About AMD K7 based systems, i have 4 boards and multiple CPUs, also a complete Athlon XP 3000+ build

Motherboard:
MSI KT6V-LSR (MS-7021, 2 boards, one begin used on the build)
MSI K7N2 (MS-6570)
ASUS A7V-E

CPUs:
Athlon XP 3000+ FSB 400 (used on the build)
Athlon XP 3000+ FSB 333
Athlon XP 2600+ FSB 333
Athlon XP 2400+ FSB 266
Athlon XP 2200+ (2 CPUs)
Athlon XP 2000+ (Thorton)
Athlon XP 2000+ (Palomino)
Athlon XP 1800+
Athlon XP-M 1500+ 45 watt
Athlon XP-M 1500+ 35 watt
Sempron 2800+
Sempron 2400+
Athlon 1200 FSB 200
Duron 700

Looks like us K7 fans had a similar PC history at one point 😀

Nice boards guys. I'm (as you probably guessed) an Abit man myself. Not an asus fan by any stretch of the imagination, but the A7N8 nforce 2 series boards are very good. Even so, they don't hold a candle to the Abit AN7 or it's little brother the NF7. Reliability-wise they are pretty similar. Abit's late socket A series does not suffer from the capacitor plague - all my KT400 / KT600 and NF2 abit boards use quality japanese caps and work great. Both the A7N8X an the AN7 have the same issue unfortunately - crappy EEPROM chips... they will go bad w/o warning, but it's a really easy fix with either an eeprom burner or a hot-flash using another board and a new (or old working) chip.

The Abit boards do some really nice things when it comes to memory compatibility and overclocking, witch is lacking on the A7N8x. The Asus nf2 is a bit picky about ram, regardless of revision (I have 2 working boards and one that needs a re-flash) while the AN7 will run anything (albeit at safer timings). I guess the only real difference between the two is how well the BIOS is made on the Abit board, and of course the uGuru witch is great to use. I actually managed to push my 2333MHz 3200+ to 15.5x166 / 1.78v stable and cool (64C max spikes, 61 average). In fact the AN7 can change the multiplier on some chips, witch I can't do on the A7N8x Deluxe or my MSI K7N2 Delta - dunno why - and it also provides better incremental FSB adjustments, and of course AGP / PCI speed locking regardless of FSB. The NF7 does not provide some of those features, but it retains the great memory compatibility of the AN7. The A7N8x is also a good overclocker - getting most 2500+ to 3200+ is fairly easy on it, but unlike the abit I have to bump up the voltage a tiny bit more, and it demands really good ram when overclocking. The K7N2 is even better when it comes to OC, but it had bad caps (replaced) and is not as reliable as the Asus or the Abit, making it more of a OC board for when I want to do FSB overclocking.

In total I have:
- 2x ABIT AN7 one boxed, one in use in my socket A machine
- 1x ABIT NF7-S
- 1x ABIT KV7 (KT600)
- 1x ABIT KT7A (KT133, ISA)
- 2x Asus A7N8X
- 1x Asus A7N8X Deluxe
- 2x ASUS A7V880 Deluxe (KT880 - great chipset)
- 1x Gigabyte GA-7VA (KT400)
- 1x Gigabyte GA-7VRXP (KT333)
- 1x MSI K7N2 Delta
- 1x MSI K7N2 Delta2 Platinum (needs new EEPROM chip, haven't had time to mess around with it)
- 1x Shuttle AK35GT2 (KT333a - best KT333 board I've ever used)
- 1x ECS K7VTA3C rev 7 I think (KT333a)
- 1x FIC AD11 (AMD 760G + VIA 686B, AGP 4X + DDR266)

Here's some hw pr0n:

Out of all those boards the only ones in use are one AN7 (the one in the machine above - 3200+ 2333MHz version + FX 5900XT), one A7V880 Deluxe (3200+ 2200MHz + radeon 9800 PRO), the Suttle AK35GT2 (Voodoo 5 rig in the signature) and the ABIT KT7A (Athlon 1333 + Voodoo 3 3000). I plan to replace the KT7A with the FIC AD11 since it's a bit nippier overall, and I won't be overclokcing the already hot 1333MHz Athlon (although it does 1400MHz easily on the AN7)

