Confusion on Intel 865PE with ICH5 southbridge.

[Joe DeFuria]

I'm starting (in earnest) researching for mynext desktop rig in anticipation of Half-Life2.

I'm nearly certain that I'll be cobbling together a P4 (865PE) based system. In my initial looks at the products, I've started to become a bit confused,with respect to IDE support.

In the end, all I "want" is support for 2 serial ATA hard drives, and 2 ATA DVD/CD drives. RAID support for the hard drives is a nice plus, but not crucial.

Simple question: 865PE with the ICH5 southbride, with NO additional 3rd party IDE controllers should be sufficient...right? I've looked at two product pages (Asus and MSI) so far, and they are very poorly laid out...highlight certain things as "optional", making it hard to discern what capabilities are possible with and without which options.

 

[Joe DeFuria]

OK....I'm starting to get a better handle on this now. There's an ICH5 southbridge (no intel Raid support) and an ICH5R southbridge, that includes Intel Raid support. (I didn't realize there were two flavors of this southbridge.)

Now, however, I am perplexed by this apparent fact:

Virtually all of the 865PE / ICH5R motherboards that I've looked at also include a 3rd party RAID controller. This seems completely ass-backwards to me. :?

I don't want any 3rd party RAID controllers in my new system.

I have found exactly ONE motherboard so far that has the ICH5R southbridge, without an additional 3rd party controller, and that's Intel's 865PERL.

It seems to have everything I want, (8 USB ports and Firewire) and nothing I don't. (multiple raid controllers, overclocking support.) Decently priced too.

I have no idea how the multiple sound codecs on the market stack up against one another....I'm not sure at this point if I'll be getting something like an Audigy or not. I actually don't need anything more than your basic 2.1 speaker support, (I hardly ever get to use any significant sound without the headphones on) but am more interested in low CPU utilization out of the sound system.

Anyone care to talk me out of this Intel board?

 

[Natoma]

You know you can disable the 3rd party RAID controller in the bios if it's that important to you and you can't find one.

If you want support for Prescott, your only choice at this point is the 8KNXP. Whatever you purchase, make sure that it supports FMB 1.5 aka VRM10.0. If it doesn't support that, then it won't have the voltage support for Prescott and you're stuck with a P4-3.2Ghz as your highest upgrade.

Intel has stated that they will add FMB 1.5/VRM10.0 support to their boards, but who knows when that will occur, and whether or not you'll actually get that board in retail.

I run onboard sound btw on my 8KNXP, thank you very much RealTek Audio, and the CPU utilization never rises above 5%, unless I'm running F@H that is. The layout is very well done and it is chock full of features. Sure it's rather expensive for a mobo, i.e. roughly $200, but you're getting a future proof board (FireWire, tons of USB 2.0 ports, external SATA ports, 4 on board SATA ports, all supporting RAID0 and RAID1, yadda yadda).

I researched about 3-5 months for my system and the 8KNXP was the best board available for what I wanted. Who knew it'd also be the only board to support Prescott atm...

 

[Florin]

Well I don't care for overclocking either but boards from manufacturers other than Intel are a little more flexible at 'bending' the rules in more harmless ways, like enabling PAT on the 865 or promising the possibility to upgrade to the first generation of the Prescott processor.

 

[Joe DeFuria]

I know I can disable devices from the BIOS, but I just don't like paying for things that I don't need, and ultimately only serve to clutter things up.

However, I am actually concerned about Prescott support. I knew lots of "first gen" 865 boards were going to have prescott issues...I didn't know that the 8KNXP was the only board right now that won't have problems.

That may indeed force me to spend $50-75 more than I was hoping. (I wanted to keep the motherboard to about $150).

It's my understanding that only "first generation" Prescotts will be socket 478 compatible. Do you (or anyone else) know what the anticipated "top speed" socket 478 prescott is currently supposed to be?

