gerry wrote on 2023-05-06, 19:03:

does the opteron feel 'responsive' enough in use though? it might emulate the experiences of early adopters of the time

It does feel very responsive but so did the Pentium 4 3.0C. It's hard to emulate any differences people might have experienced between the platforms back then as everything is somewhat different now even when we try to recreate the past.

For example the reason I have not updated the Opteron 244 3DMark scores yet and added the rest of the benchmarks using a fresh Windows XP install as I planned is VIA + AGP + ATI + XP-SP3 issues. I don't really think people experienced these problems just as much back then, at least I didn't when testing using the now wiped XP-SP2 install. Only 3dmark05 and 06 seems affected and freeze in firefly forest, games and other benchmarks run fine so back in 2003 this exact problem would have been a non issue anyhow. Back then there were always known combinations of drivers that worked but so far the MSI K8T Master2 + Radeon 9800 Pro with XP-SP3 seems to have me beaten. 😁

While I have not solved the issues I have now worked "around" them.

I noticed that I could run 3dmark05 and 06 right after I reinstalled the Omega version of the Catalyst 7.4 driver with fast writes disabled and before rebooting. This was enough to get a more representative 3dmark05 score even if it's still a bit below the "mark". I have a pretty good understanding of how these parts should perform in 3DMark so I can tell where there likely is performance left on the table. I also did new and more representative runs with the Opteron in 3DMark2001 and 03 so it's now easier too see where things are heading in the "future".

.

I found my boxed Radeon 9800 Pro (R350 128MB). I tested this card to make sure the issues I was experiencing weren't caused by issues with the other 9800 Pro.

Most of the paperwork seems to be located somewhere else but I'm confident I still have it.

.

One conclusion from my testing with the MSI K8T Master2 is that for the AMD FX51 launch in September it's probably better to emulate an AGP nForce3 or K8T800 s940 platform using s939. The only notable difference between early nForce3/K8T800 s940 and s939 motherboards is that the former uses registered memory and the latter doesn't, the performance should be very close.

I will of course still use the K8T Master2 FAR for dual CPU testing and perhaps some testing with Nvidia cards but I doubt I have the patience to keep using this board with ATI cards and as we know ATI cards were the fastest in 2003.

.

I will update the 3dmark benchmarks and add the rest of the benchmarks for the Opteron later today then it's time to build a Barton system.

Here are the rest of the benchmarks results for the Opteron 244.

In rendering and memory benchmarks the results are a mixed bag. In some benchmarks the Opteron 244 just beats the P4 3.0C in others it loses, often by a larger margin.

.

Opteron 244 Frybench x86

Opteron 244 Cinebench 11.5

Opteron 244 SuperPI 1M

Opteron 244 SuperPI 32M

Opteron 244 AIDA64 Cache & Memory Benchmark plus PhotoWorxx

.

Continued in the next post.

I added 7-Zip to the benchmark-sweet.

I will have to do I quick 7-Zip run with the P4 3.0C later before upgrading that system to the P4 3.2C.

.

Opteron 244 7-Zip x86 32M

.

Finally we have Doom III and as we all know the CPU that is faster in Doom III is the better CPU! 😁

.

What about overclocking you perhaps ask. I did run overclocked benchmarks with the Pentium 3.0C and from 3000 to 3600 MHZ it doesn't gain that much as the motherboard and memory can't run as tight timings when the FSB increase. With increased voltage this exact specimen of P4 3.0C (packing date December 2003) can run at 3800 MHz but reading reviews from the April release 3600 MHz seems more reasonable.

The Opteron 244 performance on the other hand scales great with frequency but instead we run into issues with the ATI AGP cards and the non existing PCI/AGP lock. When I did some testing using the old install with XP-SP2 before wiping the disk I could run higher FSB with Nvidia cards than with ATI cards and in its current state with XP-SP3 and driver hell it's even worse.

Next week the Geforce 5900 Ultra is released, perhaps it's worth to do some new overclocked runs then to get a fairer overclocked comparison.

.

For now here is the "magic" 20000+ points 3DMark2001 run with the Pentium 4 which would have been rather impressive back in April 2003.

This is the mark the Barton will have to try to beat! 😁

Pentium 4 3.0C @3.6 GHz Radeon 9800 Pro 3DMarl2001

.

More video card overclock... If it had been May 2003 this score would have been posted on xtremesystems.org, not here! 😁

.

[Edit]
Added another overclocked result.
[/Edit]

Skyscraper wrote on 2023-05-07, 07:14:

gerry wrote on 2023-05-06, 19:03:

does the opteron feel 'responsive' enough in use though? it might emulate the experiences of early adopters of the time

It does feel very responsive but so did the Pentium 4 3.0C. It's hard to emulate any differences people might have experienced between the platforms back then as everything is somewhat different now even when we try to recreate the past.

