วีดีโอ

That's blue allright, but I wouldn't dare put it on a pentium 3.

It is good for a mini pc but it's a bit unfortunate that it only has one pci and one ISA. It should have at least another pci or another ISA.

The cpu was also running at 5ghz since windows 2000. I was kind of fearing the same thing. The fun thing about this upgrade is that many of the applications found in the early versions of windows were still functional in windows 7 32 bit. I was following quite a few applications since windows 3.1 but time didn't allow me to include them all in this clip. I'm still considering other clips with upgrade from 1 to 11. A kind of need to do this for an AMD cpu as well.

I think that is my favorite too, but I feel that the geforce 4 ti 4400 is soo underated. For now I think the geforce 4 is the winner according to the votes, but I have something really special for the geforce 3 ti 500. Still I will use the geforce 3 for some tests in this series.

This is actually windows xp used in this clip and I'm using msi afterburner. For win98 you can use fraps but you may need to look for an older version like 1.9 or 1.6

I never knew that, that explains why there is a 610 and 710 nvidia pci.

I think you can find some drivers on the wayback machine. I also found a working bios.

:))) the terminators will be back, I also have a p4@1400 socket 478 that will be part of this series...no doubt in the tests rdram performs faster than sdram but I think they only kept it up until ddr 333mhz.

Yes, you can resort to painting. The guy that originally sold me the blue build had the psu painted blue, I never knew that. A purple build would be awesome, I think ECS make a lot of purple motherboards and I also think that creative in their early days had purple graphics cards (Palit too). There's also a rather new WD purple. But I doubt that you are going to find any purple cpus or ram. Good luck with your build.

Hi Marco, that's pretty generous, and I may take you up for the offer in a couple of years. For now I got 2 blue voodoo 2s that need a lot of work. I also got a blue hard-drive and I'm trying to get some blue radiator sdram@150mhz. I'm also moving everything into a blue-er case so I have my hands full for now.

@@retrodrive-thru47 Well, if we're not in a rush, once I have time to extract them from the board, I could send them and you can deal with them at your own convenience, to make your system just a bit more blue.

Back then I had an AMD, nothing as exotic as an Intel. The P4 were extremely expensive, and I only used it for games on the integrated SIS graphics card (equivalent to a sis 305) and a gf4 mx440 a couple of years later. I also upgraded to a sk 939 much later and by 2006 I had an agp 6600gt to keep it company. Indeed ... that system was flying.

Thank you for the tip. I'm currently looking into getting 2 blue voodoo 2 cards and maybe a different sound card and I already got a new blue case. So in a future video I'm going to upgrade this build. I'm also considering replacing the cpu with a less powerfull p3s maybe a 1266mhz so I can free up the 1400 for other projects.

Yeah, I kind of feel 4:3 is the way to go with these clasic games too. Although i have been playing NFS Porsche with the wide screen option from the graphics menu since I can remember because I felt it gave me less screen to focus on and I can take in more information. I live in Romania so EU, but I would only get a CRT if I can pick it up in person. I also picked one last week , a Philips 109B5 in a decently good shape so I will investigate that one for the next couple of weeks. Thank you very much for the offer.

Yes, I have seen those too. If I wanted to be 100% period correct I should have used the x850 with that crossfire cable you mentioned, it was even described in the manual for the motherboard. Unfortunately those cards are hard to find so I had to use the x1950. I think the mechanism for merging frames is similar for both methods with external cable(you mentioned) and what we see in the clip. I think there's also a chip on the motherboard that controls the synchronization of the cards.

For some youtube reason I can't directly respond to your comment but here it goes. You're right, the cards were, even back in the day quite rare as very few people had reason to buy them because a crossfire setup was very expensive. $1100 for two cards and crossfire motherboards were even less available on the AMD front ironically ss Nvidia dominated the motherboard space with their chipsets.

So a SSD would be reccomended? I did use a raid0 for this system and I kind of felt that the raid was on the limit actually bottlenecking the hard drives themselves. I guess that using a Ssd would not greatly increase the transfer rate but it would reduce the access time to 0ms.

