Intel 865 chipset and PAT - is it real?

[joecool]

Folks,

I've been reading on multiple web sites about the rumors that PAT can be enabled on 865-based motherboards with a BIOS update from the manufacturer. So far it sounds like ASUS and Abit have release BIOS versions for their 865-based mobo's that can benchmark at or ABOVE many 875-based boards. I'd like to get to the bottom of this and find out if it's true. If you have any information or first-hand experience with this please post here. If this is true the 865 boards could turn out to be one of the major deals of the year!!!

-Joe

PS - AT Mods, can we sticky this at the top of the forum?

 

[Vette73]

Have you seen this yet....

turn the 865 to a 875

I bet 875 sales will be dropping or the 875 prices will come down, one or both.

 

[joecool]

anybody have any thoughts on this?

 

[joecool]

come on, surely more folks have an opinion and/or experience with this!

 

[joecool]

man, i can't believe no one is replying to this ... seems to me this is the hot topic for now - we can all be getting a major deal on the 865 mobo's if they have PAT!!!

 

[wicktron]

AFAIK:
It's not PAT, it's a hack for memory optimization. It is a big deal, but in an overclocked situation, Canterwood still shines. You see, on mobos such as the P4P800 and the IS-7 which have this hack, the trick only works if and only if you run a 1:1 FSB to memory ratio. What if you were to overclock to say a 300MHz FSB. I highly doubt anyone has the RAM to run at DDR600. Thus, running a ratio to decrease RAM frequency is the only realistic scenario in this situation. In that case, the "hack" is off and the Canterwood runs circles around the Springdale.

This hack would be well worth it if somehow Asus or Abit can make it such that it works when running an asynchronous bus. Then and only then will I buy into the notion that a Springdale is a better buy than a Canterwood.

BTW, I own a P4P800.

 

[Tom]

A Canterwood isn't going to "run circles" around a Springdale, even if the Springdales didn't have "PAT" or the memory optimization as you call it.

Without this new development under ideal circumstances there was only a few percentage points difference.

 

[Vette73]

Well abit, asus, and others are said to be coming out with BIOS updatyes soon to allow the 865 to perform like a 875.

Of course intel is a little mad now as they won't their extra money, so will have to see if they come out or not now.

 

[joecool]

Originally posted by: wixt0r
AFAIK:
It's not PAT, it's a hack for memory optimization. It is a big deal, but in an overclocked situation, Canterwood still shines. You see, on mobos such as the P4P800 and the IS-7 which have this hack, the trick only works if and only if you run a 1:1 FSB to memory ratio. What if you were to overclock to say a 300MHz FSB. I highly doubt anyone has the RAM to run at DDR600. Thus, running a ratio to decrease RAM frequency is the only realistic scenario in this situation. In that case, the "hack" is off and the Canterwood runs circles around the Springdale.

This hack would be well worth it if somehow Asus or Abit can make it such that it works when running an asynchronous bus. Then and only then will I buy into the notion that a Springdale is a better buy than a Canterwood.

BTW, I own a P4P800.

hey, i'm just not sure this is true. here's what AT had to say about it in their just-published 865/875 roundup:

It doesn't take much effort to reap the benefits of PAT-like performance from the IS7, simply make sure that your memory timings are as reliably aggressive as your modules are able to withstand, the "PSB533" setting is enabled under "N/B strap" if you're using an 800MHz FSB processor, the 1:1 ratio is enabled along with the "fixed" option for your AGP/PCI bus, and the CPC (Command Per Clock) is enabled in the Advanced Chipset Features section. This applies to users that will be running their IS7 at stock speeds, but if you're overclocking (especially past 250MHz FSB) make sure you use change N/B strap back to PSB800 so as to not max out your memory too quickly. We suggest the 3:2 ratio if you're going well past 250MHz FSB.

note that they DON'T say you can't overclock; in fact, they don't even say you're stuck with the 1:1 fsb bus speed. sound to me like it's just a matter of tweaking the various settings and once you find the sweet spot the is7 will be just as good, if not better, then the ic7. anyway, i still haven't seen any benchies of both boards maxed out with the same cpu, so until someone does that i guess it's hard to say which board is the winner. still, for $40 less WITH LAN, i'm thinking it's time to trade in my ic7 for an is7 ... i doubt i'll be disapointed!

 

[extro]

I have yet to see anything that debunks the widely held belief that the Canterwood (875) has the same limitations in running with PAT enabled as the Springdale (865), i.e. that DRAM speed has to be synchronous with the CPU (1:1) or PAT can't be enabled. If that's the case, then an 865 and an 875 will be equal in overclocking ability because PAT never becomes a factor on either when the CPU/DRAM ratio is changed from 1:1.

It has now become easy to demonstrate that this is true on Springdale because the newest BIOSes allow the user to turn "PAT" on and off manually, and no performance improvement has been observed when running memory at anything other than 1:1.

