
Chris' 3D benchmark 640x480 38
PC Player 640x480 21.2
2MB DRAM
Quake 320x240 51.9
Quake 360x480 21
Doom max details 880 / 84.875
3D bench 1.0c 164.4
Chris' 3D benchmark 38
PC Player 640x480 21.5

Pentium 66
1MB Dram
Quake 320x240 18.9
Quake 360x480 8.5
Doom max details 1849 / 40.394
3dbench 1.0c 68.4
Chris' 3D benchmark 640x480 16.7
PC Player 640x480 8.2
2MB DRAM
Quake 320x240 18.9
Quake 360x480 8.5
Doom max details 1837 /40.658
3D bench 1.0c 68.4
Chris' 3D benchmark 16.7
PC Player 640x480 8.2

The difference on the Pentium 66 system is basically nil. The effect of memory interleaving is a little more pronounced on the Pentium 233, but even there it isn't earth shattering. I wonder how many would have been using an Et4000 on a system like that. Possibly people who upgraded a bit at a time. It was nice to see something finally edge out the Number Nine Vision 864.

I have an ET4000 VLB coming in a couple weeks, I may try this again with my 486 system.

Reply 1 of 4, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Yes, it's been tested in the past. The answer is that the memory interleaving has no effect in DOS. However, it does make a noticeable difference in Windows.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 2 of 4, by kixs

User metadata
Rank l33t
Rank
l33t

Are you sure there is difference between 1 and 2MB with 3Dbench 1.0c on a P-233?

I actually never tested beyond 486 class computers, but in DOS there were no gains with 2MB. But it makes up to 80% increase in Windows performance.

Requests are also possible... /msg kixs

Reply 3 of 4, by Swiego

User metadata
Rank Member
Rank
Member

I have not compared these two but have compared the ET6000 PCI against a V330 PCI against a Millennium II PCI against a Riva128 PCI on a Pentium 90 and these cards all show almost identical performance in the various DOS benchmarks so I've assumed the CPU is limiting matters quite broadly among this class of HW. Given the identical performance when varying a first order attribute like chipset, I'd be surprised if memory plays a significant role?

Interestingly, on this P90, my Millennium G200 is about 30% slower across the board in DOS. I've been wondering what drives degraded DOS performance in newer cards.... a downgrade of silicon to save costs? Sloppiness / lack of attention to detail in design? BIOS/firmware overhead/bloat?

Reply 4 of 4, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

I can test 1 Mb version on my Athlon XP 3000+ test bench, so no potential CPU bottle necks. Already did that with 2 Mb for my ever growing testing thread of PCI video cards.

I must be some kind of standard: the anonymous gangbanger of the 21st century.