A little quirky, but very good feature list and OK when you find the “sweet spot”. – Joe
SUMMARY: Different from the rev 2.0 board, but not yet ready for Prime Time?
Iwill released the third version of the XP333 – rev 2.1. This is a little different from the 2.0 version (I noticed a different capacitor placement) and the BIOS for the 2.1 is not meant for the 2.0. I have been using this board for a while to test out its various features.
Before I get into benchmarks, I want to comment on this board’s chameleon-like character and the reason for the question mark in the SUMMARY.
When I first started to exercise this board at Asynch RAM settings, it was flaky. I had artifacts on screen and it would routinely freeze in Quake and 3DMark 2001, even at 133 MHz. I went on to some other things for a few days and returned. Now it runs everything OK. I can’t explain it – it’s almost as if it “matured” while sitting unused; a puzzle.
Running the XP333 at various RAM settings was “interesting”.
One thing I found with was that changing RAM settings, or sometimes just rebooting from Windows, was an exercise in patience. From my notes:
- Set to ASYNCH 33/166, SPD/SPD, OCZ DDR333
- Benchmarks OK; reboot
- No boot, hit multiple restarts, resets to FSB 100
- BIOS change SPD to Fast, 133/166, multiplier 11
- No boot, no restart, Pull plug, reboot
- Resets to 100 FSB, SPD/SPD, all AUTO
- BIOS change 133/166, multiplier 11
- reboot OK, change to FAST
- No boot, hit multiple resets, resets to FSB 100 etc.
- BIOS change DRAM CAS 2/2
- Boots OK, run benchmarks
- Shutdown to reset, no boot, multiple resets to defaults
- BIOS change SPD to Ultra
- No boot, no resets, catatonic – board does not power up
- Wait 5-10 minutes w/o power plug, reboots
I think this has as much to do with “second tier” chipsets than just Iwill – I’ve had some of the same behavior with Shuttle’s AK31.
I did learn one thing from this – if the XP333 does not boot, hitting multiple resets will reset the BIOS to FSB 100, all auto. If you hit reset once, you MAY get a reboot at the old settings. Any RAM setting beyond FAST did not work – the board went catatonic and took 5-10 minutes to recover.
Hence the question mark: I can’t tell you unequivocally that what you’ll get will be the Jekyl Iwill or the Hyde Iwill, or that Jekyl will turn into Hyde. I can only tell you that I saw two sides of the same board, and you might also.
Asynchronous RAM Performance
Iwill just released a BIOS (1/18/2002) which allows asynchronous RAM speeds, so I tested it:
XP @ 1467 MHz, 11 x 133, CAS 2/2
Benchmark | FSB, RAM @ 133 | FSB 133, RAM 166 | % Diff |
SiSandra CPU | 4034/2021 | 4037/2023 | NA |
SiSandra MM | 8003/9244 | 8010/9230 | NA |
SiSandra Memory | 1862/1722 | 1719/1570 | -7.7/-8.8% |
Quake | 173.2/167.1 | 154.2/150.5 | -11.0/-10.0% |
3D Mark 2001 | 4489 | 4201 | -6.4% |
OCZ DDR333, SiSandra 2002.
Now, let me get this straight: So, running at ASYNCH reduces performance? Am I missing something here? Iwill recognizes this and freely admits that the ASYNCH BIOS has some bugs, which they are working to correct; However, unless there’s a significant performance improvement over SYNCH RAM settings, seems like it’s more hoopla than anything else.
Now this is not to say that the XP333 can’t run at high FSBs – it sure can. Here’s one example:
XP, 9 x 175 vs 11 x 133, RAM in SYNCH, CAS 2/2
Benchmark | 9 x 175 | 11 x 133 | % Diff |
SiSandra CPU | 4346/2178 | 4034/2021 | 7.7/7.8% |
SiSandra MM | 8622/9939 | 8003/9244 | 7.7/7.5% |
SiSandra Memory | 2265/2089 | 1862/1722 | 21.6/21.3 % |
Quake | 196.4/187.8 | 173.2/167.1 | 13.4/12.4% |
3D Mark 2001 | 4746 | 4489 | 5.7% |
NOTE: You can only run ASYNCH at 133/166; OCZ DDR333, SiSandra 2002.
