VOGONS


First post, by Almoststew1990

User metadata
Rank Oldbie
Rank
Oldbie

I've had a spare few days and, like I usually do, I found myself doing some dumb project to keep myself occupied.

I recently picked up a dual CPU Socket 370 motherboard with two 750MHz CPUs and a Willamette 1.6GHz Pentium 4 system. I thought I'd compare the two, to see which was faster in early XP gaming and general usage (or rather, prove that the P4 was faster).

The systems
In the blue corner we have an ABit V9P
2x 750MHz @ 802MHz 107MHz FSB Coppermine CPUs
768MB SDR -100MHz RAM at 107MHz
GeForce FX5900 XT (because I got this card at the same time and wanted to try it out)
WD 160GB IDE HDD
SB Live!

In the red corner we have an MSI 650 Ultra
1x 1.6GHz Wilamette CPU
768MHz DDR 266 (@ 200MHz)
The same GPU, HDD and SB Live

Usability
The Socket 370 PC was much quicker installing XP. In the OS the system certainly wasn't fast but it was 'responsive' feeling; it's hard to explain. Kind of like a modernish netbook which has slow core speed but more than one thread.
The Pentium 4 idled at 50c (with a "system temperature" of 33c) with a bigger-than-stock intel heatsink and fan. Meandering around Windows was just faster but 'laggier' than the 370 system and one of the least pleasant Windows XP experiences I've had!

The biggest issue for the S370 system was a lack of SSE2. Spotify for XP wouldn't install and Opera for XP wouldn't install. I went online using K-meleon which was bareable for the simpler modern websites.

The Socket 370 system was rock-solid stable. The CPUs (or rather, their temperature sensors located under the CPUs) were kept under 30c, and wasn't even that loud due to the large heatsinks and 80mm fans being used.
Both systems only had two USB slots; the 370 had one header (which didn't work) and the P4 system had two headers. I used a PCI USB2 card for both.

I timed unzipping a 300Mb ISO in 7z to see which would be quicker (mins.seconds):
Copperminers: 1.47
Willamette: 1.53

This was quite a nice surprise (obviously I want the Coppermines to be surprisingly good 😉)

Gaming
I wanted to try out a bunch of early to mid 2000s games which had a minimum CPU requirement between 750 and 1,600MHz to see if any could make good use of the "dual core" system. I measured CPU usage in MSI afterburner. Most games split the load between the two CPUS but usage always added up to around 100%, as in, one CPUs worth of usage. This split ranged from a flat 50% / 50% split of usage to a 80%/20% split of usage. I never saw more than 15% extra usage from the additional CPU. I suppose that was background usage which could in theory mean the other CPU was free to 'concentrate' on the game I was playing...

Games were tested at 640*480 or 800*600 and minimum - low settings. The benchmarking was locked to 60 seconds in FRAPS. I enabled hardware audio thinking this might offload a little work to the SB Live.

Here are MS Office 2000 charts:

4C3SeIq.png

(Edit- bugger the Willamette average and max frames are the wrong way around!)
BF2 shows the coppermines having better minimum, average and maximum frame rates - perhaps this game is multithreaded, as I saw the highest CPU usage in this game (it was something like 55% and 60%, so I was getting 15% extra CPU power beyond a single CPU). This had a minimum CPU requirement of a 1.7Ghz P4 and an FX5700 so I was cutting it fine. Let's have a look at the frametimes:

Lh9fUnPh.png

It's much more consistent and this was definitely felt in game. Time to move onto a train wreck, Morrowind, for which the minimum requirements are meaningless:

IZ168lh.png

This was generally unplayable for both systems. I did four laps around the Tutorial Town in that 60 second period. Let's dive into the frametimes:

d31QDxn.png

The Willamette was generally better and managed to render more frames in the 60 seconds but still had hugely inconsistant frame times.Next up, COD2:

Zq512ps.png

Surprisingly playable on both systems (minimum CPU 1.4GHz with a GeForce 2) but the Willamette has a clear win. The frametimes are also less terrible for the Willamette:

EAPHpf2.png

But still loks of spikes for both CPUs. The next game I tested was ToCA Race Driver 2. This has a requirement for a 800MHz CPU and a GeForce 2 / FX5200:

t9TqOC6.png

The Copperminers had another good show here even if the CPU usage wasn't much beyond a single CPU. How do the frametimes look?

a25L3TE.png

The Copperminers rendered half as many frames again compared to the Willamette with a much more consistent frame time in general. Next game, Far Cry which is looking for a 1GHz CPU and GeForce 2:

aZvUsUN.png

Not much between the two and surprisingly playable on the Copperminers. Next game GTA 3, whos minimum requirements are lying nearly as much as Morrowinds:

ruCVfDC.png

Both systems were very playable with neither exhibiting the sort of input lag I see on my 800MHz Slot 1. The Willamette had a texture streaming problem whereas the Coppermine didn't. The same HDD, DVD drive and other settings were used between both runs. Next game, Mafia (yet another specs lier):

