Cedar Mill Deleron at 5Ghz

[dexvx]

I was taking a break from finals, when I saw this on HardOCP, so I thought I'd repost it here:

http://www.matbe.com/actualites/11983/intel-celeron-cedar-mill/

They are based on the 65nm Cedar Mill core, with 512KB L2 cache. I don't know why people have a fetish with SuperPi, but with SuperPi 1M, it's more or less on par with a 2.6Ghz A64 or 4Ghz P4. I have no clue how it'd match an A64/P4 in real world apps, but it seems to fare far better than its Prescott brethren. I'm guessing a low-mid 3000+ A64 or a low 4Ghz P4.

Perhaps the most surprising thing is that it is on stock vcore with 50C CPU-Burn using a stock cooler.


Edit: SuperPi corrections

 

[coomar]

my opteron 146 at 3000 gets 26-27 seconds, its more on par with a fx-57 stock

 

[stevty2889]

The 65nm chips keep lookin like they will have some good OC potential, that could make them competitive again. I'm looking forward to replacing my water cooled POS pentium-D 830 with a Presler to see how high it will go. My 830 won't OC at all, but at least water cooled it's quiet and not a space heater, but still to slow to be much use compared to my X2.

 

[kamranziadar]

Well here goes Gigahertz war again.

 

[Technonut]

My rig pulls 26s in 1M @ 4.4GHz..... An OC to 5GHz at low VCore is sweet though....

 

[jazzboy]

Originally posted by: dexvx

Perhaps the most surprising thing is that it is on stock vcore with 50C CPU-Burn using a stock cooler.

Are you sure thats stock vcore? Don't forget that we're talking 65nm here - so I doubt the default vcore would be more than 1.35 at the very most. And even if the actual request voltage is 1.35, 1.456 a lot of overvolting from the motherboard.

 

[dexvx]

Originally posted by: jazzboy

Originally posted by: dexvx

Perhaps the most surprising thing is that it is on stock vcore with 50C CPU-Burn using a stock cooler.

Are you sure thats stock vcore? Don't forget that we're talking 65nm here - so I doubt the default vcore would be more than 1.35 at the very most. And even if the actual request voltage is 1.35, 1.456 a lot of overvolting from the motherboard.

I dont know, they use the same voltage on default speed and on OC speed and they are given the "default" range. I'm guessing the motherboard just defaults it to 1.45V, which is kind of high, I suppose.

I'll correct the SuperPi errors relating to proc speed.

 

[Lonyo]

Originally posted by: dexvx

Originally posted by: jazzboy

Originally posted by: dexvx

Perhaps the most surprising thing is that it is on stock vcore with 50C CPU-Burn using a stock cooler.

Are you sure thats stock vcore? Don't forget that we're talking 65nm here - so I doubt the default vcore would be more than 1.35 at the very most. And even if the actual request voltage is 1.35, 1.456 a lot of overvolting from the motherboard.

I dont know, they use the same voltage on default speed and on OC speed and they are given the "default" range. I'm guessing the motherboard just defaults it to 1.45V, which is kind of high, I suppose.

I'll correct the SuperPi errors relating to proc speed.

It's not on stock VCore at all, it even says to in the article.
Stock VCore is 1.3v, they are using 1.456v (on the CPU-Z screenie).

le Vcore de 1.3 à 1.45 volts

This overclocking was carried out while passing Vcore from 1.3 to 1.45 volts

(Nasty Google translation)

 

[stevty2889]

Even an increase of vcore to 1.45v it's still pretty good. I needed 1.525v to keep my 3.4 prescott stable at 3.82.

 

[phaxmohdem]

At the end of the day... It's still a Celeron
At the end of the day... It's still Netburst
At the end of the day... I still don't want it.

 

[Thor86]

Originally posted by: phaxmohdem
At the end of the day... It's still a Celeron
At the end of the day... It's still Netburst
At the end of the day... I still don't want it.

Agreed.

 

[dexvx]

Originally posted by: Lonyo
It's not on stock VCore at all, it even says to in the article.
Stock VCore is 1.3v, they are using 1.456v (on the CPU-Z screenie).

le Vcore de 1.3 à 1.45 volts

I interpretted as "The default vcore of 1.3 to 1.45V" as a range (instead of at), since this is pre-production chip.