I tried to do a “head-to-head” comparison – 11 x 133 and 9 x 163. What I found was that running the Iwill from 160 to 165 was not stable – either once in Windows I could not open anything, or it would freeze in Quake. Running “odd” settings was an adventure.
I earlier did a series of benchmarks with a T-Bird and the earlier 12/13 BIOS:
Synchronous RAM Performance
First I compared the Iwill against the Shuttle AK31; each board was populated with 256MB of Crucial DDR266, T-Bird 1400, Leadtek GeForce 2 Pro 64 MB and WD hard drives. Each used the same power supply (350 watts).
T-Bird @ 9.5 x 150
Benchmark | Iwill XP 333 2.1 | Shuttle AK31 | % Diff |
SiSandra CPU | 3936/1957 | 3940/1959 | NA/NA% |
SiSandra MM | 7807/9674 | 7815/9683 | NA/NA% |
SiSandra Memory | 1662/1610 | 2081/2005 | -20/-20 % |
Quake | 141.7/139.6 | 179.8/176.7 | -21/-21 % |
3D Mark 2001 | 4068 | 4668 | -13 % |
Crucial DDR266, SiSandra 2002; Best stable RAM settings: Shuttle 2522, Iwill Normal, 2.5/3.
The Crucial DDR 2.5 will not go much above 160, so to push the XP333, I used some OCZ “Performance” DDR333. The OCZ seems to do well with the Iwill but absolutely does not like the Shuttle (or vice versa) – I could not get it to work with the Shuttle at 9 x 150, let alone higher than that. However, the XP333 really liked this stuff, as shown below:
Iwill XP333 2.1 – T-Bird 1400
Benchmark | 8.5 x 166 | 8.5 x 175 | 8.5 x 180 |
SiSandra CPU | 3885/1932 | 4104/2041 | 4221/2099 |
SiSandra MM | 7705/9549 | 8142/10089 | 8374/10378 |
SiSandra Memory | 1850/1773 | 1955/1871 | 2003/1921 |
Quake | 149.1/143.9 | 156.8/151.5 | 160.8/156.2 |
3D Mark 2001 | 4223 | 4328 | 4380 |
OCZ DDR333, SiSandra 2002; Best stable RAM settings: SPD/SPD; PCI AUTO.
Running at 180 was really on the edge – I saw snow on the screen while running 3DMark 2001, so I can’t tell you that this was rock stable by any means. However, it was fun to see a board running at 180.
PCI Divisors
I tested different PCI speed settings with the 2.0 version (HERE). The net effect was to reduce performance if PCIs were set to keep them in spec at high FSBs.
I ran these tests after I updated my UXD Diagnostic card – see PCI Divisor Tests HERE.
CONCLUSIONS
My criteria is pretty simple on testing any product: How it works out of the box. Any product that has to be modified in some manner to work well is good to know, but for most users, irrelevant. When I buy anything, I expect it to work as advertised.
Sample of one testing is always an iffy proposition – what I found and the problems I had may not be representative of what others find. However, what I experienced is what I’m reporting to you.
First, I think the XP333 is hampered by the ALI chipset – it’s simply not competitive with others, and all the tweaking in the world is not going to get around it.
Second, having a lot of RAM and FSB settings and not being able to use all of them is not a positive. This may change as DDR333 and XP333 BIOSs mature, but right now it’s an event waiting to happen.
Third, I am hard pressed to see an advantage to running RAM asynchronously. I may be wrong, but unless Iwill pulls a rabbit out of the hat on this one, I don’t see where this is a plus.
Fourth, depending on which board shows up for work, you may love it or hate it. With the right RAM and settings which may be particular to your board, the XP333 may be the best you’ve had – and it may not. I must say that increasingly, this comment applies to about every board we’ve had and extensively used at Overclockers.com
Iwill has given us some great products and some not so great. I would consider the XP333 a motherboard with a lot of great features, performance that is competitive, but not top rank, and very stable when you find its “sweet spot”.
Now I know some of you will email me and say “But this website said….” I can’t vouch for any other site – I just tell you what I found and that’s it.
Iwill XP333 – PCI Divisors on p 3…
Email Joe
SUMMARY: Different from the rev 2.0 board, but not yet ready for Prime Time?
Iwill released the third version of the XP333 – rev 2.1. This is a little different from the 2.0 version (I noticed a different capacitor placement) and the BIOS for the 2.1 is not meant for the 2.0. I have been using this board for a while to test out its various features.
