ol' man: VRM 8.5 vs TUSL2-C Vtt

[MadMan007]

Hey Ol' Man I am a longtime reader (though not poster) on this and the [H] message boards and I always find your posts interesting and informative. I became very concerned when I started reading your threads regarding the Vtt voltage on the TUSL2-C since I purchased one after my beloved BH6 died, and all I wanted to do was replace the mobo and not RAM, CPU, PS, etc. If there had been any ACTUAL COVERAGE of Tualatin boards, I prob would have gone ABIT again, but at the time I did not know they even made one! Kind of hard to research when you don't have web access :)

Anyway, I first have one question then i will show you something I found at Intel while researching this.
Question: Do you know exactly what the Vtt voltage does? So far it seems you have speculated it is for the on-die cache, but have you been able to confirm this?

I did a search at Intel for the VRM 8.5 Spec. Now, it seems the searches you did were regarding the Tualatin processors themselves. I wanted to read the "low-down" on the VRM 8.5 spec itself, and what I found is quite interesting.
In this http://developer.intel.com/design/PentiumIII/designgd/24965901.pdf file look only at info for PIII "06Bxh" (Tualatin.)

On page 13 there is a reference to Vtt and indicates that it should be 1.25V. OK so far.

Page 13 section 2.2.2 tells what Vtt is supposed to do. It says that Vtt must reach a minimum (no maximum shown) of 1.13V(1.25-9%) and then 1ms later Vtt_pwrgrd is reasserted. What is Vtt_pwrgrd reasserted as?? I am not sure, but Vtt_pwrgrd is what is actually provided to the proc.

Page 14 mobo solutions should use solution 3 "independent Vcc and Vtt Regulators." Not sure what that means but it seems there is a solution with integrated Vtt and Vcc regulator that is not recommended.

Page 15 looks most interesting. It indicates that Vtt_pwrgrd can range from "1.25V-same voltage plane as processor Vtt" (see table 8, pg 15.) Then, looking at Figure 4, it seems that Vtt_pwrgrd is what is actually provided to the processor. Well, this voltage is allowed to be 1.25-same voltage as processor Vtt.

There are a few possibilities I can think of:
1) Sandra reports not Vtt_pwrgrd as supplied to processor once it is powered and running, but rather reports the initial Vtt used on the proc before Vtt_pwrgrd is reasserted. (section 2.2.2)
2) Sandra reports "AUX" as Vtt coming directly off the VRM rather than Vtt_pwrgrd, which is what is really sent to the proc.
3) The Asus board provides a Vtt_pwrgrd to the processor above 1.25 but up to the allowable processor Vtt.
4) the board provides out of Spec Vtt_pwrgrd to processor. (thought I suspect a 50% overvolt would result in many more screwed processors than we have seen, and I have a hard time believing Asus would screw up the VRM specs TAHT badly.)
Wish I had a Tualatin, then I would measure the voltage straight off the back of the socket. But alas I do not, maybe in a month or so.
I personally think (and hope :) that 1) is correct, but who knows?

PHEW sorry for the long post, but check it out and see what you think!
MadMan007

[ol' man]

4) the board provides out of Spec Vtt_pwrgrd to processor. (thought I suspect a 50% overvolt would result in many more screwed processors than we have seen, and I have a hard time believing Asus would screw up the VRM specs TAHT badly.)


People are not supposed to OC their chips you know. 1.475v would only be a 18% overvolt and only 9% over spec. Not too bad and the chip could probably handle thtfor a good while but it is when people(ie. ME) start OC'n and use 1.85v on it then if the vtt is at that voltage then I think there would major problems too.

Why does powerleap undervolt their slockets when running the tualatin? They run them at 1.32v. Also if you notice that if they ran the vtt rail off of this it would be in spec too. I am confused as too why powerleap would undervolt their chip when Intel specifically said the chip is to run at 1.475v. I have ran it at default to 1.2v but that is besides the point. I think something fishy is going on at least with powerleap and maybe ASUS too. Supposedly they had problems with their older TUSL2 and that is when the TUSL2-C was born. Look around! You cannot find a TUSL2 anywhere. I guess they used the older 815e chipset and did something with the VGA conroller so it could run the .13u core. But evidently there was still a problem with it. You cannot find one any more I think.

[ol' man]

Originally posted by ol' man


You cannot find a TUSL2 anywhere.