 

[Natoma]

Officially 3.8Ghz is the top speed for the socket 478 Prescott. However, it should be noted that I read a report on The Register (so take it for what it's worth), that Intel is considering making the top speed 4.5Ghz for Prescott on the 800Mhz FSB, at which point it will move to a 1.066Ghz FSB for Tejas.

However, this speculation was dependent on AMD executing with Athlon 64. If they execute poorly, Intel will apparently stay with the 800Mhz FSB and boost it to 4.5Ghz on socket478. If AMD executes well, they will bring out Tejas and phase Prescott on socket478 out faster.

Anyways, again, officially 3.8Ghz is the top rated speed for Prescott at 800Mhz FSB and socket478.

As for the Prescott support, yea, the only board that supports Prescott officially is the 8KNXP. Again, make sure you search for VRM10.0 or FMB1.5. If it's FMB1.0, then it will absolutely not support Prescott, no matter what the PR says. Take Asus and Abit for example of companies who say they have "future" support, but when you look at the manuals for their high end boards, they only support FMB1.0.

 

[Joe DeFuria]

Thanks for the info, Natoma!

I'll have to sit and think on this one a bit....there's probably going to be lots of new product announcements during Computex too...

 

[Natoma]

Heh. Don't get caught in the cycle of waiting for the thing that's "just around the corner"...

 

[Joe DeFuria]

Well, it seems that Half-Life 2 has now been "officially" delayed:

http://www.shacknews.com/onearticle.x/28497

This is actually "good news" for me....I can put off building my next rig a few more months yet!

This means there's a possibility I may be able to go with an Athlon 3200 or a 3 - 3.2 Ghz Prescott.

 

[stevem]

If you want support for Prescott, your only choice at this point is the 8KNXP. Whatever you purchase, make sure that it supports FMB 1.5 aka VRM10.0. If it doesn't support that, then it won't have the voltage support for Prescott and you're stuck with a P4-3.2Ghz as your highest upgrade.

VRM10.0!=FMB1.5

All i865/i875 motherboards support VRM spec 10.0 & FMB 1.0 for current P4C certification. Some existing motherboards are also likely to have VRM capacity for lower clocked Prescott PGA478. Boards with 4-phase power supplies should have an advantage here. The Gigabyte 8KNXP has an additional plugin VRM called DPS2, which allows for 6-phase supply. Total output is then 150W, which is greater than the reputed TPR of Prescott. From Intel's mods (for FMB 1.5 support) on their existing motherboards, we can see that they've only uprated their output sections (& BIOSes). In theory, the 8KNXP should be OK.

We also know from a pure VREG standoint that overclocked P4Cs are drawing in excess of 100W on existing boards. Hence, it may be possible for lower clocked PGA478 Prescotts to work in existing systems. It now depends on compliance with timings, tolerances, etc.

As for the Prescott support, yea, the only board that supports Prescott officially is the 8KNXP. Again, make sure you search for VRM10.0 or FMB1.5. If it's FMB1.0, then it will absolutely not support Prescott, no matter what the PR says. Take Asus and Abit for example of companies who say they have "future" support, but when you look at the manuals for their high end boards, they only support FMB1.0.

Where did you see official Gigabyte support for Prescott on this board? (Don't tell me it's in the manual...). Given the FMB 1.5 spec was recently released, I suspect that some vendor's boards may comply. In no case can we be definitive on Prescott PGA478 support or lack thereof at this stage, however...

BTW, I agree re: ALC655 audio (much chagrin). Grab the latest drivers from Realtek if you haven't already. Quite reasonable.

 

[Natoma]

If you want to get technical fine, FMB1.5 is not equal to VRM10.0, but FMB1.5 is in the design specification for VRM10.0.

Second, part of the FMB1.5 spec is lower, stable, voltages. Any board that supports a voltage range from 0.8325 - 1.6 is FMB1.5 compliant, which the 8KNXP is (to my chagrin when it comes to overclocking my current P4C. ).

Official support for Prescott was never stated. I read the specs for FMB1.5, read the specs for the 8KNXP, and deduced that it is indeed compliant, a few months after I purchased the board mind you.

