XMS3500 C2 test, same stick, 3 diff boards, results inside

I tested the same stick of Corsair XMS3500 C2 (Blk) on a BD7II, IT7 (original Abit board, not the MAX2), and Albatron PX845PEV PRO. I determined the max DDR, and also the fastest it would run at most aggressive settings. Interesting results (all settings totally clean in Memtest):

BD7II: maxed out at DDR412, 2-3-3-6 and 167 fsb, 2-2-2-5. This is at max 2.7 volts the board will do.

IT7: DDR454, 2-3-3-7 (can't test higher as my CPU craps out at 171 fsb) and 180 fsb, 2-2-2-5. Used 2.8 volts. Using 2.9 won't make a major difference, maybe just clean out a few #5 and #6 errors out of a stick. It won't take away that BEEEP! on boot . Will do DDR434, 2-2-3-6 at 2.9 volts.

PX845PEV PRO: DDR454, 2-3-2-6 and DDR432, 2-2-2-5. This is at 2.8 volts, the max the board can do.

Not surprisingly, when I switched to the IT7 from the Albatron board the IT7 really felt sluggish. Heck, at DDR432, 2-2-2-5, the Albatron board must have clicked off 8 Memtest loops in 35'. I think the mem timings in the IT7 BIOS is real relaxed in order to get it to run DDR well. If the IT7 boots at a DDR setting, 99% of the time Memtest will run clean. If it can't handle it: BEEEP!

The XMS3500 doesn't seem to do well at aggressive settings on the IT7, as even at DDR400, I still have to run 2-2-3-6. Even at 2-2-2-6, it maxes out around 193 fsb. The Albatron? Unreal at DDR432, 2-2-2-5. The XMS3200 C2 might do better at aggressive timings on the IT7.

I still have to install an OS in the IT7 and test 3D stability. The Albatron tested to be 3D stable. The Albatron board is a work in progress though, and might need a new rev to work the bugs out. But if you vidpin and run an Nvidia card, it smokes! It is also 3x faster in DOS at partitioning and formatting drives. It outoverclocks a BD7II by 2 mhz and I suspect the IT7 by the same amount.

Did you not try the BD7-II at 2,3,3,7? Lets make it fair.

It is true that the same stick of ram yields different results on different motherboards though. Finding the golden combinatiion is many times more productive than having any one particular piece. My BD7-II (and IT7's) loves Kingston Value Ram 2700, and produces 429-436MHz (depending on the stick) with it at the measly 2.65V Vdiimm the BD7 puts out. This is at 2,3,3,7 with the enhanced mode on.

There is a mod for the BD7-II to increase the Vdimm to whatever you want, but I haven't done it as my cpu gives up at 163MHz, coinciding with the ram.

The Value Ram cost about 100 bucks less than Corsair PC3500, but on my board at least is outperforming it. But as you have shown other boards may demonstrate different preferences, a useful point.

Larva, I found on the BD7II, 2-3-3-7 wasn't much better than 2-3-3-6; in other words, I couldn't go much farther. On the IT7, it really made a difference as I could go up to DDR442 and DDR454 by relaxing to 2-3-3-7 from 2-3-3-6. I still don't know how much higher I can go with the IT7 unless I put in a CPU that does up to 180 fsb, but I imagine at something like 2.5-3-3-7, who knows, maybe 3:4 is possible there. I don't see going up to the 3.2 volts in the BIOS will help with these Winbond chips.

Based on my results on the BD7II, I got an RMA# from Corsair. Needless to say, due to the performance on the other two boards, I think I'll keep the stick. Also the BD7II is hampered by the 2.7 volt limit, and as far as I know (many have tried), it's not possible to increase on that.

What BIOS are you using on the BD7II? I am using the stock BIOS (think it's an old one). I will update to the latest BIOS in hopes the mem timings are similar to the IT7.

The Albatron board is interesting, since when I set DDR432, 2-2-2-5 and exit the BIOS, I get beeping. I shut down, then restart and Bingo, it barges right into Win98SE and is even 3D stable. Sandra benches confirm DDR432, and like I said, Memtest is blazing fast through the loops. I dunno how the Albatron board does it.

I am also a few #5 and #6 errors away from doing DDR456, 2-2-2-6 on the Albatron. If only I could run 2.9 volts!

With my ram going up to 7 makes it work. If I go lower than 2.3.3.7 on any setting the results are poor. But of course different memory (or even a different BD7-II, although to a lesser degree) can make all the difference.