NfX6ZL9.png

Neither system was playable and the similar results makes me wonder if it was GPU limited, but it simply can't be at 640*480 lowest settings with 30mb of VRAM used... The final game of the testing was UT99:

bv6bIHX.png

Both were playable but the Willamette appeared more playable, perhaps not surprising considering this game resolutely only used one 800MHz CPU. Both recorded a very low minimum frate. The frametimes show quite a bit of variation for the Coppermines, and I've removed one anomalous frame time (in the thousands of ms) for both systems.

HaVypXw.png

Here are some totals for minimum, average and maximum framerates across each of the tested games.

SGhLbwS.png

Conclusion
Both systems provide 1.6GHz of cold, hard, turn-of-the-millennium processing power, but the Willamette is objectively faster. Not that I recommend either for an early XP gaming build. A few games appeared to appreciate the extra CPU (or XP did in the background freeing up resources for the game), but most did not. Like I thought, non of the games that this CPU can run are really multi-threaded, and multi-threaded games will need more raw grunt per core/CPU. Windows Updates were all disabled and I wasn't doing any multitasking (other than FRAPS running). I might test out the games single CPU Coppermine CPU to see how much it compares to the dual CPU set up - I've got a few more days off work and central heating is being installed so I can't really go anywhere or do anything.

Ryzen 3700X | 16GB 3600MHz RAM | AMD 6800XT | 2Tb NVME SSD | Windows 10
AMD DX2-80 | 16MB RAM | STB LIghtspeed 128 | AWE32 CT3910
I have a vacancy for a main Windows 98 PC

Reply 1 of 12, by dionb

User metadata
Rank l33t++
Rank
l33t++
Almoststew1990 wrote on 2020-07-07, 23:15:

[...]

Conclusion
Both systems provide 1.6GHz of cold, hard, turn-of-the-millennium processing power, but the Willamette is objectively faster. Not that I recommend either for an early XP gaming build. A few games appeared to appreciate the extra CPU (or XP did in the background freeing up resources for the game), but most did not. Like I thought, non of the games that this CPU can run are really multi-threaded, and multi-threaded games will need more raw grunt per core/CPU. Windows Updates were all disabled and I wasn't doing any multitasking (other than FRAPS running). I might test out the games single CPU Coppermine CPU to see how much it compares to the dual CPU set up - I've got a few more days off work and central heating is being installed so I can't really go anywhere or do anything.

This is the crux really. 2x 800MHz != 1600MHz.

Even with recent, heavily threaded games, or indeed P3/4 era multithreaded rendering applications, with the same CPU archtecture, one core at twice the speed will always beat two at half speed. The only reason this isn't a hands-down win for the P4 is that the P4 had a much lower IPC than P3, so in fact was only barely able to pull clear of a single P3 at half its clock and occasionally being beaten - with the only clear win being in something that supported SSE2.

I'm a great fan of SMP, but its added value is in multitasking environments, and if any benchmark metric is really improved, it's responsiveness, not overall frame rate. That said, given how close the single-threaded, single-tasking scores are, throwing anything else at the CPU(s) at the same time as running those games will probably result in resounding wins for the dual P3. I predict even something as simple as MP3 playback at the same time as gaming would have that effect, and running a virus scan at the same time would be day/night difference.

Reply 2 of 12, by EvieSigma

User metadata
Rank Oldbie
Rank
Oldbie

Willamette P4s are weird, man. I have a Dell OptiPlex GX400 and somehow upgrading the CPU from 1.4GHz to 1.8GHz and installing proper chipset drivers made it feel 3x more responsive running Windows 2000.

Reply 3 of 12, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

Major problem would be chipset and memory speed. Is that early SIS based motherboard picked for P4? Can't tell, because both motherboard model names are likely listed wrong.

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

Reply 4 of 12, by Standard Def Steve

User metadata
Rank Oldbie
Rank
Oldbie

I'm rocking an actual 1.6GHz (single) P3. It's a 1400-S overclocked to 1628MHz on a 155MHz FSB. It's about as fast as a 2.4B Northwood P4 in most benchmarks. The PIIIs really were awesome CPUs.

94 MHz NEC VR4300 | SGI Reality CoPro | 8MB RDRAM | Each game gets its own SSD - nooice!

Reply 5 of 12, by Horun

User metadata
Rank l33t++
Rank
l33t++

Early adopters of some of the older cpu faced this same issue. Way back in the day the new cpu did not make up for a near equal overall performance of the previous one on a good updated motherboard. Is why Soc 423 did not last, barely beat a good DX4 in overall performance. just random thoughts..