 

[stevem]

Second, part of the FMB1.5 spec is lower, stable, voltages. Any board that supports a voltage range from 0.8325 - 1.6 is FMB1.5 compliant, which the 8KNXP is (to my chagrin when it comes to overclocking my current P4C. )

FMB 1.5 spec is more than just V range. What about I? How do you know that the rise-fall times/slew/switching, etc are compliant? Stability under Prescott's load could be completely different to Northwood. Mind you, if any current board can (PTP), it's likely to be one with DPS... FMB 1.5 is just an interim (unplanned?) measure on the path to FMB 2.0 for LGA775 Prescott.

The 8KNXP isn't ideal for serious OCing. You'd need a compliant Malay P4, good DDR, & PAT (auto) off... Gigabyte's boards aren't renowned for their OCing prowess. Stability/ODM is/was their forte.

BTW, don't you list a 3.0 @ 3.05 in your sig?

 

[Natoma]

First I should have said 0.8325 - 1.6 only is FMB1.5 compliant, not that it has that range only. FMB1.5 does not allow higher voltages than 1.6v in the spec. The DPS daughterboard regulates the voltage flow to take care of the Prescott issues considering the voltage range supported in the bios by the board. This part was covered in the manual, i.e. voltage regulation by the DPS daughterboard.

Second, the overclocking on the 8KNXP is bad mostly because of the limited voltage available to give to the processor and the memory (only 0.3 overvolt allowed. Damn!!). The 3.08Ghz is a 205Mhz FSB with PAT and TP on (DDR is at 2-3-2-5 instead of the normal 2-3-3-6 timing in the SPD due to TP). I ran Prime95 on it for 12 hours and it runs rock solid with no crashes of the program up to 205Mhz with those very aggressive timings. At 206 Prime95 will crash after about 6 hours and at 207 it'll crash after 20 minutes. If I reduce the memory timings to 2-3-3-6 I can get it to 215Mhz, but all the benchmarks are lower even with the higher FSB (3DMark, UT2K3 botmatch and flybys, SiSoft Sandra) so I leave it at 205Mhz with the aggressive timings.

No matter what the timings are, it's just not stable above 215Mhz and I attribute that to a lack of voltage above 1.6.

 

[stevem]

OK...

First I should have said 0.8325 - 1.6 only is FMB1.5 compliant, not that it has that range only. FMB1.5 does not allow higher voltages than 1.6v in the spec.

Upper/lower bounds aren't a problem per se. Epox, et. al support greater ranges now...

The DPS daughterboard regulates the voltage flow to take care of the Prescott issues considering the voltage range supported in the bios by the board. This part was covered in the manual, i.e. voltage regulation by the DPS daughterboard.

You do realize that VREG is accomplished by the 3-phase VRM on the motherboard in conjunction/parallel with the VRM on the DPS module (if it's plugged in)? Only FMB spec 1.0 was available when the engineers designed these boards. They'd have to have guessed at any future revision as FMB 1.5 wasn't finalized. Also note that there's no mention of Prescott in the manual (I dunno categorically - I've never read it ). You also realise that the manual isn't a technician's manual? So how do you deduce specific capabilities based on no info? Just curious.

Care to speculate on whether the two VRM modules operate in bridged mode or not, ie 60º or 120º phase?

Second, the overclocking on the 8KNXP is bad mostly because of the limited voltage available to give to the processor and the memory (only 0.3 overvolt allowed. Damn!!). The 3.08Ghz is a 205Mhz FSB with PAT and TP on (DDR is at 2-3-2-5 instead of the normal 2-3-3-6 timing in the SPD due to TP). I ran Prime95 on it for 12 hours and it runs rock solid with no crashes of the program up to 205Mhz with those very aggressive timings.

Yes, Gigabyte never really play fast & loose with BIOS options. How do we know there aren't trace/layout issues too. Perhaps an overly complicated board (multi I/O asics, DPS, etc) is a millstone for OCing? You should pass a DMM over the board when "Top Performance" is enabled. So much for default BIOS settings... Only Epox 4PCA & (perhaps?) Abit IC7 Max3 allow VDD >0.3V, so unless you're looking to volt-mod... In passing, the Max3 was designed with Prescott design guidelines in mind...