With Asus boards if you set the timings too fast it will beep, you hit reset, and it boots. But it's doing it at slow timings. Is it possible the Albatron is falling back to safer settings on the reset? Perhaps not, but that is the Asus behavior. I find that Asus board are significantly more advance in error handling than anything else I have tried, so this characteristic may be unique to them.

Here is a schematic of the BD7-II Vdimm mod. It is known to work, and work well. The chip in question is behind the memory, near the power supply connector.

Sorry, I forgot the BIOS. I am running Abit's version BU (9/17/2002). There were some memory timing changes intended for DDR400 in the previous release.

When I boot Windows, Sandra shows DDR432, with typical 3200 score, and when I shut down and restart, the BIOS is set to 2-2-2-5. It's only when I adjust the BIOS and the computer self-boots, that I get the beeping.

Sandra sees my memory as 1.5-2-2, but I think Sandra doesn't recognize this Alb board properly. CPUZ doesn't work on this board either so I can't confirm 2-2-2-5.

I believe the beeping I am seeing is due to improper voltage being applied to both the CPU and ram on hot restarts, unless you vidpin to your stable voltage. On hot restarts, the board uses default (1.50 or vidpin voltage) to start, and them implements the BIOS voltage later (dumb, I know). And the vidpinning doesn't really increase the voltage choices in the BIOS at all. And I suspect the voltage regulator supplies both CPU and DDR voltage. I e-mailed Albatron telling them to fix the BIOS.

Right now I am vidpinned to 1.7 but my CPU needs 1.75 (to do 171 fsb anyway). So I get the hot restart problem. I will vidpin to 1.75 the next time I have the CPU out, or if my next CPU needs that. But shut down totally and press the ON button - no problem doing outrageous CPU and DDR overclocks.

So on your diagram, I just apply a resister pot to that R2 resister on the board and to ground? And I just turn the dial to set my VDIMM? If this works you are a pioneer. Maybe because people were looking into the usual method of connecting the pot to the legs of that bigger chip on the board?

Clevor said:

So on your diagram, I just apply a resister pot to that R2 resister on the board and to ground? And I just turn the dial to set my VDIMM? If this works you are a pioneer. Maybe because people were looking into the usual method of connecting the pot to the legs of that bigger chip on the board?

Click to expand...

The catch is you have to add R2 as well. This is done by clipping or unsoldering the leg of the IC and connecting one end of R2 to the motherboard trace, and the other to the leg. Then you merely connect the leg to to pot as shown.

I can't take credit for the mod (or the diagram). The info came from:

http://www.xtremesystems.org/forums/showthread.php?threadid=1348&perpage=25&pagenumber=1

Chances are very good that a combination of a vid pin mod and the above Vdimm mod would take my 1.6a higher. I just don't have the luxury of abusing these components at present, so it will have to wait for another day.

The catch is you have to add R2 as well. This is done by clipping or unsoldering the leg of the IC and connecting one end of R2 to the motherboard trace, and the other to the leg. Then you merely connect the leg to to pot as shown.

Click to expand...

Ah yes, that IS the catch. You have to do microsoldering; that's beyond my skills and equipment.

I'm banking the latest Abit BIOS handles memory the same as the IT7 BIOs, as the boards overclock very similar. Have you tried the Mr. Natural one for the BD7II that was floating around?

The IT7 I installed is selling here in Japan for around $70 U.S. because it's in the discontinued soccer ball box. This is the original board made by Abit, not ECS. It's a viable option as I was surprised to see VDIMM up to 3.2 in the old (4-02) 8H BIOs installed. I thought you needed a modded BIOs for that. Plus I simply plugged in a USB mouse and keyboard and it works fine; the mouse also in DOS. I also thought a modded BIOS was needed for that. But it remains to be seen if I have the cold boot problem.

But this board is famous for the AGP slot being soldered a bit off. Have to install the card, then force the board toward the back of the case enough so you can start the motherboard screws. Just loosening the screws didn't work; the video card just wouldn't go in the slot.

Hehe, it won't do itself, but it's not that hard. I've changed many a surface mount component with an ordinary Weller professional grade 25W iron. It really isn't that hard. Clipping the lead rather than unsoldering it makes it easier.

I've not tried the Mr. Natural BIOS.

I am pretty familiar with the IT7, although I have not owned one. I still use PS2 k/b's, so it's not for me. It is possible to raise the Vdimm to 3.2V through the BIOS, but they only yield 3.0V actual when so set. Most users note the AGP slot being off, I believe it's off on the IT7 Max2 as well. I hope the managed to correct its placement on the Max2 v2.