Before I get into benchmarks, I want to comment on this board’s chameleon-like character and the reason for the question mark in the SUMMARY.
When I first started to exercise this board at Asynch RAM settings, it was flaky. I had artifacts on screen and it would routinely freeze in Quake and 3DMark 2001, even at 133 MHz. I went on to some other things for a few days and returned. Now it runs everything OK. I can’t explain it – it’s almost as if it “matured” while sitting unused; a puzzle.
Running the XP333 at various RAM settings was “interesting”.
One thing I found with was that changing RAM settings, or sometimes just rebooting from Windows, was an exercise in patience. From my notes:
- Set to ASYNCH 33/166, SPD/SPD, OCZ DDR333
- Benchmarks OK; reboot
- No boot, hit multiple restarts, resets to FSB 100
- BIOS change SPD to Fast, 133/166, multiplier 11
- No boot, no restart, Pull plug, reboot
- Resets to 100 FSB, SPD/SPD, all AUTO
- BIOS change 133/166, multiplier 11
- reboot OK, change to FAST
- No boot, hit multiple resets, resets to FSB 100 etc.
- BIOS change DRAM CAS 2/2
- Boots OK, run benchmarks
- Shutdown to reset, no boot, multiple resets to defaults
- BIOS change SPD to Ultra
- No boot, no resets, catatonic – board does not power up
- Wait 5-10 minutes w/o power plug, reboots
I think this has as much to do with “second tier” chipsets than just Iwill – I’ve had some of the same behavior with Shuttle’s AK31.
I did learn one thing from this – if the XP333 does not boot, hitting multiple resets will reset the BIOS to FSB 100, all auto. If you hit reset once, you MAY get a reboot at the old settings. Any RAM setting beyond FAST did not work – the board went catatonic and took 5-10 minutes to recover.
Hence the question mark: I can’t tell you unequivocally that what you’ll get will be the Jekyl Iwill or the Hyde Iwill, or that Jekyl will turn into Hyde. I can only tell you that I saw two sides of the same board, and you might also.
Asynchronous RAM Performance
Iwill just released a BIOS (1/18/2002) which allows asynchronous RAM speeds, so I tested it:
XP @ 1467 MHz, 11 x 133, CAS 2/2
Benchmark | FSB, RAM @ 133 | FSB 133, RAM 166 | % Diff |
SiSandra CPU | 4034/2021 | 4037/2023 | NA |
SiSandra MM | 8003/9244 | 8010/9230 | NA |
SiSandra Memory | 1862/1722 | 1719/1570 | -7.7/-8.8% |
Quake | 173.2/167.1 | 154.2/150.5 | -11.0/-10.0% |
3D Mark 2001 | 4489 | 4201 | -6.4% |
OCZ DDR333, SiSandra 2002.
Now, let me get this straight: So, running at ASYNCH reduces performance? Am I missing something here? Iwill recognizes this and freely admits that the ASYNCH BIOS has some bugs, which they are working to correct; However, unless there’s a significant performance improvement over SYNCH RAM settings, seems like it’s more hoopla than anything else.
Now this is not to say that the XP333 can’t run at high FSBs – it sure can. Here’s one example:
XP, 9 x 175 vs 11 x 133, RAM in SYNCH, CAS 2/2
Benchmark | 9 x 175 | 11 x 133 | % Diff |
SiSandra CPU | 4346/2178 | 4034/2021 | 7.7/7.8% |
SiSandra MM | 8622/9939 | 8003/9244 | 7.7/7.5% |
SiSandra Memory | 2265/2089 | 1862/1722 | 21.6/21.3 % |
Quake | 196.4/187.8 | 173.2/167.1 | 13.4/12.4% |
3D Mark 2001 | 4746 | 4489 | 5.7% |
NOTE: You can only run ASYNCH at 133/166; OCZ DDR333, SiSandra 2002.
I tried to do a “head-to-head” comparison – 11 x 133 and 9 x 163. What I found was that running the Iwill from 160 to 165 was not stable – either once in Windows I could not open anything, or it would freeze in Quake. Running “odd” settings was an adventure.
I earlier did a series of benchmarks with a T-Bird and the earlier 12/13 BIOS:
Synchronous RAM Performance
First I compared the Iwill against the Shuttle AK31; each board was populated with 256MB of Crucial DDR266, T-Bird 1400, Leadtek GeForce 2 Pro 64 MB and WD hard drives. Each used the same power supply (350 watts).