Addendum:

No matter what the timings are, it's just not stable above 215Mhz and I attribute that to a lack of voltage above 1.6.

Disable "Top Performance" mode & CPC & relax DDR. Both of mine OC to 3.6GHz Prime95/memtest86 stable. Haven't tested higher. I run them at default speed as they're HD HTPCs. "TP" mode overvolts & overclocks everything, including fixed AGP/PCI busses...

 

[Joe DeFuria]

Looking forward to seeing some reviews of this:

http://www.abit-usa.com/products/mb/products.php?categories=1&model=147

 

[stevem]

Looking forward to seeing some reviews of this:

http://www.abit-usa.com/products/mb/products.php?categories=1&model=147

Interesting. Rotated CPU socket. Portents of BTX?

 

[Joe DeFuria]

Specifications wise, it seems to meet my needs almost perfectly...just everything I want, with almost nothing that I don't. Though I'm not directly interested in the "uGuru" aspect of the board...it could be fun to tinker around with.

* Overtly Claims Prescott Support
* Does NOT include a 3rd party Serial ATI-RAID controller (that I can tell). It uses the ICH5-5 Southbridge AFAIK.
* Realtek Codec for sound
* On-board IEE# 1394
* On-Board 10/100 Ethernet.

Hopefully, the street price is sub $150. Though being the "first uGuru" board has a potential negative price impact...

 

[Natoma]

OK...

First I should have said 0.8325 - 1.6 only is FMB1.5 compliant, not that it has that range only. FMB1.5 does not allow higher voltages than 1.6v in the spec.

Upper/lower bounds aren't a problem per se. Epox, et. al support greater ranges now...

Agreed. Merely stating that that is what the spec "requires".

The DPS daughterboard regulates the voltage flow to take care of the Prescott issues considering the voltage range supported in the bios by the board. This part was covered in the manual, i.e. voltage regulation by the DPS daughterboard.

You do realize that VREG is accomplished by the 3-phase VRM on the motherboard in conjunction/parallel with the VRM on the DPS module (if it's plugged in)? Only FMB spec 1.0 was available when the engineers designed these boards. They'd have to have guessed at any future revision as FMB 1.5 wasn't finalized. Also note that there's no mention of Prescott in the manual (I dunno categorically - I've never read it ). You also realise that the manual isn't a technician's manual? So how do you deduce specific capabilities based on no info? Just curious.

I seriously doubt Intel didn't work with the motherboard manufacturers when designing the FMB1.5 specifications. However, there is mention in the manual (and on the box btw), of VRM10.0 compliancy when using the DPS2 daughtercard VRM. Now whether or not this turned out to be a lucky guess on Gigabyte's part or not is moot imo because the board has all the prerequisites to run prescott just fine.

Care to speculate on whether the two VRM modules operate in bridged mode or not, ie 60º or 120º phase?

This I wouldn't know. There's no documentation on it in the manual, the box, or the gigabyte website.

Second, the overclocking on the 8KNXP is bad mostly because of the limited voltage available to give to the processor and the memory (only 0.3 overvolt allowed. Damn!!). The 3.08Ghz is a 205Mhz FSB with PAT and TP on (DDR is at 2-3-2-5 instead of the normal 2-3-3-6 timing in the SPD due to TP). I ran Prime95 on it for 12 hours and it runs rock solid with no crashes of the program up to 205Mhz with those very aggressive timings.

Yes, Gigabyte never really play fast & loose with BIOS options. How do we know there aren't trace/layout issues too. Perhaps an overly complicated board (multi I/O asics, DPS, etc) is a millstone for OCing? You should pass a DMM over the board when "Top Performance" is enabled. So much for default BIOS settings... Only Epox 4PCA & (perhaps?) Abit IC7 Max3 allow VDD >0.3V, so unless you're looking to volt-mod... In passing, the Max3 was designed with Prescott design guidelines in mind...