@@retrodrive-thru47 i would try an ssd for exactly that reason. better access times. raid might be a bit much. its just more interrupts and such, but i could be wrong ;) i think its a limitation of the computers of the day. i was just a kid when these systems were in common use. so take it with a grain of salt. i think the micro stuttering is the microsecond or two where the video card is loading more data into ram and has to issue an interrupt to the cpu, and the cpu then has to issue a couple to setup the data transfer into the video card. the system is in a lock state for a microsecond or two while all these pieces of hardware acknowledge interrupts and setting up dma and whatnot. this is me theorizing. i dont think anything can stop that micro stutter as its a part of the system and game programming. this is still one of the smoothest ive seen though. in the 90s that pause would have been where the cdrom or hdd were accessed to load that data. that specific action always led the system to a tiny little pause as all those interrupts , busses, and hardware did their thing.

You're welcome, glad you like it.

Thank you, I hope there's more to come.

Yes, that's what most of the people agree on. Personally I like them all good or bad, they are 20 years history at this point and I feel privileged to get them up and running one more time. (I had an AMD back then so I really can't complain).

Thank you for your reply. I always felt the opterons did overclock better than the Athlons on 939, unfortunately in this clip I don't do any overclock. As for the choice of motherboard I did collect many 939 Abits as well, I think I have several variations of the AN8. But when it came to 754 and 939 (and overclock) DFI were in a league of their own, you just have to follow Oskar Wu that left Abit at the end of the socket A period and joined DFI afterwards. I love both Abit and DFI, such a shame that they are no longer in business.

:)) hard to tell, I'll bet it isn't too much. The board itself was a bit of disappointment.

Thank you! I would recap mine as well. Can you tell me what polymer values replace what electrolytic values?

​@@retrodrive-thru47 They should have the same capacitance value or as close as you can. I have salvaged polymer capacitors from newer broken boards to repair retro ones and they work fine. Just have the same voltage or higher rating. Also you can buy cheap multi part testers online to check capacitance. Shelby on his channel Tech Tangents recapped an Abit BP6 dual socket 370 board with ploymer caps which was pretty awesome! He does some great other videos too!

@@retrodrive-thru47 Just checked and it does have the same cpu/memory layout as the Lan Party board! I must have been thinking of a different board.

​​​@@retrodrive-thru47 Just use same or similar capacitance values but same or higher voltage. I bought a pretty good cheap capcitance tester online to measure and check that the caps are within their rated value. It can also measure transitors, resistors and other components too! I have used salvaged polymer capacitors from broken newer boards and they work fine.

Unfortunately I don't really know the equivalence of electrolytic to polymer. I think the voltage needs to be the same or higher for the new capacitors. But I don't know about capacitance value. Cheers!

I almost spilled my tea. But you are right this was a disaster, there were some unfortunate souls that even bought this platform I know of at least one, still it had SDRAM and a VIA chipset. :))

@@retrodrive-thru47 Yeah SDRAM made this a whole lot more palatable. I think it just came around at a terrible time for Intel too, they were pretty blindsided by the success of the Athlon.

Sure.

It looked bulletproof back then (but it was also very expensive). I will investigate the SDRAM soon as I hope to get another motherboard.

a n i m e n i m e

I think I used 45.23 for this one.

Isn't a 9800Pro or any of the others a little bit overkill? I think I can find one that works but the P3 is out of their league and I would like to keep a minimum of realism in my benchmarks :)) I can definitely jump the horse with these graphics cards but in the end what would be the point matching parts that are so distant in time?

​@@retrodrive-thru47 early cards a pretty weak especially in games that use programmable pixel shaders like say doom3, or halo, which can obscure what those processors can do, I don't think it is unrealistic to use a 9800 pro/ geforce 5 or newer, people were still using p3 and early pentium 4 systems well into those eras.

@@retrodrive-thru47 9700PRO then. it's from 2002 so it's totally in its league. maybe a 9500pro or 9600 pro would already suffice though

The 7800gtx seems extreme for a P3 , I doubt i can get it working, maybe with an adapter, but even so....I can probably find a geforce 6 and try it with a p3 in a future video. Thanks.

​@@retrodrive-thru47 yeah finding a 7800 with the 3.3v notch is probably pretty hard, but you can find 6600s and 7600s with them fairly easily.