Intel's own description of PAT doesn't much differ from Asus' or Abit's description of their optimization. Intel says, "Intel® Performance Acceleration Technology increases memory and system-level performance by optimizing internal data paths. " Well, isn't that what Asus and Abit have said their optimizations do?

What's the truth? We can only deduce because Intel and the motherboard makers are keeping their mouths shut about it. Intel hasn't said what the exact criteria and limitations are for enabling PAT on the 875 beyond this description, "PAT is only available when the MCH is set at FSB 800 MHz and DDR 400 MHz mode" and "PAT is unique to the 875P chipset. Advanced manufacturing and test technology ensures the ability to implement enhanced logic design. These enhancements allow performance improvements of two clock cycles. These include one clock improvement to memory access and one clock improvement in DRAM chip select. These changes are internal to the chipset, allowing the external interfaces to run at standard specifications, while accelerating the internal chipset logic."

I think it's pretty clear that Intel didn't sufficiently disable PAT in the current 865 chipset packaging, though they may very well do so in a future stepping (caveat emptor), and board makers have found a way to turn it on. Intel being the master of producing great yields is probably binning a lot of perfectly good 875's as 865's to satisfy consumer demand for the lower cost and higher demand product. For all we know, the speed binning story about the 865 is just marketing bull from Intel and they're all perfectly able to run PAT enabled - for now anyway.

 

[GetInMyFatBelly]

With Intels release of a statement on this, I would run and download the beta bioses now for the boards your thinking about possibly getting. I can see sites now specializing in the PAT enable bioses for 865, because none of the companies no longer have them for download.

 

[Dance123]

Which sites are you talking about? Does it mean that if you download the latest BIOS update from the Asus site for the P4P800 you won't have PAT enabled?!

And what do you think about following article I found: Intel says there's no such thing as PAT

 

[fredtam]

I think I will wait a week to see how this all turns out. Even though I would love to get that kind of performance from a $110 board I would hate to get screwed in future bios updates. Intel Statement

 

[Tom]

After reading Intel's techno-speak laced warning, I'm more confident than ever that there isn't any difference between a Springdale chipset and a Canterwood chipset, except that Intel disabled PAT in the Springdale, for marketing reasons, not because the Springdale isn't capable.

Now, in the future that situation might change, since Intel obviously isn't too happy.

 

[fredtam]

Can someone explain to me why if it is just the unlocking of PAT in the 865 the boards do not have equal performance. It would seem that since the 865 and 875 are the same chip enabling PAT on the 865 would give it equal performance to the 875 yet it generally outperforms it even if it is by a small margin.

 

[Vette73]

Originally posted by: fredtam
Can someone explain to me why if it is just the unlocking of PAT in the 865 the boards do not have equal performance. It would seem that since the 865 and 875 are the same chip enabling PAT on the 865 would give it equal performance to the 875 yet it generally outperforms it even if it is by a small margin.

Where did you see this??

As much as I don't like it, tomshardware.com shows the 865 to perform the same as a 875 if everything is the same and the 865 has PAT on.

 

[fredtam]

Originally posted by: Marlin1975

Where did you see this??

In the roundup.

"Just like the IS7, the IS7-G is yet another motherboard that has successfully mimicked PAT (Performance Acceleration Technology) code previously exclusive to 875P motherboards. As a result the IS7-G is not only slightly faster than every 875P motherboard we've tested but, ironically, also faster than ABIT's own 875P motherboard, the IC7/IC7-G."

Quote from TH.

"Following on, in our last test, the Asus P4P800, equipped with the Intel 865PE chipset (Springdale), placed at the head of the pack in the benchmarks"

TH only tested the P4P800 with PAT enabled because Abit and others had not caught on yet. In the majority of the bench marks (of interest to me) the P4P800 outperformed all 875 boards with the exception of the P4C800. Even against the P4C800 it came out faster in several benchmarks.

 

[Dance123]

The main question is if you buy an Asus P4P800 in stores, will it have PAT enabled or not. Could it be that Asus will disable it again, because Intel could otherwise sue them. Are they legally allowed to leave PAT actived on the P4P800, but perhaps not call it PAT. I believe it has been said that PAT was enabled in press examples of the board, but what about the ones we gonna buy in stores??

Simply put, does BIOS 1007, which is the latest I believe, have PAT on or off and would there be a way to verify this somehow?

 

[joecool]

Originally posted by: fredtam
I think I will wait a week to see how this all turns out. Even though I would love to get that kind of performance from a $110 board I would hate to get screwed in future bios updates. Intel Statement

you make a very good point here ... i think there is a strong possibility that this performance boost for the 865 boards will be pulled at some point. additionally, as you also noticed, there is def. SOMETHING different between the two chipsets or they could both oc the same. overall, does seem like the 875's may be worth a $20-$30 premium for anyone who wants the maximum oc performance. my only issue at this point is that i need onboard lan. the abit is7 (865) has it and the ic7 (875) doesn't. tough call ... !

 

[Tom]

"SOMETHING different between the two chipsets or they could both oc the same. "

They do overclock the same as far as I've heard or seen. Why do you make this statement ?