First post, by sliderider
- Rank
- l33t++
This is not exactly news. Also I hope that he tested with similar video cards, or else its going to drag to one or the other. For instance the only way I found to compare my VLB system with PCI systems was to get the PCI version of my Number Nine S3 Vision 864 VLB card. Back in the days most cards came in VLB, PCI and ISA releases.
Look at my two systems for instance.
My Pentium Overdrive 83 VLB system is performing the same as a Cyrix 133mhz and my Socket 4 system equals to an AMD Am586 133mhz on Phil's VGA benchmark: https://docs.google.com/spreadsheet/ccc?key=0 … p=sharing#gid=0
I think I was using a 2MB ET4000 W32P VLB against a Trio 64 PCI.
I have Diamond branded Trio 64 in VLB and PCI but the difference in score between the VLB cards is so marginal (or nonexistent) to the point that I thought it wasn't worth swapping the cards over. Unfortunately I don't have a reliable way of pitting the VLB against the PCI card but I considered it to be more a test of the platform in general.
My POD system is the MB-4DUVC one on that chart, though it doesn't usually use the POD and doesn't do so on that chart so it's not relevant to much here, however I was planning on adding my PPro before the chart is locked, so I can add the POD and add my P66 if you wish.
i don't have them but i believe that it depends on which benchmark he run.
for "pure" fpu and alu benchmarks that don't use much memory bandwidth the pod83 could have a chance, but when it comes to memory intensive tests(superpi for example) you should keep in mind that pod83 has a quarter of pentium66's fsb bandwidth.
wrote:This is not exactly news. Also I hope that he tested with similar video cards, or else its going to drag to one or the other. For instance the only way I found to compare my VLB system with PCI systems was to get the PCI version of my Number Nine S3 Vision 864 VLB card. Back in the days most cards came in VLB, PCI and ISA releases.
Look at my two systems for instance.
My Pentium Overdrive 83 VLB system is performing the same as a Cyrix 133mhz and my Socket 4 system equals to an AMD Am586 133mhz on Phil's VGA benchmark: https://docs.google.com/spreadsheet/ccc?key=0 … p=sharing#gid=0
It does surprise, though, because you expect a Pentium running on a motherboard designed for it and with a higher fsb to be faster than one mounted on a 486 motherboard with a slower bus speed even if the core is running slightly slower. The fsb running 2x faster and with faster memory installed should make up for it but doesn't somehow. The socket 4 chipset must have inefficiencies built into the design that are pulling it down. In order to beat the POD on the 486 motherboard outright in all tests, the reviewer mentioned that he had to use a socket 5/7 motherboard with a later generation Pentium installed which to me says that the POD should be an outstanding CPU for use in a retro rig if you have to jump ahead 2 motherboard and socket generations to beat it in all testing. I'm surprised more people didn't buy PODs to extend the lives of their 486 motherboards back then if performance could be better than the first generation Pentiums because it would still be cheaper than the cost of upgrading everything else in your system. Intel should have been selling them like hotcakes.
wrote:i don't have them but i believe that it depends on which benchmark he run.
for "pure" fpu and alu benchmarks that don't use much memory bandwidth the pod83 could have a chance, but when it comes to memory intensive tests(superpi for example) you should keep in mind that pod83 has a quarter of pentium66's fsb bandwidth.
He was running Doom and Quake among the other benchmarks he was using and for the time, those were all you really needed to know whether your system could handle most contemporary games or not.
I wonder what chipsets are used in the POD system and in the Socket4 system. From posts I have seen there are some chipsets for Socket4 that REALLY suck. I know Anonymous Coward has a post or two that reflect as such. (check his Turbocharged EISA Pentium thread)
Main pc: Asus ROG 17. R9 5900HX, RTX 3070m, 16gb ddr4 3200, 1tb NVME.
Retro PC: Soyo P4S Dragon, 3gb ddr 266, 120gb Maxtor, Geforce Fx 5950 Ultra, SB Live! 5.1
wrote:I think I was using a 2MB ET4000 W32P VLB against a Trio 64 PCI.
I have Diamond branded Trio 64 in VLB and PCI but the difference in score between the VLB cards is so marginal (or nonexistent) to the point that I thought it wasn't worth swapping the cards over. Unfortunately I don't have a reliable way of pitting the VLB against the PCI card but I considered it to be more a test of the platform in general.
My POD system is the MB-4DUVC one on that chart, though it doesn't usually use the POD and doesn't do so on that chart so it's not relevant to much here, however I was planning on adding my PPro before the chart is locked, so I can add the POD and add my P66 if you wish.
First off, I liked the video 😀
That is a pretty good match on the video cards, the ET4000 W32P performs quite equal to the S3 Vision series on PCI and the S3 Trio is just a tad quicker.
The chart is locked? I don't think Phil have any plans of locking it, or have I missed something?