Considering the only addition to the Gigabyte board over other boards is the DPS2, I wouldn't attribute layout or trace issues to lack of overclocking. The rest of the board looks exactly the same as other i875p mobos.

Question though. Why do you suggest I use a DMM when TP is enabled? You think there's something awry going with the voltages being too far out of spec? I use motherboard monitor and they all appear within range.

Addendum:

No matter what the timings are, it's just not stable above 215Mhz and I attribute that to a lack of voltage above 1.6.

Disable "Top Performance" mode & CPC & relax DDR. Both of mine OC to 3.6GHz Prime95/memtest86 stable. Haven't tested higher. I run them at default speed as they're HD HTPCs. "TP" mode overvolts & overclocks everything, including fixed AGP/PCI busses...

CPC is disabled by default on the 8knxp. I get 215Mhz when I disable TP and relax the DDR timings. I was able to get it to boot into windows and run a few tests at 230Mhz FSB, but I had to change the memory:fsb ratio to 1.66 instead of 2.0.

 

[stevem]

I seriously doubt Intel didn't work with the motherboard manufacturers when designing the FMB1.5 specifications. However, there is mention in the manual (and on the box btw), of VRM10.0 compliancy when using the DPS2 daughtercard VRM. Now whether or not this turned out to be a lucky guess on Gigabyte's part or not is moot imo because the board has all the prerequisites to run prescott just fine.

Um... Perhaps we're talking cross purposes? You seem to forget that the 8KNXP & other original release i865/i875 boards debuted in Q203. The design process for eg 8KNXP started last year & was finalised by Q103 (mass production). Check the rev of your board. It's still @ 1.0. No updates since then. FMB 1.5 was finalised in Q303. Get it? Of course Intel eng/dev support their OEMs, but how do you expect them quantify a spec before it's a spec? The question then remains, why did Gigabyte embark on their DPS escapade? Marketing. A shift (forced) in focus from ODM to retail. As a top tier OEM they are privy to all sorts of nifty details. Given the TDP requirements of current & future Intel & AMD CPUs, & their extensive knowledge of the motherboard business it probably made sense. Not to the bean-counters, though! However, at no time did they officially detail support for Prescott. It could be that a single trace or component on the motherboard or a compromise in layout necesitates a rev 1.1 8KNXP for Prescott support. How good's the DPS to you then...? (Actually, it may be that a revised DPS will/may do the trick - who knows, support or lack thereof is sheer speculation right now...)

What has VRM 10 got to do with DPS? (Rhetorical) Nothing. I thought we'd established our formula earlier. If you don't install the DPS module the board is still VRM 10 FMB 1.0 certified. Nothing says it's FMB 1.5 once installed either... You've not established any prerequisites necessarily met at all...

This I wouldn't know. There's no documentation on it in the manual, the box, or the gigabyte website.

Bingo! As per every other piece of your "Prescott support" thesis for this board... Don't get me wrong, it probably does support Prescott, but not officially (yet) & nothing as definitive as you've portrayed.

Considering the only addition to the Gigabyte board over other boards is the DPS2, I wouldn't attribute layout or trace issues to lack of overclocking. The rest of the board looks exactly the same as other i875p mobos.

Ke...? You're telling me, sans DPS, an ASUS P4C800 is identical to the 8KNXP...? Exactly the same? You see no layout/trace/component dissimilarities at all? I think the Abit crew should read this & give up their day jobs...

Question though. Why do you suggest I use a DMM when TP is enabled? You think there's something awry going with the voltages being too far out of spec? I use motherboard monitor and they all appear within range.

IIRC, TP enabled:
1. overlcocks AGP/PCI to 72/36 & overvolts AGP by +0.1V.
2. overclocks FSB to 201
3. overvolts MCH +0.1V (probably due to 1.)
4. VDD is 2.65V
5. overvolts CPU 0.0250V-0.0500V

All this while it tells you your "defaults" are default... What happens when you now change default VDD to +0.3...?