My experience with ECS is so monumentally negative I wonder about Abit now that a large percentage of their boards are produced by ECS. I believe my board is an Abit produced example, but I am not positive. It did come in a decorated box, I guess the ECS version comes in a plain one. At least on the IT7. In any event, I like the board I have. It was cheap (83.50) and yielded 162fsb with my chip where the Asus P4S533 I had would only do 150. Now if they would do something about the innacuracy of the hardware monitoring...

Now if they would do something about the innacuracy of the hardware monitoring...

Click to expand...

Heh heh, I was actually surprised on this IT7: the VCORE only undervolts by .055; and the CPU temp in the BIOS only off by 10 C. Noth'in compared to my TH7II.

I believe all the BD7IIs have the bronze color PCB and were made by Abit, as were all the IT7s, no matter the packaging. The BD7II (and I presume IT7) was the highest overclocking and most stable board I've used until I tried the Albatron. The flakey 9700 Pro won't run on the Alb board, though, so I was not able to get a 3DMark2001 bench at 171, 3:4, DDR456 on this board. In Win2k I see just shy of 3500 bandwidth. I estimate a 15,400 score on a little 1.60a at 2.74 gig. Sandra scores won't show it, but running something like 2-3-2-6 will make a big diff. in 3DMark.

I know the 9700 Pro runs on Abit boards so I hope to get a run in at DDR454, 2-3-3-7, if the IT7 is stable there.

As you hinted at earlier, I believe the Abit boards do obtain better results with the XMS PC3200C2 ram, although I haven't played with any myself. This seems to be the consensous amongst the hard core Abit guys though.

Since the BD7-II isn't giving you much in the way of results in its current form, perhaps you should try the mod on it. It's really not that hard. Just get a fine tip 25W soldering iron and give it a shot. As mentioned I use a Weller unit that has served me well over the years. You sound as though you have a significant involvement in PC's, perhaps it is time to try it.

I do like my BD7-II, especially since I lucked upon cheap ram that it really likes. The results from the Value Ram 2700 have been so impressive (and consistant) that I am curious if the new Kingston HyperX PC3500C2 is any good. If it outperforms the Value Ram by any margin at all, it ought to be awesome.

Are the ECS produced boards are black in color?

I'm glad I'm not the only one that knows the ATI cards are still rough in many respects. I at one time used lots of ATI cards, but not since the advent of the Geforce. Nvidia's tend to have compatibility and driver developement edges over the ATI, I look forward to the day they re-capture the performance crown as well.

What sort of fsb does your 1.6a reach on 1.65V? If my chip would get all the way to 171 on 1.75 I might go that route, and do the Vdimm mod at the same time. I just can't afford to replace this stuff at present, so I will have to wait before turning up the wick to that degree.

The Albatron board does look promising, thanks for your report on its properties. These old BD7-II's and IT7's are amazing though, in spite of their shortcomings.

Larva, I think the Kingston 3500 must use 5-ns Winbond chips, so they are equivalent to XMS3500 C2. I did try a 3200 C2 chip on the BD7II and results were disappointing (I think it only did 193 fsb, 2-3-3-6). I have an RMAed stick on the way from Newegg.

The Winbond ram is very consistent, I think that's why Corsair is guaranteeing them at DDR434, 2-3-3-7. My TwinMOS and Corsair 512 MB sticks perform almost identical on the Albatron: DDR429-432, 2-2-2-5. Even on the IT7, I can do DDR434 at 2-3-2-6 if I use 2.9 volts.

I installed an OS on the IT7 and it maxes out at 170, 3:4. It will not do 171 as IL-2 Sturmovik locks up. I expected it as Memtest locks up here even at relaxed settings. It's the CPU maxing out. So the BD7II does 169 fsb, IT7 170 fsb, and Albatron board 171 fsb.

I have tested many 1.60a and 2.26bs, and none would do over 165 fsb with less than 1.7 volts. At 1.65 I think I can do only 160 fsb on it. Since many people upgrade CPUs in a year, why worry about killing it at 1.75 volts? It's not like I want it to last 5 years.

I benched the IT7 at DDR453, 2-3-3-7 and get 15,238 in 3DMark2001. So on the Albatron board at DDR456, 2-3-2-6 I indeed should hit around 15,350-15,400 if I could get the 9700 Pro to run on it. Very good scores on a little 1.60a at 2.74 gig. It's due to the low CAS latency of DDR (and the 3400-3500 bandwidth helps).

Cool, my 1.6a does 163fsb on 1.64 volts actual, so it's good to know it probably has more in it if/when I apply more voltage.