I stand corrected. I founda whole slew of them here

here (http://queen.pricewatch.com/search/searchcat.idq?&CiBookMark=N-19d3a68-326773-2e&CiCodePage=Windows-1252&cr=TUSL2&catd=2&cn=Motherboards&qc=%22TUSL2%22%2A%20AND%20%40ctd%202&l=34246&ne=34292&ct=Computer&CiBookmarkSkipCount=15)

These boards ironically are more expensive too than the tusl2-c. I am not sure why. Maybe they work better;)

[Godfodda]

TUSL2 = onboard graphics (+AGP slot), 815E chipset, 4 dual IDE channels (+optional RAID ), 4MB BIOS

TUSL2-C = no onboard graphics, 815EP chipset, 2 dual IDE channels, 2 MB BIOS

[MadMan007]

Well, time for the gloves to come off a bit. I felt like I was pointing out objective information for review that I hoped would not be merely dismissed. If I offend, I apologize in advance...

Ol' man, did you even bother reading the link to the Intel site I posted?? You did not address any of the points I pulled from the .pdf file as I hoped you would. I presented these facts as such with no judgement on them, as I too would like to know whether Asus has made a mistake in their VRM 8.5 implementation. However, you seem to have merely dismissed them out of hand, presumably because you have a pet theory that you do not want to see altered. And hey that is OK, but it is not scientific. This is about facts not speculation.

As for Powerleap, who knows? Does anyone have one to test? And what does an adapter for BX boards have to do with the implementation of VRM 8.5 on the TUSL2-C. It is apples and oranges. One possibility that crossed my is that to compensate for the higher current requirements of the Tualatin, powerleap lowered the voltage so as to not blow out the older VRM on BX boards. Remember, V=IR, so current and voltage are reciprocal, one goes up and the other goes down. What do you think of that as an explanation for Powerleap's strange voltage?

Read the Intel spec file and do not dismiss the facts just because they contradict what you believe. I do not know whether I am right or wrong and quite frankly do not care. I just want the truth. Maybe "you can't handle the truth" :) One program (Sandra) misreporting a voltage does not a design flaw make. Lots of other people report strange readings on all types of motherboards. After all it may very well report Vcc as "AUX" instead of Vtt, yes? unless someone has evidence otherwise.

One way to check whether Asus merely "hacked" the CUSL2 by only substituting the new B-step northbridge is to compare the VRM chips and circuits on a TUSL2-C to those on a CUSL2. I will look at mine tomorrow and tell you what the VRM chip says on it, as well as what ther components appear to be part of the voltage supply. Maybe someone witha CUSL2 can then post their findings? If they are the same, that does not mean Asus mis-implemented it either, the CUSL2 may have been designed with VRM 8.5 in mind just without the B-step Northbridge present. If they are different then it would imply that the TUSL2-C does have VRM 8.5 correctly implemented, tho it would not prove it.

The only true way to test this is for someone with a Tualatin and a TUSL2-C to measure voltage straight off the board with a voltmeter. Directions for which pins to measure can be found in this: http://forums.overclockers.ws/vb/showthread.php?s=&threadid=48446 thread. If no one else is [H] enough to do this, I will do it myself in about a month. Other than that all I have seen is taking speculations and calling them facts. Again, let's stick to the facts so we can figure this out, ok?

Thank you,
MadMan007

[ol' man]

Originally posted by MadMan007
Well, time for the gloves to come off a bit. I felt like I was pointing out objective information for review that I hoped would not be merely dismissed. If I offend, I apologize in advance...

Ol' man, did you even bother reading the link to the Intel site I posted?? You did not address any of the points I pulled from the .pdf file as I hoped you would. I presented these facts as such with no judgement on them, as I too would like to know whether Asus has made a mistake in their VRM 8.5 implementation. However, you seem to have merely dismissed them out of hand, presumably because you have a pet theory that you do not want to see altered. And hey that is OK, but it is not scientific. This is about facts not speculation.

As for Powerleap, who knows? Does anyone have one to test? And what does an adapter for BX boards have to do with the implementation of VRM 8.5 on the TUSL2-C. It is apples and oranges. One possibility that crossed my is that to compensate for the higher current requirements of the Tualatin, powerleap lowered the voltage so as to not blow out the older VRM on BX boards. Remember, V=IR, so current and voltage are reciprocal, one goes up and the other goes down. What do you think of that as an explanation for Powerleap's strange voltage?