T-Bird @ 9.5 x 150
Benchmark | Iwill XP 333 2.1 | Shuttle AK31 | % Diff |
SiSandra CPU | 3936/1957 | 3940/1959 | NA/NA% |
SiSandra MM | 7807/9674 | 7815/9683 | NA/NA% |
SiSandra Memory | 1662/1610 | 2081/2005 | -20/-20 % |
Quake | 141.7/139.6 | 179.8/176.7 | -21/-21 % |
3D Mark 2001 | 4068 | 4668 | -13 % |
Crucial DDR266, SiSandra 2002; Best stable RAM settings: Shuttle 2522, Iwill Normal, 2.5/3.
The Crucial DDR 2.5 will not go much above 160, so to push the XP333, I used some OCZ “Performance” DDR333. The OCZ seems to do well with the Iwill but absolutely does not like the Shuttle (or vice versa) – I could not get it to work with the Shuttle at 9 x 150, let alone higher than that. However, the XP333 really liked this stuff, as shown below:
Iwill XP333 2.1 – T-Bird 1400
Benchmark | 8.5 x 166 | 8.5 x 175 | 8.5 x 180 |
SiSandra CPU | 3885/1932 | 4104/2041 | 4221/2099 |
SiSandra MM | 7705/9549 | 8142/10089 | 8374/10378 |
SiSandra Memory | 1850/1773 | 1955/1871 | 2003/1921 |
Quake | 149.1/143.9 | 156.8/151.5 | 160.8/156.2 |
3D Mark 2001 | 4223 | 4328 | 4380 |
OCZ DDR333, SiSandra 2002; Best stable RAM settings: SPD/SPD; PCI AUTO.
Running at 180 was really on the edge – I saw snow on the screen while running 3DMark 2001, so I can’t tell you that this was rock stable by any means. However, it was fun to see a board running at 180.
PCI Divisors
I tested different PCI speed settings with the 2.0 version (HERE). The net effect was to reduce performance if PCIs were set to keep them in spec at high FSBs.
I ran these tests after I updated my UXD Diagnostic card – see PCI Divisor Tests HERE.
CONCLUSIONS
My criteria is pretty simple on testing any product: How it works out of the box. Any product that has to be modified in some manner to work well is good to know, but for most users, irrelevant. When I buy anything, I expect it to work as advertised.
Sample of one testing is always an iffy proposition – what I found and the problems I had may not be representative of what others find. However, what I experienced is what I’m reporting to you.
First, I think the XP333 is hampered by the ALI chipset – it’s simply not competitive with others, and all the tweaking in the world is not going to get around it.
Second, having a lot of RAM and FSB settings and not being able to use all of them is not a positive. This may change as DDR333 and XP333 BIOSs mature, but right now it’s an event waiting to happen.
Third, I am hard pressed to see an advantage to running RAM asynchronously. I may be wrong, but unless Iwill pulls a rabbit out of the hat on this one, I don’t see where this is a plus.
Fourth, depending on which board shows up for work, you may love it or hate it. With the right RAM and settings which may be particular to your board, the XP333 may be the best you’ve had – and it may not. I must say that increasingly, this comment applies to about every board we’ve had and extensively used at Overclockers.com
Iwill has given us some great products and some not so great. I would consider the XP333 a motherboard with a lot of great features, performance that is competitive, but not top rank, and very stable when you find its “sweet spot”.
Now I know some of you will email me and say “But this website said….” I can’t vouch for any other site – I just tell you what I found and that’s it.
Iwill XP333 – PCI Divisors on p 3…
Email Joe
I tested different PCI speed settings with the 2.0 version (HERE). The net effect was to reduce performance if PCIs were set to keep them in spec at high FSBs.
What I could not do was find out exactly at what speed the PCI bus was running. This required an update to my UXD diagnostics card, which I recently received, so I attempted to run the board as various settings to see how fast the bus was running. All runs were with OCZ DDR333. AUTO is the BIOS setting to automatically set PCI divisors, and “/5” indicates a manual PCI divisor setting.
