Prescott and L1 cache problems

batboy · Jan 27, 2004

Perhaps some of the initial reports saying Prescott's performance is less than stellar is due to the fact that most of the current motherboards are having problems utilizing the L1 cache.

Fugger over at Extreme Forums said this:

Lots of people will be reviewing these things and are clueless to the L1 cache problem on a lot of boards including the IC7 MAX3.

With the L1 no present the chip are crippled and I picture lots of AMD fans taking advantage of that to bench against AMD rigs to make them look good because the general public does not know how to look at the missing L1 cache and know the diff.

I have pointed this out many times on the forum, missing L1 cache hurts performance bad. I have been working with Abit on the IC7 Max3 problem and the best they can do is remove Prescott from the supported list...

Good thing I've been holding off buying a new mobo. Sounds like Asus does recognize the L1 cache, but has other issues. DFI is working on BIOS and hardware revisions to fix the Prescott L1 bug. Sounds like everyone that has tested a Prescott agrees it runs a little warmer than the Northwood. No surprise to me since they use more watts.

That's the bad news, now the good... Fugger also said this:

Clock for clock the Prescott is faster than the Northwood due to the increased cache and branch prediction but the longer pipeline has its drawbacks. But this is done in order to ramp up speeds.

If you don't know who Fugger is, he is an Intel CPU Gawd. He owns a record breaking rig and has tested the Prescott. Since he's under a NDA from Intel, he can't publish his result until the official release date. The whole forum thread can be found at this link:

http://www.xtremesystems.org/forums/showthread.php?s=&threadid=27233&perpage=25&pagenumber=1

xtrmeocr · Jan 27, 2004

hes also notorious for making claims and never backing them up

oc_newbee · Jan 27, 2004

xtrmeocr said:
hes also notorious for making claims and never backing them up

So your saying that what Fugger said is non sense. Why?

batboy · Jan 27, 2004

Everyone is entitled to their own opinions, but I think Fugger is quite knowledgeable. We'll see what he posts about the Prescott in a week and then we can decide if he was able to back up what he said with benchmarks and other proof.

oc_newbee · Jan 27, 2004

batboy said:
Everyone is entitled to their own opinions, but I think Fugger is quite knowledgeable. We'll see what he posts about the Prescott in a week and then we can decide if he was able to back up what he said with benchmarks and other proof.

I agree with you on that one. With all the knowledge that Fuger got, I know that he could tell which is good and what not. So let just hope that those numbers won't let us all down.

DayUSeX · Jan 28, 2004

yeah weird, i dont know him so he could be an intel fanboy, but i find that rahter odd that these companies woudlnt know about hte "bug" before they were released

Flakk · Jan 28, 2004

batboy said:
Good thing I've been holding off buying a new mobo. Sounds like Asus does recognize the L1 cache, but has other issues.

asus p4p800 has issues with the prescott? whats wrong? and... is it possible they will fix what's wrong in bios updates?

does this mean i can't keep my p4p800 dlx if i want to see full performance from a prescott cpu?

dropadrop · Jan 28, 2004

DayUSeX said:
yeah weird, i dont know him so he could be an intel fanboy, but i find that rahter odd that these companies woudlnt know about hte "bug" before they were released

He works for intel.

Krowa 02 · Jan 28, 2004

Whatever he says I take with a grain of salt, becuase as mentioned

But I do hope they fix prescott for my ic7 cause id sure like to use one in the future

.

nikhsub1 · Jan 28, 2004

It seems as if the bug is with the mobos not recognizing the extra cache, not a bug with the CPU. Am I missing something?

batboy · Jan 28, 2004

That's my understanding too, that's it's not a problem with the CPU itself. I'm not sure we know for sure, but it's sounding like several of the motherboards are not recognizing the L1 cache correctly. Whether this can be fixed with a BIOS update or if it requires a hardware revision is a different question. The point I draw from this is it might explain a little bit about why the Prescott is not performing as well as some folks had hoped. Let's hope that's the situation. I'll be glad when Feb. 2 rolls around and we can start getting some real data on the Prescott. This being left in the dark and all the second hand speculation is killing me.

Albigger · Jan 28, 2004

nikhsub1 - from what I understand, you are correct in thinking that the bug is with the motherboards. I believe the one board (DFI?) has issued a hardware and BIOS update fix for the problem.