Usually msi afterburner(where there is a lot of info) or fraps. The system takes a performance hit, I've investigated in one of my first clips, the best retro OS... it's usually 2% for Fraps and 5-7-10% for Afterburner depending on the information displayed.

In order, Ignition, Revolt, Need for Speed Porsche unleashed, Stunt GP and RC Cars in the first part and Return to castle Wolfenstein and Medal Of Honor Allied Assault in the second.

Thank you, that's really insightful. I knew of some asus motherboards that don't require pairs of rdram but I thought that was an exception. As for Sse2 I didn't quite mean the games(this was a failure on my part for using a weaker graphics card), the encoding program had a checkbox to use sse2 on the P4 but true enough the performance wasn't that great compared to just SSE maybe a 5-10%. (I also would expect other programs like cinebench or superpi to take advantage of sse2). Do you know of the reason why encoding was better on the P4? Was it the longer pipeline, the better prediction branch? Cheers! P.S. As for the connector I suppose there is a problem on the motherboard as some capacitor was hanging on just one leg in that area where all PSU connectors were.

@@retrodrive-thru47 Why Tsunami MPEG is faster on P4 even without SSE2? I can't say without profiling on both systems. But i had worked on mpeg2enc about 20 years ago and the slowest part by far was the motion vector estimation, which is just... a lot of subtractions and comparisons. A LOT OF THEM since you're checking dozens of potential matches per block, for every potential match, counting all pixels. This was the area i back then was curious in improving by mashing a SUSAN corner detector into it, which... almost worked :D i wasn't very good :D Most of the encoder is in fact not at all pipeline sensitive or prediction sensitive, merely relying on raw arithmetic throughput - i think mostly MMX if implemented. I don't actually remember that much from the era.

You can get a 20-Pin to ATX 20-Pin & AUX 6-Pin Cables

I'm personally dealing with the same issue with PSUs. Good thing that there's an adapter for that. I didn't know it existed.

Yes reading about the differences I've come to the same conclusion, the shorter pipeline and bigger cache helps the P3 over the faster P4.

Nice!

I'll do some more comparisons. Now I've got a SDRAM sk 423 too and I can do a proper comparison.

Sure

May very well be. I has some competition from socket 8 and socket 754.

Terminators isn't quite the correct name. RDRAM is a bus that is terminated on the mainboard. To fill out unused memory sockets, you use CRIMMs which are continuity only, not termination. When CRIMM are missing, the termination is of course missing as well, you get a floating far end. There can only be one set of terminators, but any number of RIMMs or CRIMMs, so this setup is more flexible. Termination is resistive, not capacitive.

@@SianaGearz oh? Please explain further. I thought they were boring old decoupling caps on a card.

@@myne00 CRIMM card has no soldered-on components at all, it merely connects socket's input lines to socket's output lines. RDRAM bus is connected in series from one chip to another; and when a chip is missing in the chain, you still have to make those connections to pump the data through. The RDIMM has an only 16bit narrow bus, and the socket is long due to loop through outputs. In contrast SDRAM DIMM sticks (disregarding dual channel) are all mostly connected in parallel as opposed to in series, with just a couple unique address lines per chip or stick to differentiate them. The socket's data bus width is 64 bit.

Yes, it would be a good ideea to test them for resistance with a multimeter before powering on any board. I do have several 512mb pairs. I usually like to keep it a bit real and keep the memory somewhat close to what people could afford at the time. It was also a good ideea to present the terminators because I've learned some things about them from the comments. I never knew that there were 1gb modules. Cheers!

No.

I would think so too. Since they do the same function I would replace them all in that area, and around CPU too. I'm not really trained to identify which capacitors belong to which area that's why it would be difficult for me to identify those responsible for RAM.

@@retrodrive-thru47 really it's the larger ones around those primary system components that are doing the heavy power line filtering and supply to those components, I've fixed boards that weren't stable or didn't post before by swapping the caps around those areas

Almost sound like my path. In 2002 I had an XP 1600+, in 2004 I upgraded to a Barton core 2800+ and in 2006 to an X2 4000+ that also felt like a major upgrade for me too.