Iwill XP333 2.1 ASNYCHRONOUS BIOS
FSB/PCI Divisor | PCI MHz |
133/Auto | 33.31 |
133/166 | 33.31 |
135/Auto | 33.78 |
133/Auto | 33.31 |
140/Auto | No Boot |
140/Auto | No Boot |
150/Auto | No Boot |
166/4 | 41.58 |
180/Auto | 30.07 |
There were a number of FSBs that simply would not work – sometimes it would take 3 boots to get a reading. The upshot of all this is that running the XP333 2.1 at AUTO settings is the safest way to get as many different speeds as possible. Also note that at the AUTO settings, the PCI bus will be in spec at 166 MHz and probably at 200 MHz as well.
The Asychronous BIOS is buggy and I went back to the last BIOS released – December 13, 2001.
Iwill XP333 2.1 12/13 BIOS
FSB/PCI Divisor | PCI MHz | SiSoft MEM | Quake |
10×132/Auto | 44.07 | NA | NA |
8×166/Auto | 33.22 | 1932/1773 | Freeze |
8×166/4 | 41.58 | 1889/1750 | Freeze |
7×170/Auto | 34.07 | NA | NA |
7×170/6 | 28.4 | NA | NA |
7×175/6 | 29.27 | NA | NA |
7×180/5 | No Read | 2072/1892 | Freeze |
7×180/6 | 30.07 | NA | NA |
7×180/Auto | 30.04 | 2072/1893 | Freeze |
8×180/Auto | 30.04 | 2105/1931 | Freeze |
I think one thing you notice is that “stable” is something I did not see too often. Notice that from 100 to 132 MHz, the AUTO setting appears to be set at 3 and switches to 4 at 133 – no surprise.
However, at settings approaching 180 MHz, it looks like the AUTO setting switches someplace between 170 and 180 MHz from 5 to 6. I could not pin it down because the UXD card would not boot at many of these settings.
CONCLUSIONS
Iwill’s XP333 rev 2.1 does give users a lot of options. The PCI divisor switch works well, although running way out of spec (e.g., 133/2) does not appear possible. Finding a stable setting looks to be something of a challenge. How these options translate into stable settings at high performance levels is something each user must determine.
Another Look, by Default, on page 4…
Ed Note: Every once in a while, Default will grace our pages with second looks at motherboards; we’ll call it
DEFAULT’S DEMENTIA
A while back Joe asked me if I would be interested in giving an opinion on several motherboards. We exchanged a few emails and kicked it around a bit. Joe said he got the feeling I didn’t want to work for anyone or have to respond in a timely manner.
This is very true. I enjoy a unique status. I don’t depend on anyone for anything. Therefore no one can influence me. If I want to spend weeks kicking a board around no deadlines, no problem.
I came up with the idea “Second Opinion”. Occasionally if Joe has an interesting board after he beats up on it, I get the remains for a second opinion. Without a doubt, the XP333 fits this category.
First Things First
I got the board on a Friday afternoon and, as usual, I resisted the temptation to stick it in a case and see what happens. I always spend some time going through the manual several times. There is also a fold out installation guide. The first time DIYer should find this a valuable aid. It’s obvious IWill designed the XP333 for the DIYer and PC Enthusiast.
The one thing I really liked were the jumpers. For a brief moment I thought about replacing them with old jumpers when I send the board back. IWill also included a little packet with three spare jumpers in it. I figured I might be able to get away with pilfering them, but my conscience overuled me. If you have changed jumpers as much as I have, you will definitely like these jumpers.
I spent some time rereading reviews which yielded very little (other than Joe’s) of what I wanted to know.
The IWill forums yielded a wealth of information. As usual, some loved the board and some hated it. What piqued my interest was some had sucess with one BIOS and others had success with another. Normally there is a bit of this, but I found a lot of it. There were also a lot of problems clearing CMOS. Some had no problem running a high FSB and others could barely get off the ground.
Next Things Next
Two days later, I decided to approach this like I would a development project when I was working as a Lab Rat. Pick the objective and eliminate variables that could influence the outcome. The objective was to investigate the FSB capabilities and PCI dividers. I used a 200FSB T-Bird at or around 1GHz. I didn’t want an overclocked CPU.
If I couldn’t achieve a high FSB, I didn’t want it to be because the CPU gave out first. For a video card, I used an older 32Mb GTS I knew could handle an AGP bus speed of 80 from another board I had been working on. The only variables I had to deal with now were the BIOS and the RAM.