For example the reason I have not updated the Opteron 244 3DMark scores yet and added the rest of the benchmarks using a fresh Windows XP install as I planned is VIA + AGP + ATI + XP-SP3 issues. I don't really think people experienced these problems just as much back then, at least I didn't when testing using the now wiped XP-SP2 install. Only 3dmark05 and 06 seems affected and freeze in firefly forest, games and other benchmarks run fine so back in 2003 this exact problem would have been a non issue anyhow.

a good point, as much as we try we wont really get an equivalent experience, just something close to it

back then i'd imagine almost every pc user was still just fine with p4 and athlon xp, the bright new future wasn't a common experience

I was never hyped with Pentium 4´s, at the time circa ~2002, i had a pentium 3 system and a friend had a pentium 4 "willy" socket 423 - 1.7 ghz, and we spent a weekend in a lanparty benchmarking both systems, pentium 4 was very "meh", not disappointing, but meh, we didnt know nothing about SSE2, large pipelines, and tech crap, just it was not as fast as intel promised.

Gumur.gurl wrote on 2023-05-10, 01:54:

I was never hyped with Pentium 4´s, at the time circa ~2002, i had a pentium 3 system and a friend had a pentium 4 "willy" socket 423 - 1.7 ghz, and we spent a weekend in a lanparty benchmarking both systems, pentium 4 was very "meh", not disappointing, but meh, we didnt know nothing about SSE2, large pipelines, and tech crap, just it was not as fast as intel promised.

I "upgraded" from an Athlon Thunderbird to a Willamette Pentium 4 2.0 right when it was released. It was through an "employer - worker rent scheme" paid mostly by the government and as it usually goes with such schemes the computers ended up being as expensive as practically possible (expensive for the government that is). It was a Fujitsu Siemens system with a Geforce 3, 1024MB memory and so on, totally ridiculous specs for 2001.

The Willamette might have been kind of shit but it inhabited a lot of really nice systems.

.

.

In forum threads back from 2002 - 2003 you always see references to the "Chieftec 360W". It's usually mentioned in a negative way as an excuse/cause for everything from instability or limited overclock to memory failing or motherboards blowing up. What kind of destructive beast is this "Chieftec 360W"? Lets find out!

.

The most common Chieftec 360W unit sold, at least in Europe during 2003 was without a doubt the HPC-360-202. I saw these units made by Sirtec everywhere back then and I still sometimes see them in dumpster finds and such now 20 years later.

Were the HPC-360-202 as bad as you would think from reading old forum threads and such? No not at all, they are in fact pretty solid and from my experience rarely have bad caps . When I can't find a spare recapped Tagan 380W/480W unicorn (so 99% of the time) the Sirtec 360W units are my go to power supplies for using with power hungry old systems that use 5V for powering the CPU (among many other things directly on or connected to the motherboard).

One funny thing is how the passive PFC seems to be marketed heavily, not really a sign of quality these days.

.

.

This is the motherboard I wanted to use for the Barton Athlon XP 3200+ testing, the Abit AN7. The AN7 supports high enough voltages, especially memory voltage to be able to push a Barton hard enough for it to compete with the P4 3.0C and the Opteron 244. Sadly the cooler I want to use, a Zalman doesn't fit because of clearance issues around the socket. I hate the "heavy duty" clips many other high end Socket A coolers use and as this board is pristine and I don't want to damage the socket I will use another, lesser board.

.

Here is the cooler I'm going to use, a Zalman CNPS7700-AlCu. This cooler is from 2004 but other similarly performing coolers in the same series were available 2003.

.

On another note reading an old Abit AN7 review made me realize that some reviewers really pushed the settings and optimized as much as they could both at stock and overclocked settings for their reviews. I think I will have to revisit the Pentium 4 and Opteron systems and do some more testing when I'm done with the Barton...

So here is finally some results with the Barton XP3200+ released 20 years ago give or take a few days.

The testing didn't go to plan at all and I have been unexpectedly busy with IRL/Work stuff this last week. The Asus A7N8X is severely limiting the overclocking with its 5V design for powering everything on the motherboard including the CPU. The issue is not that the PSUs 5V rail is drooping but that it's impossible to get enough power to where it needs to be.

Other issues includes lack of SSE2 which makes it impossible to run some benchmarks and for some reason the HDD performance is so terrible that the virus scan test in PCMark05 doesn't even manage to run one full cycle before timing out...

If I had more time I'm sure the HDD issue could be sorted but I want to squeeze in some testing with the Geforce FX5900 Ultra as well as this card also was released pretty much exactly 20 years ago this weekend. If I want some results with the FX5900 Ultra before going away on vacation this Thursday i better use a platform less prone to "issues" so I will use the Pentium 4 on the P4C800.