On the CPU front I have:

Duron 700
Duron 750
Duron 850
Duron 900
Duron 1100
Duron 1400
Duron 1600
Duron 1800
Athlon 900
Athlon 1000
Athlon 1100
Athlon 1333
Athlon XP 1500+
Athlon XP 1600+
Athlon XP 1700+
Athlon XP 1800+
Athlon XP 1900+
Athlon XP 2100+
Athlon XP 2200+
4x Athlon XP 2400+ (Thoroughbred core, FSB 266, 256KB L2, 2000MHz)
Athlon XP 2400+ (Barton core, *half disabled L2 cache, 2000MHz, FSB 266)
2x Athlon XP 2500+
2x Athlon XP 2600+ (Thoroughbred core, 256KB L2, 2083MHz)
Athlon XP 2600+ (Barton Core, 1917MHz)
Ahlon XP 3000+ (FSB 400 model)
Athlon XP 3200+ (OEM FSB 333 model, 2333Mhz)
Athlon XP 3200+ (400MHz FSB model, 2200MHz)

Has anyone else used Lubuntu with there K7, the lubuntu site says the k7 has problems with flash http://lubuntu.net/

But in my opinion lubuntu runs really well on these processors, very well in fact, facebook is smooth saling and youtube is playable, more so then windows XP.

Oldskoolmaniac wrote:

Has anyone else used Lubuntu with there K7, the lubuntu site says the k7 has problems with flash http://lubuntu.net/

But in my opinion lubuntu runs really well on these processors, very well in fact, facebook is smooth saling and youtube is playable, more so the windows XP.

I use my machines for games, witch is difficult at best on linux. I stick to win98se or winXP depending on the machine's specs. I do run steam on my 3200+ systems and a few games from their library, like carmageddon 2, earth 2150 and so on. I've never actually tried to run a youtube clip on them as a matter of fact.

I once had Age of empires and a few other games running on wine, but I had a dual boot with xp and performance under lubuntu was pretty good.

My build at the time was a 2.2GHz 400FSB, 2GB RAM, dual 1TB western digital blue drives with the sata150 jumper in place and a Radeon HD2400 Pro (little overkill, but it was the only high end card that I had)

As for windows 7 on the k7 it ran pretty good, but don't try youtube with this OS. Internet explore is the only browser that runs smooth on win7 other browser will crawl like firefox will crash all the time, but oddly enough firefox runs well under xp and ubuntu.

kanecvr wrote:

Looks like us K7 fans had a similar PC history at one point :) […]
Show full quote

Looks like us K7 fans had a similar PC history at one point 😀

Nice boards guys. I'm (as you probably guessed) an Abit man myself. Not an asus fan by any stretch of the imagination, but the A7N8 nforce 2 series boards are very good. Even so, they don't hold a candle to the Abit AN7 or it's little brother the NF7. Reliability-wise they are pretty similar. Abit's late socket A series does not suffer from the capacitor plague - all my KT400 / KT600 and NF2 abit boards use quality japanese caps and work great. Both the A7N8X an the AN7 have the same issue unfortunately - crappy EEPROM chips... they will go bad w/o warning, but it's a really easy fix with either an eeprom burner or a hot-flash using another board and a new (or old working) chip.
...
...
In total I have:
- 2x ABIT AN7 one boxed, one in use in my socket A machine
- 1x ABIT NF7-S
- 1x ABIT KV7 (KT600)
- 1x ABIT KT7A (KT133, ISA)
- 2x Asus A7N8X
- 1x Asus A7N8X Deluxe
- 2x ASUS A7V880 Deluxe (KT880 - great chipset)
- 1x Gigabyte GA-7VA (KT400)
- 1x Gigabyte GA-7VRXP (KT333)
- 1x MSI K7N2 Delta
- 1x MSI K7N2 Delta2 Platinum (needs new EEPROM chip, haven't had time to mess around with it)
- 1x Shuttle AK35GT2 (KT333a - best KT333 board I've ever used)
- 1x ECS K7VTA3C rev 7 I think (KT333a)
- 1x FIC AD11 (AMD 760G + VIA 686B, AGP 4X + DDR266)

Here's some hw pr0n:

...
...