Joe sent me the board with the Dec. BIOS. Some users tried the Jan. BIOS and went back to the Dec. BIOS and visa versa. After about a day and a half of getting nowhere special, I tried the Jan. 18th BIOS and found it worked better. Another day and I got onto the board and things started to happen.
Does this get your attention??
This is with a 191FSB setting in the BIOS. I began having delusions of grandeur. Would it be possible to hit the 200 mark? I couldn’t do it, but I think it’s just a matter of time before someone does. All it’s going to take is the right stick of RAM and a serious, advanced user.
Now is when I start to get the emails. What BIOS settings did you use? While this may work sometimes, it’s not going to work with this board. The BIOS settings will depend on what RAM you use. Now I get emails what RAM did you use? That’s not going to work either, because two sticks of the same brand can perform differently.
I recently read a review of ram where Crucial PC2100 did 188 FSB and Corsair PC2400 did 187 FSB and OCZ PC3000 did 188 FSB. I used Crucial PC2100, Corsair PC2400, and Kingmax PC2700. I hit the high with the Corsair.
To Async Or Not To Async
My findings were the same as Joes. There was a drop off in performance with Async. All three brands of RAM worked in Asyncronous mode. If I can believe what I read, IWill is working on this. Time will tell.
PCI Dividers
I have no way to acurrately measure them, but it was obvious they worked. After screwing with them, I found the best option was to leave the BIOS on Auto. There were times when 3DMark would drop off and then pick back up again as I worked my way up in FSB with the same memory settings. This indicated a change in the AGP bus speed, so I knew they worked.
Getting into BIOS and Clearing CMOS
My experience was about the same as Joes. I couldn’t find any one thing that worked consistently. There were times when I could just hit reset and get back into BIOS. There were times when I could just hold the insert key and hit reset to get back into BIOS. There were times when multiple hits of reset would get me back into BIOS. Sometimes, just shutting down power and starting up again would get me into BIOS.
There were times when I had to do all of the above in random sequence and eventually one would work.
Clearing CMOS was another adventure. There were times when I could just power down, move the jumper, hold it for a few seconds and I was back in business. There were times when I had to pull the power plug, wait five or ten minutes and I would get back in. There were also times when I had to repeat the sequence several times for it to work.
One time was particularly disconcerting. The board would start to post and then hangup. There was no way to get into BIOS. I remembered reading a post in a forum about the BIOS getting corrupted and having to get a new EPROM.
I wrote it off as just being a screwed up BIOS and not using the right combination to clear CMOS. For a few moments, I found myself wondering if BadFlash.com had an overnight delivery option for a new EPROM, but I got it cleared eventually after doing all the above in various random combinations.
Eventually I found what worked the best most of the time was shutting off the power, holding the insert key and power back up. When it got stubborn, isolating from the AC line, waiting five or ten minutes, using the jumper to clear CMOS and then powering up worked.
Last Words
I used the board four days for surfing and gaming without a single crash. What amazed me about this board is although I could not get a tick above 191FSB, I could loop 3D Mark and do gaming at this FSB without a hint of instability.
I did not resort to any third party programs to tweak the BIOS. I just slowly and methodically worked my way through the BIOS until I found the right combination.
No benchmarks were included because this was not the objective of the article. This was not a review per se. Also benchmarks with a 1GHz T-Bird and a 32Mb GTS could be misleading to some. I will (no pun intended) say the numbers I did get were good for the combination I was using.
IWill said this was a board designed for the Enthusiast and this is what you get. This is not a rockem, plug and play, point and click out of the box solution.
This is a board for the serious, advanced user and perhaps the wannabe hardcore looking to learn. If you feel you have the skills, or you are up to the challenge, this is the board for you.
ADVANCED TWEAKING CONTINUED page 5…
Disclaimer:
I do not work for anyone or get paid for anything I write. The XP333 2.1 was supplied by Joe Citarella at great risk to Overclockers.com. Everything I write in articles or say in emails is Public Domain for the reader to do with as they wish.
Default, aka John R. Abaray
DEFAULT’S DEMENTIA
First a special thanks to Doug Panting for the links and information he provided. From the emails I received, some of the forum hardcore felt no one had shown what the board could really do. Joe suggested I keep the board another month to see what I could do with it and do a revisited article, so it fit right in.