It's an interesting video, but I'm not sure if it's a fair comparison, because isn't VLB technically faster than PCI? If I really wanted to compare the two, I would have used all the same parts, aside from the motherboard and CPU of course. I also would have used an ISA video card, since it would be compatible with both systems, and the slower speed of the ISA bus would make CPU-bound performance gains more noticeable.
wrote:It does surprise, though, because you expect a Pentium running on a motherboard designed for it and with a higher fsb to be faster than one mounted on a 486 motherboard with a slower bus speed even if the core is running slightly slower. The fsb running 2x faster and with faster memory installed should make up for it but doesn't somehow. The socket 4 chipset must have inefficiencies built into the design that are pulling it down. In order to beat the POD on the 486 motherboard outright in all tests, the reviewer mentioned that he had to use a socket 5/7 motherboard with a later generation Pentium installed which to me says that the POD should be an outstanding CPU for use in a retro rig if you have to jump ahead 2 motherboard and socket generations to beat it in all testing. I'm surprised more people didn't buy PODs to extend the lives of their 486 motherboards back then if performance could be better than the first generation Pentiums because it would still be cheaper than the cost of upgrading everything else in your system. Intel should have been selling them like hotcakes.
I think it really depends on which chipsets and motherboards that are being used on both the P66 and the POD. The early Pentium chipsets were pretty much crap compared to what came later as they were in many regards just 486 chipsets that were modified to run with the Pentium. Anonymous Coward's Socket 4 EISA system and this ISA/VLB system both use the Opti chipset and from his info they are pretty slow. Gona also have an early Opti chipset and he also reports quite bad memory performance. I've tried to show this in my Socket 5 & 7 benchmark thread. Here you'll see quite a difference from the early Socket 5 chipsets compared to the later ones. Many of the boards tested use chipsets also available on Socket 4. SiS 50X, Ali 1451, Opti Viper are all Socket 4 and 5 chipsets. Unfortunately no other Opti chipsets available in the bench. Seeing the few other benches that exist I*m inclined to conlude that the Intel 430LX Mercury chipset is the best there is for Socket4.
What I'm trying to say is that when running the POD in later 486 chipsets like the SiS 471 or SiS497 the hardware is more mature and you get quite good performance out of it, even though with the limitations. The POD was actually quite a nice upgrade IF you had a system that could take advantage of it. The cache settings and video card also makes quite a difference. If you had a W32P or another quick video card, then you'd get more out of your POD than with an ISA graphics card.
It's an interesting video, but I'm not sure if it's a fair comparison, because isn't VLB technically faster than PCI? If I really wanted to compare the two, I would have used all the same parts, aside from the motherboard and CPU of course. I also would have used an ISA video card, since it would be compatible with both systems, and the slower speed of the ISA bus would make CPU-bound performance gains more noticeable.
PCI and VLB are pretty much equal on the earlier boards/cards if you used the same card/chipset. I've done quite amount of research into this to compare the 3D Blaster VLB with the Diamond NV1 in NASCAR Racing and Battle Arena Toshinden. For more info see this thread.
To clarify a few things;
At 33MHz, which I used in that video, VLB is alleged to have the same throughput as PCI in ideal circumstances. I could have used ISA video but I suspect the score would drop similarly on both machines.
The chipsets in use are UMC 8498F (POD) and Intel Mercury 430LX (P66). The cache on both is 256K and cannot be changed on the Pentium due to it being a Batman's Revenge board; the cache is hardwired to it.
Unfortunately I can't test SuperPI. The POD isn't running Win9x (It's running Chicago 189) and the P66 simply won't as far as I remember.
Having used both systems for extended periods of time, I can report that the POD does generally fare better here, the performance generally seems more consistent.
I think that's everything I meant to address... I've just woken up. I'm glad people liked the video. I'm not at home at the moment so I'm currently incapable of running further tests.
wrote:Unfortunately I can't test SuperPI. The POD isn't running Win9x (It's running Chicago 189) and the P66 simply won't as far as I remember.
SuperPI run on 3.x with Win32s so I would be surprised if it didn't work.
And what do you mean a P66 "won't"?
wrote:Intel should have been selling them like hotcakes.
Agreed. Perhaps this is why there are so many of these 83mhz versions on eBay. I don't see the 63Mhz version quite as often, however. I know these overdrive processors were pretty expensive back in the day, and I think Intel really wanted people to upgrade to new architecture instead of prolonging the lives of their Socket 3 systems.
My Retro B:\ytes YouTube Channel & Retro Collection
Would the 32K L1 cache of the POD be the major factor in making it faster than a P66? P60 and P66 only have 16K of L1 cache. I have payed alot of DOOM on both systems and found that the POD 83 is faster, even the POD 63 feels faster than Socket 4.
wrote:Would the 32K L1 cache of the POD be the major factor in making it faster than a P66? P60 and P66 only have 16K of L1 cache. I have payed alot of DOOM on both systems and found that the POD 83 is faster, even the POD 63 feels faster than Socket 4.
If L1 cache is a major factor, the change should be just as much as moving from Pentium 166 to Pentium 166MMX.
Concerning POD, I think these CPUs were, as most other iOD CPUs, characterised by three words: overadvertised, expensive, late. Check out this: http://redhill.net.au/c/c-6.html
wrote:These came out so late and at such a high price as to be effectively useless as anything but a marketing tool. But in their proper role, as a marketing tool, they were one of the all-time classics: pure, one hundred percent vaporware from start to finish, neatly garnished with a sprig of FUD.
So no, they never sold well and were never supplied well either. Besides, they are much more mature than Socket 4 pentiums.