Hate posting a reply and then have to edit it because it made no sense 😁 First computer was an IBM 3270 workstation with CGA monitor. Stuff: https://archive.org/details/@horun

Reply 7 of 12, by auron

User metadata
Rank Oldbie
Rank
Oldbie
EvieSigma wrote on 2020-07-08, 01:33:

Willamette P4s are weird, man. I have a Dell OptiPlex GX400 and somehow upgrading the CPU from 1.4GHz to 1.8GHz and installing proper chipset drivers made it feel 3x more responsive running Windows 2000.

not that unexpected given that it was a very different architecture which windows 2000 predated by almost a year... having device manager presumably full of yellow exclamation marks would be another indicator that things aren't working right.

Reply 8 of 12, by Almoststew1990

User metadata
Rank Oldbie
Rank
Oldbie
The Serpent Rider wrote on 2020-07-08, 01:51:

Major problem would be chipset and memory speed. Is that early SIS based motherboard picked for P4? Can't tell, because both motherboard model names are likely listed wrong.

The board is actually called a MSI 645 Ultra, using a SIS 645 chipset, I called it 650 in the benchmarks. The motherboard is an early Socket 478 motherboard that appears to only accept 400MHz FSB CPUs. I dropped the memory speed to 200MHz so it was the same as the SDR100MHz RAM (sort of). It also "doesn't do" ACPI power management, which XP was upset about when I was installing the OS, and when I turn the PC off it says "it's now safe to shut down your computer". Definitely something not right there.

EvieSigma wrote on 2020-07-08, 01:33:

Willamette P4s are weird, man. I have a Dell OptiPlex GX400 and somehow upgrading the CPU from 1.4GHz to 1.8GHz and installing proper chipset drivers made it feel 3x more responsive running Windows 2000.

I deliberately went with XP as I have so much experience with it and the AGP and IDE drivers were a non issue.

dionb wrote on 2020-07-07, 23:38:

This is the crux really. 2x 800MHz != 1600MHz.

Yeah, I wasn't expecting it to be 100 parallel usage or anything. In the CPU-Z benchmark it scored 11 points for single-threaded tasks and 15 points for multi-threaded tasks, so a theoretical maximum boost of 46% by having 100% extra CPU on board.

Ryzen 3700X | 16GB 3600MHz RAM | AMD 6800XT | 2Tb NVME SSD | Windows 10
AMD DX2-80 | 16MB RAM | STB LIghtspeed 128 | AWE32 CT3910
I have a vacancy for a main Windows 98 PC

Reply 9 of 12, by matze79

User metadata
Rank l33t
Rank
l33t
EvieSigma wrote on 2020-07-08, 01:33:

Willamette P4s are weird, man. I have a Dell OptiPlex GX400 and somehow upgrading the CPU from 1.4GHz to 1.8GHz and installing proper chipset drivers made it feel 3x more responsive running Windows 2000.

1.4Ghz Williamette to Northwood ?
Its even more worse if one has SD-RAM PC133 P4 mainboard..

https://www.retrokits.de - blog, retro projects, hdd clicker, diy soundcards etc
https://www.retroianer.de - german retro computer board

Reply 10 of 12, by EvieSigma

User metadata
Rank Oldbie
Rank
Oldbie
matze79 wrote on 2020-07-08, 15:26:
EvieSigma wrote on 2020-07-08, 01:33:

Willamette P4s are weird, man. I have a Dell OptiPlex GX400 and somehow upgrading the CPU from 1.4GHz to 1.8GHz and installing proper chipset drivers made it feel 3x more responsive running Windows 2000.

1.4Ghz Williamette to Northwood ?
Its even more worse if one has SD-RAM PC133 P4 mainboard..

Nah, Intel made a 1.8GHz Socket 423 chip. I dunno why when they already had a 1.7GHz, but yeah.

Reply 12 of 12, by debs3759

User metadata
Rank Oldbie
Rank
Oldbie
Almoststew1990 wrote on 2020-07-08, 07:27:
dionb wrote on 2020-07-07, 23:38:

This is the crux really. 2x 800MHz != 1600MHz.

Yeah, I wasn't expecting it to be 100 parallel usage or anything. In the CPU-Z benchmark it scored 11 points for single-threaded tasks and 15 points for multi-threaded tasks, so a theoretical maximum boost of 46% by having 100% extra CPU on board.

A boost from 11 to 15 = 1.364, or 36.4%.

Pretty sure some apps would be much better, depends how well they multitask on the platform.

See my graphics card database at www.gpuzoo.com
Constantly being worked on. Feel free to message me with any corrections or details of cards you would like me to research and add.