Read the Intel spec file and do not dismiss the facts just because they contradict what you believe. I do not know whether I am right or wrong and quite frankly do not care. I just want the truth. Maybe "you can't handle the truth" :) One program (Sandra) misreporting a voltage does not a design flaw make. Lots of other people report strange readings on all types of motherboards. After all it may very well report Vcc as "AUX" instead of Vtt, yes? unless someone has evidence otherwise.

One way to check whether Asus merely "hacked" the CUSL2 by only substituting the new B-step northbridge is to compare the VRM chips and circuits on a TUSL2-C to those on a CUSL2. I will look at mine tomorrow and tell you what the VRM chip says on it, as well as what ther components appear to be part of the voltage supply. Maybe someone witha CUSL2 can then post their findings? If they are the same, that does not mean Asus mis-implemented it either, the CUSL2 may have been designed with VRM 8.5 in mind just without the B-step Northbridge present. If they are different then it would imply that the TUSL2-C does have VRM 8.5 correctly implemented, tho it would not prove it.

The only true way to test this is for someone with a Tualatin and a TUSL2-C to measure voltage straight off the board with a voltmeter. Directions for which pins to measure can be found in this: http://forums.overclockers.ws/vb/showthread.php?s=&threadid=48446 thread. If no one else is [H] enough to do this, I will do it myself in about a month. Other than that all I have seen is taking speculations and calling them facts. Again, let's stick to the facts so we can figure this out, ok?

Thank you,
MadMan007

I have never said anything was a fact.

It is pretty much a fact that powerleap does undervolt their cpu's in the slocket adapters they have. WHy I am not sure but the only way it makes sense if they run the vtt off the vcore rail. If they do it then it may be possible for some MOBO makers to do this route that is all I am saying. It may save them a few bucks. Best put your gloves back on and grab your hockey stick. Lets get back to work;)

I have simply said what I understand. I read through I guarentee what you posted and quite frankly man you are above me if you can understand all that without studying it a little bit. I didn't have the time. I think what you say is what should be done. Sorry if I don't understand circuts as well as you;) I do know something fishy when I see it though like powerleap running their chips at 1.32v. That is weird. That has ben pretty much confirmed by some people in Taiwan and would explain why people cannot OC their chips on them.

When I sit back and think about it I really would like to think that ASUS would not do that with their mobo's but until someone can take that measurement we will not know for sure. I am wondering if the sensor for measureing the vcore and vtt is in the cpu like the internal temp sensor. If it is then I would think maybe sandra is right. If not then your guess is as good as mine. You don't know and I don;t know without testing the mobo on the back side. I RMA'd my tusl2 before I put all of this together pretty much. All I knew I didn't like the way my chips stablitiy went down day by day. I know many people here that have had their chip running at 1.8v but they are not experiencing the stuff I did. I also ran prime all the time and I know it heats up the cache alot cause of use.

So tell me then does the vtt supply the cache with its juice seperate of the vcore or is it something else? I do know by looking at the diagrams I have had at the other thread which is where you should have posted all this that the vtt pins and the cache are on the same side this is why I am assuming it does this.

Do you think that is a safe assumption?

If so It would explain why my chip got bad so fast possibly.

[MadMan007]

Sorry bro, it is just that from a lot of your posts you are telling people TUSL2-C=teh devil ;), without any proof one way or another, so it seemed to me you were putting together a few random bits and making an important conclusion. Damning a product on incomplete facts seems premature. Who knows maybe you had a bad CPU or mobo?? it happens...
I am no circuit expert, but in that Figure 4 diagram there is a line from Vtt_pwrgrd to the CPU, but not from Vtt *itself* to the CPU. Of course, it may just not be shown in that diagram, especially since that diagram is describing the operation of Vtt_pwrgrd and not Vtt itself.
As for what Vtt does, all I can tell from the VRM 8.5 specs is that it is used as an initializing signal for the VID pins. I am not even sure whther Vtt actually provides any *power* to the CPU or whether it is just a signalling voltage. I think it is a signalling line to tell the mobo whether it has a Tualatin (.025V steps) or a CuMine (.05 V steps.) I am not an electrical engineer though so I cannot say for sure.
As for Powerleap, I am inclined to think they had to have a strange voltage to run the CPU from the voltages and currents supplied by the Slot1 interface. Maybe they do in fact use the Vcc line to run their Vtt signal as well? Does not mean Asus does tho.

