Orientation.

This post will be constantly updated.

(April 2023)
At this point in time 20 years ago the Intel Canterwood chipset (875P) just had been released on April 14 along with the Pentium 4 Northwood 3.0 GHz. This platform was also relevant for workstations and small servers with Xeon Prestonia and Gallatin CPUs.
On the AMD-side the Barton Athlon XP 3000+ was the top CPU and the nForce2 was the best chipset. For workstations AMD had the 760MPX chipset and Athlon MP (2600+) CPUs. This Saturday (22-04-23) marks the 20 year anniversary of the launch of AMDs 64-bit K8 architecture in the form of the Opteron 240, 242 and 244.

Relevant platforms during the spring of 2003.

Intel.
875P - Canterwood
845xx - Brookdale-xx
E720x - Granite Bay, Plumas and Placer. (workstation/server)
860 - Colusa
850(E) - Tehama(-E)
VIA P4X333/400
SiS 645DX/648

AMD.
AMD 760MP(X)
AMD 8000
VIA KT333/400, KT600 in May.
nForce2
SiS 746FX

Relevant video cards during the spring of 2003.

ATI Radeon 9xxx
nVidia FX series
(9800Pro launched in Mars, FX5900Ultra in May. 9800XT and FX5950Ultra are only minor clock-bumps released a few months later.)

Relevant reviews.

Barton 3000+ review
https://www.anandtech.com/show/1066
Pentium 4 3.0C
https://www.anandtech.com/show/1093
Intel Canterwood (875p)
https://www.anandtech.com/show/1094
nForce2
https://www.anandtech.com/show/1005
AMD Opteron (K8)
https://www.anandtech.com/show/1098
Intel Prestonia and Gallatin
https://www.anandtech.com/show/1032
ATI Radeon 9800/9600 Pro
https://www.anandtech.com/show/1077
nVidia Geforce FX5900 Ultra
https://www.anandtech.com/show/1104

A Toms Hardware video showing an Opteron system just before the release 2003
https://www.youtube.com/watch?v=xDZbhftYyNk

Level1Techs on Youtube made a video on K8 and x86-64
https://www.youtube.com/watch?v=rMvwHMuTLGw

Reserved for index.

Reserved for links.

Current guidelines, requests and misc info.

I will try to get a socket 940 system running this weekend in time for the 20 year anniversary of the AMD K8 release. In time there will be some benchmarking to compare these "new" CPUs against other "current" platforms. Perhaps other members have suitable systems to help us get a more complete picture of the hardware landscape during the spring of 2003.

When it comes to true period correct K8 motherboards from April 2003 that's a hard nut to crack. I'm thinking any Socket 940 (not 939) motherboard without PCI-E will do as both nVidia and VIA at least had pre-release boards with AGP out during Q2 2003. I think VIAs K8T800 chipset on paper actually launched with the Opteron but in the real world they were hard to find during Q2.

When it comes to CPUs any single core 0.13um s940 Opteron could (/should/would) be made into a period correct one by lowering the multiplier.

i had completely forgotten about the Opteron! this is interesting because of its 'modern obsolete' status

these very early forays into the new world are largely forgotten today, i hardly ever see very early 64 bit stuff now

Yes, but however they lacked of the instruction CMPXCHG16B. So they were 64bit processors, but they were limited to Windows x64 8.0 and EOL on other Windows x64 with Microsofts patches first.

Disruptor wrote on 2023-04-18, 21:30:

Yes, but however they lacked of the instruction CMPXCHG16B. So they were 64bit processors, but they were limited to Windows x64 8.0 and EOL on other Windows x64 with Microsofts patches first.

Well if the issue can't be bypassed I guess Windows 7 will have to do for testing newer stuff! 😁
I remember reading something about this instruction somehow causing issues for some people with much newer CPUs when Windows 8.1 rolled out.

.

I spent yesterday evening hunting parts in my basement and attic.

So far I have found.

MSI K8T Master2-FAR. The last time this motherboard was in action was in May 2015 benching Doom 3 for the Doom3 performance thread, I hope it still works.

2x Opteron 248. These are the CPUs I used for the Doom 3 benchmarking. I also found 2x Opteron 250 (0.13um).

I have not found any of my boxes with decent AGP cards from the correct era but I did find two Radeon 9800 Pro, one without a cooler and one with the cooler unattached. I think they are from a scrap lot I bought but I don't remember testing them, with luck one could be persuaded to work.