Nice stuff.

I have some of those motherboards but many of my boards have bad caps.

Here are some more pr0n:

This is the motherboard wich got me interested in revisiting Socket-A again, the Asus A7V KT133 (April/May 2000). I posted a picture of this board in the dumpster find thread before I tested it. I had much fun testing it and it turned out to work perfectly well. I already had two of these motherboards, also dumpster finds and now I have three.

Asus A7V rev 1.02 VIA KT133.

Another Asus A7V series board, this one uses VIA KT400 and I do not think I have tested it yet. This motherboard is also a dumpster find.

Asus A7V8X Gold-UAY rev 1.04 VIA KT400.

Here we have an Epox VIA KT400 board. I have no idea where I got it, it's probably a dumpster find.

Epox EP-8K9A7I VIA KT400.

Here is a motherboard I probably will not use as it needs a full recap. If it would have been a newer version I would probably have ordered caps right away but now I'm not sure this board would be any better than an Asus A7N8X. I will get caps for it some time in the future though when I need to order caps for some other motherboards. I have no clue where I got this motherboard but no one would throw a board like this in a dumpster?

DFI LanParty NFII Ultra REV_A Nvidia nForce2 Ultra (with bad caps).

That DFI board is awesome.

I'm making some progress.

Athlon XP systems mostly runs of the +5V rail, many modern PSUs have weak +5V rails. Digging through a pile of junk power supplies I found a 15+ year old Chieftec 350W PSU which seems decent. I don't remember the exact model name but I took a picture of the inside when I inspected the unit. The 12V line was kind of weak only capable of delivering 15 amps but it should be enough for a Radeon X1950 Pro 512.

The PSU

Don't you think it's a good idea to put a Radeon X1950 Pro 512 AGP card on an untested KT600 board with burn marks in one of the memory slots? Don't worry the X1950 Pro 512 is also untested and also has burn marks, on both it's molex jacks so it's fine! 😜

The system "assembled".

Smoke test! ...

And we have lift off! Literally, the CPU fan was set to 7000 rpm!

The POST screen.

Windows XP just finished installing. 😀

... I wouldn't run a X1950 on a measly 350w psu... even the X800 series needs a 400w psu in conjuncture with a good CPU and some add-in cards. I'd say look for a 400 or 450w PSU for the x1950 and 3200+

kanecvr wrote:

... I wouldn't run a X1950 on a measly 350w psu... even the X800 series needs a 400w psu in conjuncture with a good CPU and some add-in cards. I'd say look for a 400 or 450w PSU for the x1950 and 3200+

I think it will be fine.

A PCI-E Radeon X1950 pro 256MB needs about 80W. I guess the 512MB AGP version could be a bit more power hungry but not much and it will take some small amount of power from the 3.3V and 5V rails. 12V x 15A = 180W, there will be about 100W left over on the 12V rail for the HDD, CD-ROM reader and CPU fan.

This PSU can deliver 30A on the 3.3V rail and 30A on the 5V rail which I think is more important, in any case I will notice if it dosn't work.

Skyscraper wrote:

I think it will be fine. […]
Show full quote

kanecvr wrote:

... I wouldn't run a X1950 on a measly 350w psu... even the X800 series needs a 400w psu in conjuncture with a good CPU and some add-in cards. I'd say look for a 400 or 450w PSU for the x1950 and 3200+

I think it will be fine.

A PCI-E Radeon X1950 pro 256MB needs about 80W. I guess the 512MB AGP version could be a bit more power hungry but not much and it will take some small amount of power from the 3.3V and 5V rails. 12V x 15A = 180W, there will be about 100W left over on the 12V rail for the HDD, CD-ROM reader and CPU fan.