.

Lets start with the 3dmark runs.

Athlon XP 3200+ "Barton" Radeon 9800 Pro 3DMark2001

Athlon XP 3200+ "Barton" Radeon 9800 Pro 3DMark03

Athlon XP 3200+ "Barton" Radeon 9800 Pro 3DMark05

Athlon XP 3200+ "Barton" Radeon 9800 Pro 3DMark06

.

All in all the 3DMark results are probably what could be expected.

When it comes to memory heavy benchmarks the Barton XP 3200+ actually did rather well. How the Athlon XP handles rendering is somewhat unclear because it couldn't run Cinebench 11.5 or Frybench as the Barton core lacks SSE2.

.

Athlon XP 3200+ "Barton" SuperPI 1M

Athlon XP 3200+ "Barton" SuperPI 32M

Athlon XP 3200+ "Barton" 7-Zip x86 32M

Athlon XP 3200+ "Barton" Aida64 Cache & Memory Benchmark plus PhotoWorxx

.

And finally Doom III, here the Barton can't keep up at all with the Pentium 4 and the Opteron 244.

.

I might add some overclocked results later today but for now I got a lawn to mow.

Skyscraper wrote on 2023-05-14, 07:52:

I'm sure there will be plenty of 32bit Windows XP benchmarking in the thread, not only because single core K8 CPUs are a bit too slow for Windows 6.x -x64 but also because the competition couldn't run a 64-bit OS.

Ahem... a 478 SOCKET - 64 bit - pentium 4, exist.

im running my benchmark and installing games, to try, with my all full of power INTEL EXTREME GRAPHICS 2.

Gumur.gurl wrote on 2023-05-14, 08:17:

Skyscraper wrote on 2023-05-14, 07:52:

I'm sure there will be plenty of 32bit Windows XP benchmarking in the thread, not only because single core K8 CPUs are a bit too slow for Windows 6.x -x64 but also because the competition couldn't run a 64-bit OS.

Ahem... a 478 SOCKET - 64 bit - pentium 4, exist.

im running my benchmark and installing games, to try, with my all full of power INTEL EXTREME GRAPHICS 2.

Yes but not in 2003, 20 years ago.

The 64 bit version of the Prescott s478 CPU (OEM only) showed up late 2004 or early 2005.

That said I do want a 64 bit s478 CPU but they are kind of rare, I think Intel made them exclusively for IBM.

Skyscraper wrote on 2023-05-14, 10:20:

Yes but not in 2003, 20 years ago. […]
Show full quote

Gumur.gurl wrote on 2023-05-14, 08:17:

Skyscraper wrote on 2023-05-14, 07:52:

I'm sure there will be plenty of 32bit Windows XP benchmarking in the thread, not only because single core K8 CPUs are a bit too slow for Windows 6.x -x64 but also because the competition couldn't run a 64-bit OS.

Ahem... a 478 SOCKET - 64 bit - pentium 4, exist.

im running my benchmark and installing games, to try, with my all full of power INTEL EXTREME GRAPHICS 2.

Yes but not in 2003, 20 years ago.

The 64 bit version of the Prescott s478 CPU (OEM only) showed up late 2004 or early 2005.

That said I do want a 64 bit s478 CPU but they are kind of rare, I think Intel made them exclusively for IBM.

lOL, RIGHT , they just glued a prescott to a 478 pin socket.

Gumur.gurl wrote on 2023-05-14, 11:41:

lOL, RIGHT , they just glued a prescott to a 478 pin socket, also games of 2003 are great to play 20 years after with low end hardware of the time, in hl2 i get 20 fraps and call of duty 30-40 fps, EZTREME GRAPHICS 2 vga was such a monster card for its time, lmao. i Cant run cb15 because is 64 bit only aaaaaaaaaaaargh!

Well Socket 478 was the launch platform for the Prescott so not much gluing needed to be done.

The Prescott 1M core has the hardware present to support 64 bit but Intel in their vast wisdom disabled 64 bit support for the first models of the s775 Precott P4 and (almost) all Prescotts for s478 .

Vogons gets more insteresting™ by the day... 😜

Skyscraper wrote on 2023-05-14, 12:23:

Well Socket 478 was the launch platform for the Prescott so not much gluing needed to be done. […]
Show full quote

Gumur.gurl wrote on 2023-05-14, 11:41:

lOL, RIGHT , they just glued a prescott to a 478 pin socket, also games of 2003 are great to play 20 years after with low end hardware of the time, in hl2 i get 20 fraps and call of duty 30-40 fps, EZTREME GRAPHICS 2 vga was such a monster card for its time, lmao. i Cant run cb15 because is 64 bit only aaaaaaaaaaaargh!

Well Socket 478 was the launch platform for the Prescott so not much gluing needed to be done.