It would be nice to find my spare Geforce FX5900 Ultra but if I don't I do know in which system the other one is located.

.

My main worry at the moment is the multiplier control. I'm at work reading up on issues people had with the MSI K8T Master2-FAR back in 2003. As the board supports the Athlon FX in single CPU mode I figured the board has multiplier control so I would be able to underclock the Opteron 248 from 2.2GHz to 1.8GHz (the speed of the 244). Now I read that at the time of the release of the Athlon FX51 MSI had not implemented multiplier control in the BIOS and I have no clue if they later did. I never had any reason to change multiplier with this board as I do not own a s940 FX CPU and the Opterons multipliers can't be increased. Perhaps lowering the multiplier can be solved with some "Cool and Quiet" software.

In the worst case I will just have to Ebay a couple of Opteron 244.

.

Edit:

I just caved and ordered 2x Opteron 244 for a totally unreasonable 45 euro shipped... The "real value" is ~2 euro each so money well spent... It's for science! 😁

The 248 is probably a better "bin" of the Sledgehammer core compared to the 244 available at launch so this way we will (in theory) get more accurate overclocking numbers. In the real world the overclocking with all of these CPUs on this motherboard will be limited by the small amount it's possible to increase the reference clock. If I remember right the BIOS does support locking the PCI and AGP-bus but early steppings of the VIA-chipset doesn't!

These CPUs won't be here in time for the weekend but that won't stop me from getting the K8-system up and running as long as the motherboard still works. I guess I will have to start the benchmarking with some of the competing platforms if I can't lower the multiplier with the Opteron 248.

I just noticed that the official launch date seems to be the 22th of April 2003 and not the 23th (the date Anandtech posted their review). Not that this matters much 😁

Just as a heads up - none of the 939 and 940 chips have the instructions required for 8.1 and 10. Even the dual core chips lack CMPXCHG16B - I tried 8.1 and 10 IoT LTSC on my 64 x2 4600+ (Manchesterl) as well as a Venice 3200+. Neither were able to get as far as showing the Windows logo and restarting.

Only AM2 (a.k.a. K8 "F" stepping), and later AMD platforms support Windows 8.1 x64 and Windows 10 x64.

Good time for you to make this thread. I am putting together a system with a MSI K8T Neo2 for XP and 98 gaming, X800 GTO AGP and an Audigy 2 ZS, these are really underappreciated platforms for the technology they pioneered that we pretty much depend on now (multicore, 64-bit instructions).

I still have Opteron down as 'modern' in my mind 😀
We had HP Opteron servers at where I worked back 15 years ago and great they had loads of cores, but our trading platform needed 1 fast core for much of its job and that struggled on these CPUs.

I still have some huge dual processor single core Opteron boards with lots of RAM slots in my collection.

SteveC wrote on 2023-04-19, 18:00:

I still have Opteron down as 'modern' in my mind 😀
We had HP Opteron servers at where I worked back 15 years ago and great they had loads of cores, but our trading platform needed 1 fast core for much of its job and that struggled on these CPUs.

If they had many cores but low clock speeds they were probably Magny Cours Opterons. They could indeed be considered "modern" and there are probably some left in operation somewhere.

The early single core K8 Opterons from 2003 on the other hand... 😁

.

Unknown_K wrote on 2023-04-19, 18:08:

I still have some huge dual processor single core Opteron boards with lots of RAM slots in my collection.

I also have a couple of really huge dual socket s940 boards, sadly both of them are PCI-E based so likely from year 2005. I should have a smaller ASUS K8N-DL somewhere but it's also PCI-E based.

The only true year 2003 socket 940 board I own is the ATX-sized MSI K8T Master2-FAR. It's a nice board but because of the small footprint only one CPU has memory slots and a single 4-phase VRM powers both CPUs! Other than the size and the limitations it brings the most serious issue seems to be that MSI quickly abandoned the board to focus on newer products so every BIOS after version 1.3* is totally broken with dual CPUs. The reason later BIOS versions only works as intended with one CPU is probably because of the intern that was responsible for coding only had access to a single one (and that there is a single CPU version of the motherboard). The board only supports 0.13um CPUs as the VRM can't handle the higher current draw of the lower voltage 0.09um CPUs.