I waded through the various tweaks, BIOS hacks, and mobo mods sent to me to see what worked and what didn’t – again at risk to OC.com. The last time I put that line in my disclaimer I was joking. This time, I’m serious. There is always the chance you can smoke your board or get a bad BIOS flash and then it’s BOO HOO time.
At this point, I was really into the board and figured if something happened, I’d just buy him a new one.
Risky Business
The usual disclaimer: The Author assumes no responsibility. If you try some of these mods or tweaks and screwup your board and blah blah, blah blah, blah blah. The usual “YOU HAVE BEEN WARNED”.
Credits
I’ve seen some of these mods popup in other posts with no reference to where they found it. I’ve also seen another site present the Vmods when it’s obvious they are not the originator. It’s impossible to know the originator, but I give credit to where I found them. It’s called courtesy, something that seems to be disappearing in today’s world.
Vmem
This is where it starts. I said in my last article, I felt RAM was holding the board back. This proved to be true. If you move both jp10 jumpers to the right (5-6/2-3), you can get a tenth more to 2.8v – found in a post by lallend. A post on 2/18/2002 by Biggins (also goes by Bigsy) listed a number of mods. You can check out his site HERE.
With a resistor from jp10 5-6 to ground, you can get enough juice to melt your RAM if so desired. This is a no-solder mod – just stick one end into the jumper.
Values listed by Bigsy are:
- 1.2k = 3v
- 1 k = 3.04v
- 820 = 3.1v
- 680 = 3.15v
- 560 = 3.2v
- 470 = 3.29v
- 290 =3.24
Just remember, resistors have a tolerence and can be more or less. I checked my resistors with a VOM and with a 1k resistor, I got to 193FSB. With a 560, I got to 197. With a 500, I got to 200FSB in BIOS which was 199.74 in WCPUID. I was hesitant to put anymore juice to the RAM, but I wanted to hit 200FSB/DDR400.
Northbridge
My home is setup for solar heat and at times, if I don’t pay attention, it can get quite toasty. When the ambient got to 26C, I would have to back off on the FSB three or four notches. Since I was either a little over or underclocked, I knew it was not the CPU. The northbridge was suspect.
I had heard there was no thermal paste under the northbridge. I took it off and found Joe had put a layer of Artic Silver under it. It had to be Joe because I don’t believe any Mobo maker is going to use A-Silver. I cleaned it off, checked the HSF surface and, as I suspected, it wasn’t flat – it was concave. I spent the better part of an hour lapping it and gave it another coat of A-Silver. No charge Joe. It was enough to take me over the top.
At this point, I switched to a 1 GHz Duron Morgan. I’ve always been a big Duron fan and this one would do 1275 @ 2.04 Vcore easily with air cooling (MCXC370). I had read in a hardware site testing Kingmax DDR400 that it would be impossible to achieve this with a 200 FSB Duron. The difficult I do immediately, the impossible takes me a few days more longer.
What amazed me again is I could run 3DMark, do surfing and gaming at this FSB without a lockup or BSOD.
This is not the top FSB. I did this with Corsair XMS2400. My PC2700 worked at PC2700, but not much more. With the right stick of RAM and a 266 FSB CPU, I have no idea where it will top out at.
I wanted to do a set of benchmarks with the Morgan at 1.2 GHz+ with my GF3 Ti200. At 200 FSB+, the Morgan didn’t want to go there, although it would do it at a lower FSB. With the bridges painted and plus 10% in BIOS, the best I could get, including the undocumented jumper under the first PCI slot, was 2.03. It’s possible with more Vcore I could have done it.
However there was another problem: The GF3 Ti200 already had Vmods to get it to run at 220/540. When I stuck the card in the board, it went crazy. It wouldn’t hold a multiplier. Every reboot would give me a different one. I even got a post at 1330 MHz. If I backed off on the VIO, everything went back to normal.
It put me in a bit of a quandry: Not only would I have to put the iron to Joe’s board, I would have to undo the Vmods on the GF3 Ti200. I vacilated for a while and then decided against it. The original objective was FSB, and I had the DDR400+ I was looking for.
I didn’t want to put this one in because it’s at 1114 MHz and I wanted one around 2500/2200. I decided to put it in to give you an idea of what you can do even at lower MHz. I compared it to the NFlop just for fun. Remember, this is with a CPU you can get for around $50USD or less.