The Prescott 1M core has the hardware present to support 64 bit but Intel in their vast wisdom disabled 64 bit support for the first models of the s775 Precott P4 and (almost) all Prescotts for s478 .

Vogons gets more insteresting™ by the day... 😜

Im installing and benchmarking the next 3 games .

For comparison, a slightly newer processor with Radeon 9800 Pro 😉
Dual channel memory is not supported..

Tom.. wrote on 2023-05-14, 13:28:

For comparison, a slightly newer processor with Radeon 9800 Pro 😉
Dual channel memory is not supported..

Nice!

.

.

Here are some overclocked results with the Barton Athlon XP "3200+".

The actual CPU I'm using is a Mobile Barton 2400+ or 2500+ but all types of Barton can be made into all other types of Barton by manipulating bridges on the CPU package. After a certain date AMD introduced a "super lock" locking the multiplier on desktop Bartons and the way to get around this was to turn your CPU into a Mobile CPU thus allowing changing multipliers on the fly in the OS. Sadly changing multipliers in the OS didn't work with nForce2 based motherboards so you had to either buy an always fully unlocked real Mobile Barton or hope that you would get an unlocked desktop CPU. I didn't want any hassle so I picked one that was sure to be unlocked but the overclock is very typical for a 3200+ .

The motherboard I'm using, the Asus A7N8X-E Deluxe do not have all signal lines for setting the multiplier connected to the socket so I'm limited to multipliers of 12.5x or less. Because this CPU isn't that great and the motherboard can't deliver the power needed to push it beyond 2500 MHz the multiplier limitation is a non issue. This is the very motherboard that for no good reason killed my rare 166 MHz FSB 2333 MHz HP OEM version of the 3200+, the only CPU "I" ever killed. Had I used a "better" motherboard for this testing I would probably aimed for something like 12x217 for 2600 MHz which wasn't an unreasonable even though admittedly good overclock back then.

The memory I'm using is an unmatched pair of what I think is Winbond UTT. These modules must be the worst UTT (BH5 UnTesTed) in existence as they are only rated at 2-3-3-6 timings and wont do over 200 MHz with 2-2-2-x timings at 2.8V. The specimen of A7N8X motherboard I'm using will not do much over 200 MHz FSB fully stable with the stock BIOS so the crappy memory is also not really an issue just beware that you could get a Barton running faster/better than this even back in 2003.

.

The Asus A7N8X-E Deluxe.

Two unmatched Geil sticks that just manages to run at 200 MHz with 2-2-2-x timings.

.

.

In 3DMark2001 the Barton Athlon XP with the Radeon 9800 pro see decent gains but it can't reach the overclocked Pentium 4.

In 3DMark03 the overclocked Barton and Radeon 9800 Pro performs really well.

The overclocked Barton is stable enough to run with 100% load for extended periods of time but I would not burden this old hardware with a power virus such as Prime95 or the like.

Overclocked Barton Athlon XP SuperPI 1M

Overclocked Barton Athlon XP SuperPI 32M

Overclocked Barton Athlon XP 7-Zip x86 32M

.

And finally sadly no, overclocking the Barton Athlon XP does not make it a good CPU for Doom III.

I have such an AMD Barton 2500+@3200+ processor. The motherboard is very slow, slower than the KT600 based ones I had.

Tom.. wrote on 2023-05-14, 16:20:

I have such an AMD Barton 2500+@3200+ processor. The motherboard is very slow, slower than the KT600 based ones I had.

The Asus A7V880 is nice! I have one as well but as you say it isn't the fastest board.

I'm sure there are BIOS mods out there that could improve things and in the end any 200 MHz FSB capable Socket A motherboards stable enough to run video cards with PCI-E ---> AGP bridges are good motherboards.

Skyscraper wrote on 2023-05-14, 16:46:

Tom.. wrote on 2023-05-14, 16:20:

I have such an AMD Barton 2500+@3200+ processor. The motherboard is very slow, slower than the KT600 based ones I had.

The Asus A7V880 is nice! I have one as well but as you say it isn't the fastest board.

I'm sure there are BIOS mods out there that could improve things and in the end any 200 MHz FSB capable Socket A motherboard stable enough to run video cards with PCI-E ---> AGP bridges are good motherboards.

Could you do a sound test in 3dmark03 and show me the result? I think there might be something wrong with my motherboard..

Tom.. wrote on 2023-05-14, 16:53:

Could you do a sound test in 3dmark03 and show me the result? I think there might be something wrong with my motherboard..

I had not really planned to setup a system with the A7V880 until it's 20 years since its (the chipset) release.

If I find the time after dealing with the Geforce FX5900 Ultra and the Pentium 4 3200 I could perhaps do a quick build in advance so you don't have to wait 8 months or so 😁

What is the issue with your board?