* There is supposedly a somewhat buggy BIOS version 1.35 beta that actually works with 2 CPUs but I never found it, version 1.4 and 1.5 have broken SMP support. If anyone should happen to find this BIOS let me know! The reason I want it is that this BIOS supports the (0.13um) Opteron 250, version 1.3 supposedly doesn't.

.

I found the HDD with XP I used for the Doom 3 benchmarking with the K8T Master2 (I do actually mark them) and the board booted XP at first try. For some reason the mouse would not work, neither PS/2 nor USB but the issue solved it self after 10 or so reboots and some swearing.

edit +l

Interesting topic to say the least! The hardware from that era can hardly be used to run Windows 6.X or later; modern internet and today era software. They are single core so are strongly limiting performance and IMHO are best used with XP with the exception of dual socket systems that can relatively still be used on the modern internet. Modern Windows starting with Vista have such a large overhead that performance is crippled; I tested 3d mark 2001 on the following: Athlon XP 3200; 2*1gb RAM; Abit AN7; geforce 7800gs; WD raptor and the score was something like 50% lower on 7 as far as I can remember - I have to look for the screenshots.; the same thing happened when i tested the Pentium 4 with 2gb, same video card on the Abit IC7 - MAX3.
I have all the Abit boards from socket 754 with the Athlon 64 CPU's but I would not try something newer then XP because they only have 2 DIMMS so with 2gb of ram performance with NT 6.X would be terrible however on XP they simply fly!

Here is a picture showing the setup as of now... perhaps not very aesthetically pleasing but...

I need to find some nice DDR400 (ECC REG) memory, I'm sure I have some somewhere. For some reason I used DDR333 (ECC REG) when I last used the system for the Doom3 benchmarking but I don't remember why. Sadly CAS 2.5 or CAS 3 is the only relevant user configurable memory setting in the BIOS so if one want CAS 2 or change memory multiplier that needs to be solved through the SPD.

.

Level1Techs on YouTube just made a video on K8 and x86-64, I added a link in the orientation post.

nd22 wrote on 2023-04-20, 15:11:

Interesting topic to say the least! The hardware from that era can hardly be used to run Windows 6.X or later; modern internet and today era software. They are single core so are strongly limiting performance and IMHO are best used with XP with the exception of dual socket systems that can relatively still be used on the modern internet. Modern Windows starting with Vista have such a large overhead that performance is crippled; I tested 3d mark 2001 on the following: Athlon XP 3200; 2*1gb RAM; Abit AN7; geforce 7800gs; WD raptor and the score was something like 50% lower on 7 as far as I can remember - I have to look for the screenshots.; the same thing happened when i tested the Pentium 4 with 2gb, same video card on the Abit IC7 - MAX3.
I have all the Abit boards from socket 754 with the Athlon 64 CPU's but I would not try something newer then XP because they only have 2 DIMMS so with 2gb of ram performance with NT 6.X would be terrible however on XP they simply fly!

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.

You would not believe how fast the K8T Master2 FAR boots up Windows XP SP3. (from a spinning disk). The board pretty much POSTs the instant you push the button which is kind of strange for a dual CPU s940 motherboard.

Skyscraper wrote on 2023-04-20, 17:15:

Here is a picture showing the setup as of now... perhaps not very aesthetically pleasing but...

A beautiful setup.jpg

some serious cooling going on there!

this will be good ... 😀

gerry wrote on 2023-04-20, 19:46:

some serious cooling going on there!

this will be good ... 😀

I first tried this CPU cooler... it was not very impressive...

.

.

The weather is rather great here at the moment but it's going to get (alot) worse in the coming days letting me spend more time tinkering.

As today is the actual year 20 year anniversary of the K8 architecture and the x86-64 standard I thought that I would at least post a picture showing a motherboard and CPU available at launch. In reality the Opteron 244 CPUs I ordered will not be here until next week so one of my Opteron 248 acts as a stand-in. The MSI K8T Master(2) FAR board was probably hard to find in April 2003 but it couldn't have been that many weeks away as it was one of the boards used for AMD FX51 previews.

Some progress is being made!

After testing depressingly many broken Radeon 9800 cards I found a working 9800 Pro 128MB.

.

I have also found some working PC3200 ECC REG memory and the dual Prestonia system housing them.

I think the K8T Master2 actually posts a bit slower with the faster memory.