This PSU can deliver 30A on the 3.3V rail and 30A on the 5V rail which I think is more important, in any case I will notice if it dosn't work.

they say 80w, but my X1950 pulls up to 11A off the 12v molex connector - that's about 132-133w... that's coming pretty close to that theoretical 15A limit. The geforce 6800 has a similar power draw. Give it a go and see what happens. Even if you see magic smoke it should not damage anything but the PSU (in theory).

kanecvr wrote:

Skyscraper wrote:

I think it will be fine. […]
Show full quote

kanecvr wrote:

... I wouldn't run a X1950 on a measly 350w psu... even the X800 series needs a 400w psu in conjuncture with a good CPU and some add-in cards. I'd say look for a 400 or 450w PSU for the x1950 and 3200+

I think it will be fine.

A PCI-E Radeon X1950 pro 256MB needs about 80W. I guess the 512MB AGP version could be a bit more power hungry but not much and it will take some small amount of power from the 3.3V and 5V rails. 12V x 15A = 180W, there will be about 100W left over on the 12V rail for the HDD, CD-ROM reader and CPU fan.

This PSU can deliver 30A on the 3.3V rail and 30A on the 5V rail which I think is more important, in any case I will notice if it dosn't work.

they say 80w, but my X1950 pulls up to 11A off the 12v molex connector - that's about 132-133w... that's coming pretty close to that theoretical 15A limit. The geforce 6800 has a similar power draw. Give it a go and see what happens. Even if you see magic smoke it should not damage anything but the PSU (in theory).

So far so good! 😀

I'm done with installing all benchmarks and I'm now testing to see that everything runs as it should before I go on a hunt in my box with socket A CPUs for a Duron with Morgan core. 😀

The video card driver ended up beeing Catalyst 9.3 as the 10.2 driver seemes more targeted at HD3x0 and HD4XX0 cards.

I am doing my "test" benchmark runs with a Thoroughbred Athlon XP.

3dmark2000, 3dmark2001, 3dmark03 and 3dmark05 runs fine but 3dmark05 needs the registered version otherwise the scores aren't shown offline so I will not include it in the benchmark template. Doom 3 (1.0) runs fine but I still need to test Quake, Quake II, Quake III (32bit) and Unreal Gold (32bit). All games tested will be tested at 640*480 as it's the CPU we are testing, Doom 3 (1.0) is set to the "medium template".

Other benchmarks are SuperPi 1M, PCMark2002 (CPU and memory) and 7-Zip (32MB).

The testsystem for the basic CPU scaling tests if someone wants to compare scores or add CPUs I do not have or can mimic using unlocked CPUs with the same core.

K7 motherboard with VIA chipset. KX/KT133 for Slot A, KT266 or better for Socket A.
A K7 CPU suitable for the motherboard running at stock speed*
Enough memory to not run out of memory! 😁 DDR single channel for Socket A and normal SDRAM for Slot A. At 100 mhz FSB the memory should be set to 133 MHZ otherwise 1:1 ( CL2-2-2-X ideally).
Radeon X1950 Pro 256MB or 512MB using Catalyst 9.3 and ATI Tray Tools to disable vsync.
Creative Live!, Audigy or Audigy2.
Some kind of storage device.
Windows XP SP3. Killing processes is fine as it does not affect benchmark scores much but saves on memory.

*Using unlocked CPUs to mimic other CPUs with the same core is fine as is using a Barton 2500+ (11x166) as a Barton 3200+ (11x200). Overclocking the motherboard FSB from 166 to 200MHz on a KT400 motherboard or from 133 to 166 MHz on a KT266/KT333 motherboard is only fine if the motherboard has the correct divider as VIA chipsets do not have a PCI lock.

Benchmarks results of all kinds and with all kinds of K7 hardware are welcome but for benchmarking the K7 CPU performance scaling from 1999 to 2003 the hardware needs to be somewhat consistent.

