Official Presler/Cedar Mill Overclocking & Issues Thread

So far based off what we know the initial reports are somewhat mixed. The number of people who have actually bought these chips are small so the odds that either something Ross has done caused his chip not to scale well or that he recieved a bum chip are pretty high.

We will know soon enough in the coming days. Batboy's chip should arrive on monday and mine around that time period as well.

______________________

READ THIS BEFORE YOU BUY

If you do not have the following items do NOT buy a Pentium D this applies to both the 8xx series and especially the 9xx series. (you will only be asking for trouble)

• You MUST have a ATX 2.0 compatible PSU and/or proper adaptors

• It must have both a 24pin ATX, 8pin P4 connector, and a 4 pin molex or adaptors needed to convert improper connections to the ones listed here

• Yes you MUST connect ALL 3 power connectors to get stable results

• Your PSU MUST have atleast 30A on the 12v+ (or split but single rail is always prefered)

• You MUST have adequate cooling (XP120, Ninja Scythe, TT Typhoon and above. Zalman doesnt cut it)

• You MUST be fully prepared to dissapate 150W of heat from this CPU for proper operation

• Water Cooling IS prefered and reccomended

• You MUST be careful when applying voltage to the CPU. These chips pull incredible amounts of power and you WILL kill your motherboard if you do not sink/cool the FETs above 1.5v

Rolling Averages

Review Sites:

=Default Voltage=
4.16ghz @ 1.3v - Pentium D 955XE (Legit Reviews)
4.02ghz @ 1.3v - Pentium D 955XE (Tom's Hardware)
4.00ghz @ 1.3v - Pentium D 955XE (Xbit)
3.50ghz @ 1.3v - Pentium D 955XE (Anandtech)

**AVERAGE: 3.92ghz on stock cooling @ stock volts

=Max On Stock Cooler=
4.76ghz @ N/A - Pentium D 955XE (Legit Reviews)
4.26ghz @ 1.37v - Pentium D 955XE (Xbit)
4.26ghz @ 1.44v - Pentium D 955XE (Tom's Hardware)
4.25ghz @ 1.37v - Pentium D 955XE (Anandtech)

**AVERAGE: 4.38ghz on stock cooling @ 1.39v
_______________________________________

OCF Members:

=Default Voltage=
3.73ghz @ 1.3v - Pentium D 930 (Ross)

=Max On Stock Cooler=
4.05ghz @ 1.45v - Pentium D 930 (Ross)

I wonder what the difference is between the top of the hill chip and your basic 9xx series chip that would cause this difference ?

diehrd said:

I wonder what the difference is between the top of the hill chip and your basic 9xx series chip that would cause this difference ?

Click to expand...

Well I am hoping there isnt otherwise im gonna have to order some KY before the Fedex guy gets here. If im gonna get pwned again im sure as hell going to ease the process anyway I can.

Most likely this is what you and I discussed, leakage issues from poor design. I dont think that suprises anyone here.

My greatest fear lies with Yonah. Im afraid it will have similar scaling issues making it DOA for people like us.

Sentential said:

Well I am hoping there isnt otherwise im gonna have to order some KY before the Fedex guy gets here. If im gonna get pwned again im sure as hell going to ease the process anyway I can.

Click to expand...

LOL Hopefully it doesn't come to that! Who knows, maybe you'll have better luck with yours. At least we all hope you do.

dylskee said:

LOL Hopefully it doesn't come to that! Who knows, maybe you'll have better luck with yours. At least we all hope you do.

Click to expand...

You and me both. I just wish the statistics were on my side

while the end results may not be what we were expecting or hoping for, i think its a little early to say. we've got one member with first hand experience with a retail chip. i think that's too small a pool to draw conclusions.

STILL i agree, things aren't looking too hot so far

I'm having trouble finding any other Presler results from production 920, 930, and 940 processors (not talking about extreme edition 955 or engineering samples). There must be other pioneers out on the net.

We have discussed the probability of speed binning in some of the other threads. I'm convinced that's part of the situation. Sent proposed some interesting speed binning theories in another thread. In a nut shell, he said some of the 930 CPUs are bound to be failed 940 cores, which I tend to agree. What remains to be seen is how many? Also, are a significant number of 920 CPUs also failed 930 cores?

Let's look back to exactly two years ago when the first Prescott core processors were released. I jumped on the first 2.8E available and was a bit disappointed. Intel still was working out the bugs of their new 90nm technology. Finally, after a couple of stepping releases and improvements, we starting seeing some nice overclocks.

Fast forward to the present time. It's possible that history repeats itself. The Presler core is brand new 65nm technology. Intel is basically using the Presler and Cedar Mill releases to perfect the 65nm cores for the up coming Conroe that will have a brand new architecture and should put Intel back in the saddle again.

So, as with most early production releases, the hype and expectation will probably exceed reality. The silver lining is that the Presler is better than the first dual core Smithfield. Early benchmarking indicates very nice improvements. Looks like that 2X2M cache is paying off and the next stepping will bound to be better.

For those looking for insane CPU clock speeds, you better wait for the bugs to be ironed out and get the single core Cedar Mill when it's released. I for one, hesitated on getting a 8XX, but I jumped on the Presler. I have a 940 on the way right now. Looks like several of us will be getting Preslers delivered next week. Mine will be here on Jan. 9th. Let's reserve final judgement when we start getting more info. As any good scientist or statistician will tell you (btw I'm a scientist), the more data you have... the better.