I am interested to see correct (with L1 cache enabled) results from this prescott, and trust most of Fugger's comments - for the most part he seems to know what he's talking about. However others think that he makes claims he can't back up, and I do recall him saying something about 5-5.5 GHz, and as for that, I'll believe it when I see it.

markodude · Jan 28, 2004

Batboy - know how you feel, I was this far (markodude makes his fingers show a very small distance) off getting one last night from my contact, but he could not get hold of anyone and is off to the states for a big conference today and will be back in a week by which time we should know what the situation is anyway.
He was certainly miffed when I showed him the inquirers benchmarks, but do we know if they had a board which ran L1 cache properly - I will ask Mike if he knows now.
If 5ghz is possible it will be on some crazy cooling, I reckon s775 b4 we see 5.5ghz on sustainable cooling.

oc_newbee · Jan 28, 2004

Originally posted by Albigger
and I do recall him saying something about 5-5.5 GHz, and as for that, I'll believe it when I see it. [/B]

I don't remember him saying that prescott will reach 5-5.5ghz in the link above. Those word were coming from someone else not Fugger. Don't quote me on this but, he might have said that with the 3.4 Prescott and some super duper good cooling then it might be able to reach mid or high 4 gig, but never anything like 5 or above ghz.

xtrmeocr · Jan 28, 2004

one of the reviews out there was done on a LGA755 prescott and a socket T gransdale mobo. they got pretty much they same numbers, are we to believe that intels own socket t motherboards also dont see the L1 cache?

batboy · Jan 28, 2004

Good question. Are the intial benchmarks with or without the L1 cache working properly? Not sure anyone knows that yet.

The 5 gig speculation was what someone thought it would take to beat the Super Pi record I think. The Prescott will ramp up to abpout 4.2 gig by the end of the year. At that time a super cooled Prescott will probably make 5 gig. Right now they are a new immature core. We will be lucky to get 4 gig out of them for the first couple months.

I'm assuming from his comments that Xtrmeocr is an AMD fanboy. Therefore I take anything he says with a grain of salt... maybe the whole salt shaker. I love how they like to mention that "one of the reviews out there..." but they never post a link.

batboy · Jan 28, 2004

Here is a new article comparing the 3.2 Northwood to the 3.2 Prescott. Still no mention on whether the L1 cache is working right though. The results look bad for the Prescott at first glance.

http://www.theinquirer.net/?article=13847

Albigger · Jan 28, 2004

oc_newbee said:
I don't remember him saying that prescott will reach 5-5.5ghz in the link above. Those word were coming from someone else not Fugger. Don't quote me on this but, he might have said that with the 3.4 Prescott and some super duper good cooling then it might be able to reach mid or high 4 gig, but never anything like 5 or above ghz.

Originally posted by Fugger on Xtremesystems
I think 5.5Ghz will not be a problem for the chip rather the mobo's FSB capabilities.

I am not sure if the problem can be fixed in bios or it is a chipset problem from that rev.

...page 3 of the thread that was linked, and:

Originally posted by Fugger on Xtremesystems
Once we start hitting 5Ghz + AMD will lose pi rankings quicly.

...page 4 of the thread, and:

Originally posted by Fugger on Xtremesystems
It will take 5Ghz + to take the top spots of pi with the prescott core.

5.5Ghz will take out number 1

Its not a matter of if, but rather when. There are issues that need to be worked out and trust me you will have all the answers here.

...page 5 of the thread.

I'm not sure on his timescale, but I'm sure that won't happen without super cooling and I don't think it'll happen very soon (at least not in the next few months). Though he's had his hands on one, and I'm just speculating, so what do I know?

Also, I would have to agree with xtrmeocr if there really is a review like he says (link please?)

Besides that, prescotts have been around for a while, there probably would have been some mention of this problem (if it really is a problem) a little earlier than 1 week before release of the product. As JCVIggen says from that thread on Xtreme:

Originally posted by JCViggen on Xtremesystems

Originally posted by skate2snow on Xtremesystems
But like FUGGER said: P4C800 can monitor the L1 but its not working.

He said something like that.

Click to expand...

I talked to ASUS in Taiwan and they told me that was BS.

They have been working with prescotts for a while, it is not possible for the board to not use the L1 cache.

OC Noob · Jan 28, 2004

The "L1 chache bug" sounds like an Intel fan grasping at straws, but for the sake of my IC7-G I hope its true. I got this thing because the 64 bit AMDs were going to require a new socket and most people believed prescott would run in 875 socket 478 boards. Comon sense dictates a bug like this would be found since mobo companies have been testing the chips for a while. I guess we'll see. Hopefully the chips will be atleast be better then Northwoods and OC like a mother. Although the IC7-G doesn't support it officially I'll bet it will still work and I plan on getting one to make my 4 ghz dream come true.

So they didn't add any extra registers or anything like that to Prescott, heh?

Flakk · Jan 28, 2004

OC Noob said:
The "L1 chache bug" sounds like an Intel fan grasping at straws

true it kind of does, but it also makes some sense and seems as likely as not. because as with any new cpu, a new bios revision must be written to support it. to make use of the extreme edition p4, you must be using a certain bios revision, and the same will be with prescotts. they're not officially out yet, so maybe motherboard companies will release it when its time... yes i could be grasping at straws =) you'd think that mobo companies would have this bios revision out already for the few people who are testing these cpu's early

it's hard to wait, but we just have to wait and see. go enjoy life for a week and come back, and all the answers will more than likely be here

Prescott and L1 cache problems

Senior Moment

Member

Member

Senior Moment

Member

Member

Member

Member

Member

Unoriginal Macho Moderator

Senior Moment

Member

Member

Member

Member

Senior Moment

Senior Moment

Member

Member

Member

Similar threads