@@retrodrive-thru47 quite similar indeed. I stuck with AMD for the longest time, even through their poor performers however I never cared about having the fastest, I always aimed for price/performance. After the 939 x2 3800 I upgraded to an Asus M2N-E with Nvidia Chipset. Had an Athlon X2 5200 which lasted me a while, later maxed the board to a Phenom X4 9950 (Mediocre but still a solid upgrade over the athlon) Later I got an AM3 board and went Phenom II 955 and later upgraded to FX8350 who everyone knows was flawed... but it was just a CPU drop for a significant upgrade over the 955. AMD's first low end phenom and low end FX were indeed complete trash and buggy. Like Phenom1's without the 50 at the end (Phenom 9100, 9500, 9600) or the original bulldozer fx chips

i had that cpu for a while too but mine was running 2.4ghz. i literally only had to upgrade from it when team fortress 2 came out and only ran decently on core2 machines (i bought an athlon x4 before and it was trash, i kept it for like 3 months before going intel). back then the source engine was single threaded by default and there was a hacky way to enable multi core that doubled performance (mat_queue 2 on console) but it was unstable.

No.

@@CovenantAgentLazarus Your mom says "Yes"

@@phazonclash No.

Even the "gooder" P4s were still not great. I replaced a 2.8E with an Athlon XP 3000+ and it was so much nicer to use. My pet peeve is the denorm bug, where P4 (all revisions) raises an FPU exception on too broad a range of values which don't require re-normalisation, but it thinks they might, by an order of magnitude vs. any other processor. It is particularly nasty in the earliest ones but it's still pretty bad on all of them.

I ran with a 2500+ Barton @3200+ for years. It was great until I needed a 64 bit CPU with SSE2/SSE3 for Darwin/Mac OSX (early hackintosh).

For now, I do plan to revisit this to try the 77.77 driver as it looks that it has support for the 6800 and others.

Tualatin was the same as Coppermine clock for clock however. The main reason for existance at all was for Intel to trial the 130nm process. Also, Tualatin came out within 2001, the same as Willamette. Also, Intel recalled all Coppermine PIII 1133 as every single example was defective.

@@dabombinablemi6188 The PIII 1.4GHz you have there appears to be the S model. These had improved L2 cache doubling the capacity. The Tualatins all also had a more modern data prefetch management. So you do tend to see clock for clock differences than compared with the Coppermine. That PIII was released in January 2002 the day after the Northwood P4 2.2GHz. So the Northwood really was it's stablemate. The 1.4Ghz Tualatin Celeron is later still at May 2002. The Willamette P4 1.4GHz was first released back in 2000. Even these earliest Willamette P4s pre date the very first Tualatin Core PIII by about half a year. What you have done is taken the fastest PIII from 2002 and compared it to the slowest P4 from back in 2000. Your tests are still very interesting though. I wonder how that PIII would have compared with Willamette P4s at 1.9GHz and 1.8GHz as they would have probably been more price comparable to the PIII you had in their day. It would be interesting to see how the slowest 1.6GHz and fastest 2.2GHz Northwood released the same month as your PIII would compare too. It would give a good indication as to how well or bad these late PIIIs were actually doing back in their day. I Wish I had some in my collection now as you have got me interested.

I thought the idea was to compare three CPUs running on the same clock, to show that clock-by-clock, P4 was slower than previous generation. Which was quite unusual at the time, to say the least. At least P4 was able to run on much higher clock speeds later on.

Yes, it's geforce 3 ti territory. It would have been a better choice for these tests.

Communications and Networking Riser. It was supposed to be used with modems, sound cards, network cards. But, considering that more and more components started being integrated into the chipset even at the time, considering that the motherboard shown in the video has 5 PCI + 1 AGP slots, and considering that PCI peripherals dominated the market back then (there were even people still using ISA cards at the time), the CNR slot was more of a solution in search of a problem.

@@thomasmittelwerk410 Owned several pc compatibles since mid 90s but I've never heard of such thing...guess it was short lived and not much popular like rambus

@federicocatelli8785 many boards had it, few ever used it.

Some sort of cut down PCI. Others have named it. It could be found on many motherboards around 2000, Amd or Intel. I could only see soundcards and modems for it.

Glad you enjoy it!