What do you guys want - 5ghz straight out of the box - seriously if Intel could do that dont you think they would market the chips at that speed? We know the process is doing well since the 960 has been announced at 3.6ghz Dual Core. It is only logical that we will see higher clock frequencies and benchmarks from 9xx than 8xx - and really 9xx is only a stop-gap until conroe.
Basically this is running a pair of 65nm die shrunk Irwindale Xeons on one socket - I almost got the exact same PCMARK score with 9xx I did when I was running Dual Xeons a year ago.....and the 9xx does not cost nearly as much as a pair of 3.6 2Mb Xeons, plus the boards are much cheaper and more overclocker friendly.
Most people have been running single core - dont you think the 9xx will put a lot more strain on your PSU/motherboard/VRMs at the same frequency since it is running 2 cores? Batboy are you considering upgrading your Antec 550w before getting this CPU? Worth considering as I reckon we can pull up to 20A with just the CPU when you get near 5ghz.
And remember 9xx is only officially supported at 1066FSB on the 975x chipset - perhaps there is a reason that we struggle to get the dual cores to such high FSBs, Intel had to do a board spin to support 1066FSB on the dual core even though such a speed was allowed on a single core processor on 955X and 925XE chipset.
Just some thoughts.....perhaps it will take a bit more effort, time, and modifications/upgrades before we see stellar overclocks from this CPU.

markodude said:

What do you guys want - 5ghz straight out of the box

Click to expand...

Umm, yeah!!
With that being said, you bring up some very good points markodude. I guess we were expecting a little too much considering people here were almost hitting 5GHz with phase with the 6xx and 8xx series so i guess we were expecting the magic 5GHz with the new series of chips. There is still a lot of people waiting for their turn, so time will tell.

I'm not a hardcore OC'er like a lot of people here so I'd be pretty happy with 4.0GHz before major tweaking. I am curious about the power consumption of my soon-to-be rig but 33-34A should be ok on the +12 I would think ..? (OCZ 520W is what I was planning on)

markodude said:

Batboy are you considering upgrading your Antec 550w before getting this CPU? Worth considering as I reckon we can pull up to 20A with just the CPU when you get near 5ghz.

Click to expand...

Not planning on upgrading the PSU unless I have to. These Antec units tend to be conservatively rated and it has handled my current 570 running at 4.4 gig without a hiccup (can bench at 4.5 gig with watercooling if I give it lots of voltage). My Presler 940 goal is to run at least 4.4 gig everyday, so that should not be any more watts than my current system.

DsN, as long as you get the Powerstream model 520w, you should be ok with your posted goals.

batboy said:

We have discussed the probability of speed binning in some of the other threads. I'm convinced that's part of the situation. Sent proposed some interesting speed binning theories in another thread. In a nut shell, he said some of the 930 CPUs are bound to be failed 940 cores, which I tend to agree. What remains to be seen is how many? Also, are a significant number of 920 CPUs also failed 930 cores?

Click to expand...

It all depends on how they bin and how often. Granted they will probably be more careful with the first lot but we will see. Having effectivly the whole series with a similar package date will give a fairly good indication of the overall health of Intel's 65nm process.

If all of our chips scale linerarly according to type that paints a pretty grim picture for Yonah and possibly Conroe if it faces similar issues. However if the numbers are across the board that generally tends to mean that yeilds are higher and that its simply luck of the draw.

Ross I hate to say it but I would immediately put your chip in either the for sale section here or ebay. Take a $30-50 loss and give it another spin once we figure out where the sweet spot is.

batboy said:

My Presler 940 goal is to run at least 4.4 gig everyday

Click to expand...

Right now the rolling average for 955XEs is just short of 4.4 on stock cooling. The odds of you hitting much higher than that are significantally in your favor. However in mine its going to get dicey. Either I will make out like a bandit or be doomed to defeat. I do not have the luxuary of a middle-ground for this processor. Its either going to be a home run or complete crap since Im pulling from either unbinned chips or the scum of the bucket.

My goal is to get at the very least 4ghz stable. Im hoping that I will atleast get that but I am very concerned that it will not happen. We will see soon enough.

Also we need to post where we bought our chips from. Ive seen many occasions where a store gets a very nice batch and another gets a lousy one.

Do we know the eta for single cores? if 9xx chips are gonna get stuck in a rut il go single core if its not gonna take forever....

Vrykyl said:

Do we know the eta for single cores? if 9xx chips are gonna get stuck in a rut il go single core if its not gonna take forever....

Click to expand...

They come much later as far as I know. As late as Q2 and Q3 of this year. They are a ways off which is why I sprung for a 920

The 6x1 are Cedar Mill - 65nm and available now or very soon.

Yeah, i didn't really go for the 8xx series when it was first released, but i have a 930 on the way from mwave.com myself. something about the new die shrink i find attractive. Its going to go on my p5wd2-p with my TDX danger den water block. I'll be sure to report my results when it arrives (tuesday the 10th) my 630 was never rock solid stable over 3.8-3.9, so if i can get a solid, stable 4.0ghz, i'll be satisfied. Anything higher would be a bonus.

Mikey - your benchmarks will be a lot better with the 930 at 4ghz than the 630 at 4ghz. I myself have gone from 3.46EE to 3.73EE to 840D to 950D in this same mobo, and improved on all benchmarks with each upgrade.
Plus multi-tasking is a lot better on the dual cores, here I am with the CPU running 2xSETI and 1xPrime95, its still perfectly usable

Yeah, i do benchmarks for fun, but its how it handles my every day tasks that i'm more worried about. i don't do anything super taxing other then some light 3d gaming.

I've heard from other people that the dual core is much smoother (hard to believe for how smooth everything already is) and i've figured that 2 cpu's running at 4.0ghz would be a lot more responsive then 2 virtual CPU's (HT enabled) at 4.0ghz

EDIT: Mark, the 950 does not have HT, right? its just the 955xe?