X48 chipset is X38 upgrades, inherits X38's Bug?

[cadaveca]

@ 1.8v.

I know it could mean good as well, if they "official spec" it, each chip should be able to do it, but compounded by the binning of the chipset for this function in the first place, as well as the additional voltage requirement(.3v over stock), as said before it is looking towards not scaling as well as say, P35.

We could also say that it's possible that part of the reason for the lower FSB speeds than P35 is because of tighter timings, which, given the added performance, seems a bit realistic, however that then brings to mind the idea thta X38 is nothing more than 975x with added functionality and nothing more.

But it does scale quads pretty good too, so...


It's a mixed bag, for sure, but given the performance and huge increases that intel has been offering over the last 18 months, it seems a bit of a let-down that it doesn't really add much to what's already out there. Chipset's fine and dandy, but I know that myself and many others would like to see intel really increase the FSB of thier cpu products, as it comes as such a benefit when running quads on FSB...

And yet again, only 1600mhz ram speed? one Dimm per channel, @ .3v increase? that's lame. Ram manufacturer's starting to sell higher than 1600mhz DIMMS, with no board support...

[rob[GL]]

I don't care, because by the time DDR3 is "mainstream" enough and cheap enough to buy (like current DDR2 being $75CAD for 2x1GB) which will probably be another year+ away... The next chipset will be out. I think I will skip X48 all together. I may upgrade my P35 to X38 however. Just for the increase bandwidth on the PCI-E lanes for crossfire. My PC never sleeps. heh

[Origin_Unknown]

you can only get 1600mhz with one dimm? am i reading that last table right?

[cadaveca]

Per channel, yes, and needs 1.8v over 1.5v stock.

[Charles Wirth]

Wow, more wrong on wrong.

X38 does support 1600FSB now, my boards fire up on that strap and I can run 4 banks filled @ 1600Mhz tight timing, in fact I am running at 466FSB 1933Mhz stock voltage on the NB and 2v on the memory.

[Shintai]

Clarification is always better than guesses

[cadaveca]

Wow, more wrong on wrong.

X38 does support 1600FSB now, my boards fire up on that strap and I can run 4 banks filled @ 1600Mhz tight timing, in fact I am running at 466FSB 1933Mhz stock voltage on the NB and 2v on the memory.

Hi Fugger, have you done any long-term video renders with that box? I am having issues, as are others, with long-term stability using all banks populated.

What chipset voltage are you using? nV cards, I assume?

I'm sorry for supposedly implying something that may not seem truthful, however I was given "chipset bug on warm boot" that leads to this issue. Running 4 sticks of D9's @ 1200/2.4v is unstable. reboot by system leads to a weird boot...system seems like it's only half there...you know...you click...and there's a delay...shut down and restart fixes issue.

There must be some initialization or something that isn't happening on warm reboots. It FEELS like the P965 pci-e bug, but that didn't affect stabilty as much as this bug seems to.

Hopefully i can get to the system today and I'll try to get some screens of what happens, but screens don't capture slow system response.

[Aivas47a]

Per channel, yes, and needs 1.8v over 1.5v stock.

All the DDR3 1600 or faster ram I've seen for sale so far carries a voltage spec of 1.8 or higher, so this doesn't seem like a chipset problem.

[cadaveca]

"X48 VCCM @ 1.8v" says differently. Also note that I had made mention of this problem long before the above info was posted, so I'm not getting this from some rumour or website...systems I've built with this chipset are having these issues. I hope a bios fixes it, but i think it's not quite that simple, myself.

[Charles Wirth]

FLG,

http://www.xtremesystems.org/forums/...5&postcount=11

Resolved with bios, yes there was an instability problem at first.

No one can test the 1600FSB stap yet so the jury is still out, but the boards I have tested there is not a problem with that strap and no voltage increase is needed till 475+ provided the ram can do that speed. Drop in low in modules and forget about the 1600+ Mhz.

The PCIE problem is not something I am chasing or tested, high speed ram stability with all four slots filled I am.

Believe me or not, I been there to test first hand.