I totally agree that in 5 years a 1.6a will be a memory, but I just had 3 back surgeries in 8 months, complications out the wazoo, and don't have a job at present. I simply can't afford to replace a cpu right now, even at the low price of modern components. I only paid 140 for the 1.6a many months back, but even this is exceeds my toy budget at the present.

Your BD7-II does seem to be somewhat under par. Most folks seem to be getting more from it with XMS PC3200C2 than 193. I'd volt mod that sucker and get it moving My Value Ram is good for 214-216 (depending on the stick) at 2,3,3,7 on 2.65V actual on my board, and I am hoping that the volt mod and say 2.9V will allow it to keep up the 1.6a when I go to 1.75V on it.

Some of the Value Ram PC2700 out there is 6ns Winbond, but the sticks I have used bear Kingston logos and dissimiliar markings. I have no idea what my chips are, but seem to perform at an identical level with the Windbond equiped version based on the reports I have been given. I agree it is likely the PC3500 is 5ns Winbond, or the Kingston labled equivalent.

Thanks for your input, good luck with the Albatron.

Ignore my sig, I have a P4 1.6a and BD7-II

Well today has been both good and bad, discovering this thread I guess now I know I can't increase my Vdimm on my board (ignore sig; I have a BD7-II also) -- just don't have the skills either.

At first when I saw this I was thinking I could use my Larry Quinn Vmod thing but oh well, that looks impossible. It really irks me that this board's Vdimm sucks so bad. All I want is 2.8 Volts.

Thanks larva for digging up that info. I searched and searched the web when I got my board about a month ago and never found anything about a Vdimm mod, I figured it just didn't exist.

Would you guys say this is still one of the better boards around for OC'ing? I max out (Prime stable) at 172 FSB (2760 MHz; 1:1) with 1.71 Vcore. It's good and all but I'd like more flexibility with my ram. It'll go up to about 180 and then craps out due to lack of voltage (it's Corsair 3200 C2).

Yes, even with its limitations the BD7-II is capable of impressive performance and stability. Since the only only cure for your quandry (other than the Vdimm mod) is to buy a new board, you have little to lose if you muck it up. I would recommend getting a good quality soldering iron, some quality solder with a little silver content, and a little practice on an old board. The mod isn't really that hard, and might gain you the 3:4 at 172fsb with your ram. At which point your BD7-II would be capable of slaying all but the mightiest of giants (those packing more that 2.7GHz).

When you RMA'ed that stick to newegg what did you say was wrong with it?? Do they accept "didnt overclock well"?

I cannot say enough good things about Newegg!!!

I simply told them on my BD7II, it did not meet Corsair's guaranteed specs. They sent me an RMA# no problem.

Did you know I also sent back a 3500 C2 Plat stick that did even better than the stick I tested on this thread??? It did DDR423, 2-3-3-7 or so when the one I tested on this thread did DDR412, 2-3-3-7. Shoot, who knows how high it would've gone on the Albatron. Oh well, I hope the replacement one does at least as well.

I could send the Plat stick back since it still did not meet Corsair's DDR434, 2-3-3-7 spec.

I got a Gigabyte Maya Pro card that will not run on the Albatron board but runs on all my other boards. Gigabyte has not replied to e-mails. I complained to Newegg and they gave me an RMA# even though it is beyond the 30-day return period. Unfortunately this item can only be replaced, not refunded, so I am hoping for the latest revision board which may have fixed the problem.

Needless to say, you can't go wrong with Newegg!

Hey Larva and Grungee, you can get great performance from the little 1.60a. I find that it is only around 200 points behind a 2.26b at same fsb in 3Dmark, and I'm talking the 15k range. You can still do this at 1:1 at fast CAS settings.

I checked my notes and I see one run at 169, 1:1, 2-2-2-6 where I clocked a 15,241 score. Of course this is with a 9700 Pro overclocked to 355/345. I lose 500 points unoverclocked. So even at 1:1 you can get stellar scores.

Note an IT7 at 170, 3:4, 2-3-3-7 gets me 15,238, however that card was clocked to 341/341. Figure 355/345 would give me maybe 15,300-15,350? Hey, the BD7II at 1:1 ain't that bad.

I think my BD7II board has a BIOs that is lousy at high DDR but extremely stringent at CAS.

I figure the Alb at 171, 3:4, DDR456, 2-3-2-6 should net me at least 15,500 if I could run the 9700 Pro at 355/345. These are all pretty good scores as normally guys who run 2.9-3.0 gig are in this territory.