MBM depends on the ini file for configuration. How good is the software interface with the ITE 8712F sensor? How good is Gigabyte's board level implementation? BIOS? Which monitoring ports are enabled, etc, etc? What temp does 25C portray in everybody's system...?

Another little quirk is that TP can only be enabled if CL=2. By design, according to Gigabyte tech.

CPC is disabled by default on the 8knxp.

No it ain't... Guess what? It's your components/setup...

I get 215Mhz when I disable TP and relax the DDR timings. I was able to get it to boot into windows and run a few tests at 230Mhz FSB, but I had to change the memory:fsb ratio to 1.66 instead of 2.0.

How long did you research...? Perhaps we should take this to PM land.

Overall, you're pulling my leg...? Right...? You wanted to outpoint Russ/Demalion?

 

[Natoma]

What has VRM 10 got to do with DPS? (Rhetorical) Nothing. I thought we'd established our formula earlier. If you don't install the DPS module the board is still VRM 10 FMB 1.0 certified. Nothing says it's FMB 1.5 once installed either... You've not established any prerequisites necessarily met at all...

This I wouldn't know. There's no documentation on it in the manual, the box, or the gigabyte website.

Bingo! As per every other piece of your "Prescott support" thesis for this board... Don't get me wrong, it probably does support Prescott, but not officially (yet) & nothing as definitive as you've portrayed.

The 8knxp box and manual both state that the board is only VRM10.0 compliant when the daughtercard is used. Considering FMB1.5 is part of the VRM10.0 spec, it deems conclusive that that means using the daughtercard will enable FMB1.5 which is a prerequisite for Prescott.

Considering the only addition to the Gigabyte board over other boards is the DPS2, I wouldn't attribute layout or trace issues to lack of overclocking. The rest of the board looks exactly the same as other i875p mobos.

Ke...? You're telling me, sans DPS, an ASUS P4C800 is identical to the 8KNXP...? Exactly the same? You see no layout/trace/component dissimilarities at all? I think the Abit crew should read this & give up their day jobs...

I said it looks exactly the same, not that it is exactly the same. The layout of the P4C800 looks identical to the 8KNXP save for the DPS and the fact that the 8KNXP has 6 DIMMS while the P4C800 only has 4.

Question though. Why do you suggest I use a DMM when TP is enabled? You think there's something awry going with the voltages being too far out of spec? I use motherboard monitor and they all appear within range.

IIRC, TP enabled:

1. overlcocks AGP/PCI to 72/36 & overvolts AGP by +0.1V.
2. overclocks FSB to 201
3. overvolts MCH +0.1V (probably due to 1.)
4. VDD is 2.65V
5. overvolts CPU 0.0250V-0.0500V

All this while it tells you your "defaults" are default... What happens when you now change default VDD to +0.3...?

Where'd you get this information?

MBM depends on the ini file for configuration. How good is the software interface with the ITE 8712F sensor? How good is Gigabyte's board level implementation? BIOS? Which monitoring ports are enabled, etc, etc? What temp does 25C portray in everybody's system...?

I actually had to change the default sensor in order to get accurate readings. I'm at work right now so I can't give you the sensor name.

Another little quirk is that TP can only be enabled if CL=2. By design, according to Gigabyte tech.

Where are you getting this? I haven't been able to find any documentation on TP anywhere.

CPC is disabled by default on the 8knxp.

No it ain't... Guess what? It's your components/setup...

So the bios is lying to me?? :?

I get 215Mhz when I disable TP and relax the DDR timings. I was able to get it to boot into windows and run a few tests at 230Mhz FSB, but I had to change the memory:fsb ratio to 1.66 instead of 2.0.

How long did you research...? Perhaps we should take this to PM land.

Overall, you're pulling my leg...? Right...? You wanted to outpoint Russ/Demalion?

You forgot WaltC and DaveH in that one.... Yes we can take this to PM since we aren't even discussing the original topic anymore anyways.