There were several times I thought about getting an overnight for Corsair PC2700 and a XP CPU, but I always thought better of it. With the new Horses and DDR400 memory just around the corner, it would be like throwing money away just to prove a point. I haven’t bought any new gear recently because I’d like to know what’s going to work with what first.
{mospagebreak}
Disclaimer:
I do not work for anyone or get paid for anything I write. Everything I write in articles or say in emails is Public Domain for the reader to do with as they wish.
Default, aka John R. Abaray
DEFAULT’S DEMENTIA
Vcore
In a post by Umekami10, there is a reference to a jumper under the first PCI slot. This is not on all boards.
This is with the jumper connected:
With the Vbridges connected on my CPU plus 10% in BIOS, I could get .01 more Vcore. If you don’t have the bridges connected, this might get you up to 2.03 Vcore.
I have no idea what else it might raise the Voltage on so, I did this one last.
HIC Chip
Pin 7 to ground with a 47k pot in between will allow you to get enough Vcore to do what ever you want, including cooking your CPU. Make sure the pot is set on max resistence when you start.
AGP
Found in a post by Crusader: Using Rivatuner RC10, you can get an improvement in AGP by enabling AliAGP. This option is not included with all drivers. I tried it with 27.20 and it works. You will not find it with an earlier driver like 21.85. You’ll just have to try it yourself and see if it works with your favorite driver.
Go to power user and click the magnifying glass. This will match the database to the driver you are using. Scroll down to Global and click the + sign. Besides enable ALiAGP, this will give you enough options to thoroughly screw up your driver. If you get a lockup, just reboot using F8 for safe mode and go back in, using Add/Remove to get rid of your screwed-up driver and you’re back in business.
BIOS Hacks
Biggins/Bigsy also gave some BIOS Hacks using WPCREDIT and WPCRSET:
- 61-6c
- 78-80 or C0
- 7B-57
- D5-70
- 8B-60
Depending on your PCI divider, you may not be able to change some of the settings.
This is BIOS 0116 with hacks. With the hacks, it’s a bit faster than 0118. Without the hacks, it’s slower.
Hacks 78-80,7B-57,8B-60
Just use edit and click what you want to try. The change is instant but not permanent – a reboot will get you back to normal.
To get a permanent stick, use WPCRSET. Exercise a bit of caution, cause if you set unstable settings and can’t get back into BIOS, it’s BOO HOO time.
It’s simple enough to use – just enter your settings and reboot. Notice it identifies 98SE as Win NT. Support for this program was discontinued a couple of years ago – there are no plugins for anything newer.
You can find out more about the ALi Chipset and posssibly more hacks HERE.
Something I almost forgot:
Which DIMM slot?
I don’t see this mentioned much anymore.
When DDR boards first came out, I learned early on the size of the memory stick and placement in the DIMM slot can affect how high you can go in FSB. The simplest way to explain it: The length and layout of the traces can effect electrical characteristics. A good example was the AK31V.2 which showed a big jump in KT266 performance with a redesign of the board layout.
The most recent example of this is the NFlop. Stuffing all the slots hurt performance. In the last article, I got nowhere using the first slot. I got to 190 FSB using the third slot. This time I got to 200+ using the first slot.
There are a lot of variables that come onto play. The fatter the sticks and the more slots used will increase the difficulty of hitting high FSBs. You’ll just have to try it and see what happens.
Tweak Time
A lot of newcomers have no idea of how much time can be spent tweaking and modding to get a board working at a level of excellence – not to mention the risks involved.
If you don’t enjoy the challenge, the journey will be filled with frustration. You’re not going to be happy with the trip or the end results. It’s not for everyone.
The way I See It
This board has been maligned to some degree and not without reason. There was a tremendous amount of Hype before the board was released and expectations were high. Initial reviews were of less than stellar performance. It was only when the Hardcore Elite began to work their magic the true potential of this board came to light.
Not being able to adjust the FSB in asyncronous mode is holding this board back. This will be a feature on some of the KT333 boards coming to market, along with Vcore to 2.4v amd Vmem to 3.4 – if I can believe what I read. Again, if I can believe what I read, they will have a 1/5 PCI divider where the XP333 offers a 1/6 divider.
IWill has the chance to make this board the FSB King with the right BIOS. Time will tell.
Disclaimer:
I do not work for anyone or get paid for anything I write. Everything I write in articles or say in emails is Public Domain for the reader to do with as they wish.
Default, aka John R. Abaray
Be the first to comment