Edit

Lets skip Quake 1, GLQuake and my X1950 Pro are not best friends. It's probably something trival but it runs at a million FPS with all K7 CPUs anyhow. Quake II, Quake III and Unreal Gold runs perfectly in Windows XP out of the box so to speak.

Now I just need to find a Morgan core Duron, unlock it if necessary and start benching.

/Edit

Edit2

WFT, this upper end ECS Photon AF1 VIA KT600 motherboard do not have mutiplier control in the BIOS (or with jumpers), the same effing brain fart as with the Asus A7M266. Perhaps I should have checked that before choosing this motherboard. It's possible to change the default multiplier using the bridges on the CPU but when I have the heatsink off I might aswell change CPU, I should have most of the Morgan Durons I think.

I will have to order a nice Thoroughbred Athlon XP-M as it should be possible to change mutiplier using software on KT600 motherboards with those, that will take care of all Thoroughbred Athlon XP models. I have a boat load of Thoroughbred CPUs but as there are 20 or so speed grades I would go insane if I actually had to change CPU or at least remove the heatsink for every model.

/Edit2

Yeah, you can't change multipliers on most KT400 / KT600 boards. They were considered by manufacturers as "budget chipsets". Not even my KV7 has a multiplier setting in bios - or my NF7-S witch is a higher end nforce board. The only boards that I know of that support multiplier changes are:

- Abit AN7
- Asus A7N8X Deluxe (regular version and early revisions DO NOT support multiplier changes)
- ASUS A7V880 Deluxe
- MSI K7N2 Delta
- MSI K7N2 Delta2 Platinum
- Shuttle AK35GT2

These are all NF2 boards, except for the A7V880 witch uses the VIA KT880 chipset, and the shuttle AK35GT2 witch is a KT333a. The rest of my boards (see list above) do not support multiplier overclocking.

If you are trying testing the cpu under unreal couldn't you select software as the render so it would be pure cpu.

Skyscraper wrote:

The video card driver ended up beeing Catalyst 9.3 as the 10.2 driver seemes more targeted at HD3x0 and HD4XX0 cards.

I am doing my "test" benchmark runs with a Thoroughbred Athlon XP.

3dmark2000, 3dmark2001, 3dmark03 and 3dmark05 runs fine but 3dmark05 needs the registered version otherwise the scores aren't shown offline so I will not include it in the benchmark template. Doom 3 (1.0) runs fine but I still need to test Quake, Quake II, Quake III (32bit) and Unreal Gold (32bit). All games tested will be tested at 640*480 as it's the CPU we are testing, Doom 3 (1.0) is set to the "medium template".

In my AGP tests, I used Cat 9.1 for DX9 and Cat 10.2 for DX10 cards (9.1 was compatible enough for all games/programs I used).

[Bolded part]How is that a problem ?
Source : LINK ?

I would try Half Life 2 : Lost Coast or FEAR (both have build-in benchmarks).

kanecvr wrote:

Yeah, you can't change multipliers on most KT400 / KT600 boards. They were considered by manufacturers as "budget chipsets". Not […]
Show full quote

Yeah, you can't change multipliers on most KT400 / KT600 boards. They were considered by manufacturers as "budget chipsets". Not even my KV7 has a multiplier setting in bios - or my NF7-S witch is a higher end nforce board. The only boards that I know of that support multiplier changes are:

- Abit AN7
- Asus A7N8X Deluxe (regular version and early revisions DO NOT support multiplier changes)
- ASUS A7V880 Deluxe
- MSI K7N2 Delta
- MSI K7N2 Delta2 Platinum
- Shuttle AK35GT2

These are all NF2 boards, except for the A7V880 witch is a VIA KT880 board. The rest of my boards (see list above) do not support multiplier overclocking.

That sucks, you learn something new every day. My only experiences with newer Socket A stuff have been with Nforce2, it seems I have not missed much of value... I will be on the lookout for an ASUS A7V880 Deluxe motherboard, KT880 boards seem to perform well in the reviews.