Now, this is taking a leap for me, BUT why would a processor be designed to run with different voltages going to different parts of the chip (main CPU and cache)? that seems extremely awkward to me. Would be like having some engine cylinders run with 87 octane and some with 93 octane, with 2 fuel pumps, etc. Then again, chips are not engines either, it just seems odd to me.

Will post my TUSL2-C VRM chip info and what not soon when I open my computer!

MadMan007

[Rmcky]

MadMan007, I've been following this too and tend to agree with you. I think Ol'man should spell out his real motivations, because there is more to this than meets the eye. Maybe he was hoping for a bigger refund on the board he abused so badly. I think he already got his abused CPU replaced, but his ethics are really none of my business.

Anyway, I own a TUSL2-C and it's one of the finest boards I've ever owned. Am I able to take my Tualatin to 1.6? No, but by the same token, I'm not that bothered by that, either. Will it run stable at 125mhz FSB? You bet.

You know, there are thousands of these boards out in the field and with all the reading I've done, as far as I can tell there is only "1" person who's disappointed with theirs. I have a feeling that if this keeps up, the suits he keeps joking about will be paying him a visit, to serve the court papers.

You know, critisizing a product with facts is one thing, but putting forward the amount of effort this one man crusade has, on pure speculation, just isn't quite appropriate. I hate to admit it, but I'm afraid I'd get a little chuckle if he did have to face the music.

It's obvious he's an excitable boy, but a dose of reality might rein him in a bit. Just my opinion, I could be wrong.:)

Rick

[ol' man]

Originally posted by Rmcky
MadMan007, I've been following this too and tend to agree with you. I think Ol'man should spell out his real motivations, because there is more to this than meets the eye. Maybe he was hoping for a bigger refund on the board he abused so badly. I think he already got his abused CPU replaced, but his ethics are really none of my business.

Anyway, I own a TUSL2-C and it's one of the finest boards I've ever owned. Am I able to take my Tualatin to 1.6? No, but by the same token, I'm not that bothered by that, either. Will it run stable at 125mhz FSB? You bet.

You know, there are thousands of these boards out in the field and with all the reading I've done, as far as I can tell there is only "1" person who's disappointed with theirs. I have a feeling that if this keeps up, the suits he keeps joking about will be paying him a visit, to serve the court papers.

You know, critisizing a product with facts is one thing, but putting forward the amount of effort this one man crusade has, on pure speculation, just isn't quite appropriate. I hate to admit it, but I'm afraid I'd get a little chuckle if he did have to face the music.

It's obvious he's an excitable boy, but a dose of reality might rein him in a bit. Just my opinion, I could be wrong.:)

Rick

Yeah you are wrong.


madman, a chip and a car are two different things.


As long as I get my cash back from my RMA I guess with dealing with people like you all I could care less anymore. Burn up your chip possibly! Don't check it with a voltometer! I don't care, I know my chip is running in spec. Do you?

I got other stuff to do than argue why some vtt pins on the socket run at 1.25v and others at vcore what ever that may be. I mean they put the pins on the chip for a reason. DO you think the voltage magically changes once it hits the cpu?

How is that for excited.

And yeah the ASUS TUSL2-c is the devil, it sucks for OC'n too. Why do you need 3.5v for the I/O voltage. That is what they set it too by default. Seems my ST6 runs with 3.3v just fine.

[DSTA]

Hm, I'm not sure VTT_PWRGD has anything to do with this issue.

Here's a description taken from the Tualaron datasheet (http://developer.intel.com/design/celeron/datashts/29859601.pdf):

"The VTT_PWRGD signal informs the platform that the VID and BSEL signals are stable and should be sampled."

IMO this means that VTT_PWRGD (VTT powergood) is just a "switch", i.e. the mobo's & VRM's job is to switch VTT_PWRGD to high once VTT is stable - otherwise no boot.

Things that would be interesting:

a) Someone please grab a voltmeter and measure. Please! I'd do it but I have neither a TUSL2 nor a Tualaron.

b) Failing that, someone take a look at the implementation of the systems monitor chip used on TUSL2. Is it possible that it just reads the wrong voltage?

c) See if it possible to get "closer" to the hardware of the sys monitor chip buy using the sensors tools available with Linux.




My personal opinion on the matter: emprical results say there is something subpar with the TUSL2 WRT overclocking. That's a fact. But somehow I doubt ASUS would deliberatly overvolt VTT in the situation of a 0.13 CPU being used at default voltage.