[cadaveca]

If you say it's fixed, then I guess it will filter through to other boards hopefully sooner rather than later. I assumed that the two were related...

[Charles Wirth]

If you match my known good config, I can assure you success. That was the intention of my post.

I have about 400+ hours on many X38 boards (over 3 months continuous) resolving many DX38BT issues that Intel missed. My bad for not including you in our notes.

I would not attempt an OC demo (onstage live for Intel in front of thousands of people) with an unstable machine or one with any outstanding issues. I know this combo is up to snuff. I am not into committing suicide.

[cadaveca]

Hi fugger...personally, I don't tihnk you are at fault here...obviously there are some isues still around, probably on boards that you aren't using, or fixes haven't been implemented yet...some people have issues getting just 4 sticks stable, nevermind @ high speeds.

It's not fair to take it out on you, however you also shouldn't take it personally. Your capabilities are not something I myself question...but I know you can either pick up a phone or send an email, and it gets to who really needs to hear it.

Had i seen your post with the P5E bios, I wouldn't have posted on this issue period, but I am sitll having issues getting X38 stable, and same issue across different boards from different OEMs. So it's either a bios teething issue, or a chipset issue...and given the info provided on "X48", it really seems hard to beleive it to be bios alone.

So yeah, you got a bios, but does it fix it, or is it a workaround? And as such, are we gonna see it on these other obards, or is that up to the OEMs themselves?

I saw the stage demos...most of those benches wouldn't even run if system wasn't close to fully stable, so well done, I suppose, records fell... but we'd all like a crack at it too, ya know?

[Charles Wirth]

I have a definitive answer to why P5E3 boards did not run four sticks stable, this was resolved in bios and I feel confident the problem was addressed and fixed for Asus.

The 0503 bios was specific to resolve this problem.

I'm dealing with this exact same thing on Nehalem.

I will shoot of an Email to K7, my testing may be beneficial to your group after all.

Match my combo and enjoy.

[Blauhung]

/sigh.... the few boards and the minuscule work (by comparison) that you have done with these systems doesn't compare to the thousands of boards/parts that Intel goes through, not to mention OEMs, during the design and validation phase. But hey.... if you assure us that it'll work on EVERY board in existence, the that's good enough for me.

Oh... and just a bit of perspective here. I know of one OEM that is willing to hold off a product launch because of an issue that happens once out of 3000 reboots. THAT'S the kind of standard that Intel must validate to, and if they can't, and it's because of a feature, then they simply won't support it.

/nod

Intel does not officially support DDR3-1600 as part of the full spec on the the X38, that is completely left up to the OEM/mobo makers to support. Intel although can support DDR3-1600 on the Bone Trail boards because they do not represent the entire production run of all X38 chipsets that will make it into consumers hands.

[Aivas47a]

"X48 VCCM @ 1.8v" says differently. Also note that I had made mention of this problem long before the above info was posted, so I'm not getting this from some rumour or website...systems I've built with this chipset are having these issues. I hope a bios fixes it, but i think it's not quite that simple, myself.

That's cool, I don't doubt the existence of the issues you've seen. I was only making an observation on the memory voltage spec. That is, the fact that Intel's specs for the X48 call for 1.8v for memory running at DDR3 1600 seems OK to me since that's what the memory manufacturers are recommending anyway. But, I'm probably not following your point.

[cadaveca]

Aivas47a, vccsm is Voltage Current Control Supply Minimum, not VDIMM. First board I can think of that had this user-adjustable was a ULi chipset...

Anyway, this helps ensure proper current to DIMMS for high-speed clocks, not voltage. I could explain more, but I don't tihnk it's needed, other than to say that the "M" in VCCSM is "minimum". Just seems to me the hold-up time is too long @ high speeds for going back down to 1.5v, and hence the increase. The same problem could potentially affect greater # of DIMMs, and the current requirement on the circuit would be higher, and maybe VCCM is too low for multi-DIMMs as well.


At least that's what I thought...fixable in bios, maybe, maybe not, but Fugger says it is, so I trust his word.