It seems I will have to juggle CPUs.

agent_x007 wrote:

In my AGP tests, I used Cat 9.1 for DX9 and Cat 10.2 for DX10 cards (9.1 was compatible enough for all games/programs I used). […]
Show full quote

Skyscraper wrote:

The video card driver ended up beeing Catalyst 9.3 as the 10.2 driver seemes more targeted at HD3x0 and HD4XX0 cards.

I am doing my "test" benchmark runs with a Thoroughbred Athlon XP.

3dmark2000, 3dmark2001, 3dmark03 and 3dmark05 runs fine but 3dmark05 needs the registered version otherwise the scores aren't shown offline so I will not include it in the benchmark template. Doom 3 (1.0) runs fine but I still need to test Quake, Quake II, Quake III (32bit) and Unreal Gold (32bit). All games tested will be tested at 640*480 as it's the CPU we are testing, Doom 3 (1.0) is set to the "medium template".

In my AGP tests, I used Cat 9.1 for DX9 and Cat 10.2 for DX10 cards (9.1 was compatible enough for all games/programs I used).

[Bolded part]How is that a problem ?
Source : LINK ?

I would try Half-Life 2 : Lost Coast or FEAR (both have build-in benchmarks).

I had no idea they were giving out keys for free now, I keep up appearances... 😀 Well then 3dmark05 can be included I guess. My USB stick with 3dmark99 to 3dmark06 and some other old benchmarks has remained more or less unchanged the last 7 or 8 years or so.

I'm not sure I own Half-Life 2: Lost Coast but perhaps benching FEAR is a good idea, I will install it and se how it runs. [Edit] I looked up HL Lost Cost, it seems to be a Steam exclusive? I do not use Steam. [/Edit]

Oldskoolmaniac wrote:

If you are trying testing the cpu under unreal couldn't you select software as the render so it would be pure cpu.

Yes I suppose I could but I doubt the X1950 Pro is bottlenecking much at 640*480. Video card driver overhead exists in the real world so it could just aswell be included in the game benchmarks.

I'm not sure the motherboard designers from MSI that started working for ECS and designed the Photon boards were that good.

The ECS Photon AF1 is a board with some quirks so to say. If I disable the onboard bloat I don't need, lite network card and com ports I lose the Sound Blaster Live! aswell. Even if I disable the onboard devices before I insert the Soundblaster Live! it refuses to work. The only way to get the Live! working again is resetting the BIOS, and then just live with the extra clutter or disable all the extra crap in the Device Manager.

I'd love to hear how you go with unlocking.

I do have a KT600 board and it does have multiplier options in the BIOS, but my pencil trick attempts didn't do anything. So I also just got a few CPUs (not every single model) as they are quite cheap, but it would be indeed much easier to just select the multiplier 😀

F.E.A.R. has some input device bug that makes it inconsistent with benchmarks. Especially with any Logitech devices.

I believe a Steam user fixed this issue and it's now found on the PC Gaming Wiki: http://pcgamingwiki.com/wiki/F.E.A.R.#Low_frame_rate

Even without any Logitech devices, without this fix you can get inconsistent and lower benchmark results.

Other benchmarks I like using on XP machines:

- Doom 3
- Far Cry (There is a benchmark tool floating around)
- HL2 Lost Coast
- X2 The Threat
- Serious Sam Second Encounter

The pencil trick doesn't work with Bartons and Thoroughbreds. Graphite isn't able to bridge the pits made by AMD when they laser trimmed the switches on the CPU package. Generally, late model socket A unlocking usually requires filling material like superglue or epoxy then a dab of conductive pen ink.

gdjacobs wrote:

The pencil trick doesn't work with Bartons and Thoroughbreds. Graphite isn't able to bridge the pits made by AMD when they laser trimmed the switches on the CPU package. Generally, late model socket A unlocking usually requires filling material like superglue or epoxy then a dab of conductive pen ink.

I can confirm this. But if you're lucky enough, you can come across unlocked desktop bartons like I did.

I'm mostly interested in the early ones. Spitfire, Morgan and Thunderbird cores.