[MadMan007]

Well, maybe the TUSL2-C is slightly behind the ST6 for OC'n (which would be typical for Asus vs Abit products after all) But I have seen ol' man complain about his PCI or USB crapping out at 38 MHz. Well, I just realized I have been running mine at 40 MHz for months with no problems. I think he got a bad (or worse than usual, for OC'n) TUSL2-C.

That is all,
MadMan007

[ol' man]

Originally posted by MadMan007
Well, maybe the TUSL2-C is slightly behind the ST6 for OC'n (which would be typical for Asus vs Abit products after all) But I have seen ol' man complain about his PCI or USB crapping out at 38 MHz. Well, I just realized I have been running mine at 40 MHz for months with no problems. I think he got a bad (or worse than usual, for OC'n) TUSL2-C.

That is all,
MadMan007

Well maybe it is around 40MHz then I never really messed around in that area. I kept everything at 133fsb or close to it. The ASUS is 134, you have no 133fsb unless you want 1/3 dividers but then you are back to original problem of the chipset configuration not handling it. I have heard it is right in the 38~40MHz area for many people. They can never get it higher. That is the bottom line. There is one thing to boot your chip at 44MHz pci and actaully run windows with it. I know my ST6 has no problems with 44MHz pci but I know the tusl2 craps out early. I feel that has nothing to do with this problem though but it might but i doubt it. There may not even be a problem. But given that most buy the celery to OC I am not going to say get the TUSL2 anymore cause I got it and it OC'd very cr***y. Many others say the same. That is not a falacy.

[Godfodda]

Hope you guys don't mind if I stray off-topic a bit.

I like reading about this and find myself learning a couple of things. But if it turns to a flame war, the mods will pull it and the discussion is over. Play nice, please. Some of us would like to see the data.

[Rmcky]

Godfodda, I agree, this is interesting and I'd like to know more because I own the board in question. I wasn't looking to flame or even imply that there is anything wrong with being excitable. As a matter of fact, I admire Ol'Man and his overclocking efforts. I've learned quite a bit of useful stuff from his experimenting. All I was saying is that accusations like these should be based on facts and not speculation. Also, it seems Asus should be involved in this or at least have the option of explaining it or defending themselves. Hopefully, more info will be forthcoming from them in the near future. One thing I am curious about, though, is if Ol'Man's replacement board will have the same tendancies. There's a possibility his first board had some sort of defect and that he could change his mind about how doggy they are. As I said, I own one of these and I still feel it's one of the finest boards I ever used. My vtt may very well be out of spec and, if so, I'd certainly like to be aware of it, if for no other reason than to be sure I don't cause myself additional problems by increasing the voltage past the reasonable point.

I hope Ol'Man will follow through with this until everything's answered to everyone's satisfaction. Even so, I still have a feeling that the board simply doesn't read vtt at all and that other programs, such as Sandra and MBM, are picking up the core voltage twice from the sensor and parking one of them in the vtt spot. At least that's what I'm hoping the final verdict will be.

In the mean time, Ol'Man, keep up the good work, just try to keep it real.:)

Rick

[ol' man]

Originally posted by Rmcky
Godfodda, I agree, this is interesting and I'd like to know more because I own the board in question. I wasn't looking to flame or even imply that there is anything wrong with being excitable. As a matter of fact, I admire Ol'Man and his overclocking efforts. I've learned quite a bit of useful stuff from his experimenting. All I was saying is that accusations like these should be based on facts and not speculation. Also, it seems Asus should be involved in this or at least have the option of explaining it or defending themselves. Hopefully, more info will be forthcoming from them in the near future. One thing I am curious about, though, is if Ol'Man's replacement board will have the same tendancies. There's a possibility his first board had some sort of defect and that he could change his mind about how doggy they are. As I said, I own one of these and I still feel it's one of the finest boards I ever used. My vtt may very well be out of spec and, if so, I'd certainly like to be aware of it, if for no other reason than to be sure I don't cause myself additional problems by increasing the voltage past the reasonable point.

I hope Ol'Man will follow through with this until everything's answered to everyone's satisfaction. Even so, I still have a feeling that the board simply doesn't read vtt at all and that other programs, such as Sandra and MBM, are picking up the core voltage twice from the sensor and parking one of them in the vtt spot. At least that's what I'm hoping the final verdict will be.

In the mean time, Ol'Man, keep up the good work, just try to keep it real.:)

Rick

Well I RMA'd the board for a refund. I only had it for 1 week.