 

[Lonyo]

Originally posted by: dexvx

Originally posted by: Lonyo
It's not on stock VCore at all, it even says to in the article.
Stock VCore is 1.3v, they are using 1.456v (on the CPU-Z screenie).

le Vcore de 1.3 à 1.45 volts

I interpretted as "The default vcore of 1.3 to 1.45V" as a range (instead of at), since this is pre-production chip.

Current 90nm are at 1.4v so I'd be suprised if 65nm went to 1.45v.

Today's Prescotts vary in their operation voltage, but the chip to which we are comparing our Cedar Mill sample has a 1.400V core voltage. The 65nm Cedar Mill sample has a 1.300V core voltage.

In this case, we compared Presler to a Pentium D that featured a 1.3625V core voltage. Our Presler sample, however, featured a 1.300V core voltage - identical to that of a single core Cedar Mill.

Linky
Also from a French speaking person (for better translation):
that overclocking has been realised by changing the vcore from 1.3 to 1.45 volts by using a ventirad (didn't know what that was) intel

 

[mamisano]

Still a lot of voltage for a 65nm chip, my Opteron 165 sits at 1.35v stock on a 90nm process.

 

[coldpower27]

Originally posted by: mamisano
Still a lot of voltage for a 65nm chip, my Opteron 165 sits at 1.35v stock on a 90nm process.

As has been said in the article, default voltage of Cedar Mill is 1.3V.

 

[dmens]

So the rumors were true. I heard cedarmills have been hitting 5ghz on air using a mobo codename "Badaxe"... apparently it targets the enthusiast market.

 

[BrownTown]

Those are some pretty impressive numbers for a celeron, whats this processor supposed to cost, it might actually not suck like all the other celerons...

 

[Zap]

Hey hey, not all Celerons suck. Here's the Celeron Hall-of-Fame rollcall:

300A@450
366@550
533A@800
566@850
1000A@1.33
1.1A@1.46
1.2@1.6
low end Celeron D 533MHz FSB@800MHz FSB
various mobile Celeron (Northwood 256k cache) @up to 3+GHz

 

[Griswold]

My original VW Beetle will also go 300kph if I drop it from a plane flying at 30'000ft!

 

[blckgrffn]

Originally posted by: Zap
Hey hey, not all Celerons suck. Here's the Celeron Hall-of-Fame rollcall:

300A@450
366@550
533A@800
566@850
1000A@1.33
1.1A@1.46
1.2@1.6
low end Celeron D 533MHz FSB@800MHz FSB
various mobile Celeron (Northwood 256k cache) @up to 3+GHz

My 1.8 mobile at 3.6ghz, was, um, pretty fast And it cost $35.

Can hardly tell the difference with my 3.4 p4.

 

[mamisano]

Originally posted by: Zap
Hey hey, not all Celerons suck. Here's the Celeron Hall-of-Fame rollcall:

...
366@550
...

Just set up a Linux Server running Dual Celeron 366@550 on a BP6. Had to re-cap the board, but $15 and an hour later it runs like new

 

[RichUK]

Originally posted by: Zap
Hey hey, not all Celerons suck. Here's the Celeron Hall-of-Fame rollcall:

300A@450
366@550
533A@800
566@850
1000A@1.33
1.1A@1.46
1.2@1.6
low end Celeron D 533MHz FSB@800MHz FSB
various mobile Celeron (Northwood 256k cache) @up to 3+GHz

A friend of mine still has one of those

 

[Phoenix 97]

I seem to have had really poor luck with overclocking anything that wasn't mine in the past few years.
In 2000 I went for the Celeron 566 @ 850 route but the motherboard I was using was so unstable it hard-locked on me like 3 times a day, I also ended up with corrupted data here and there. I was using a board with the abysmal VIA Apollo Pro 133A chipset.
Later on I got myself a Tualatin Celeron 1200 with a spiffy 256KB L2. The system ran just fine on a new i815EP board but the 1/4 PCI divider wouldn't kick in until about 127 MHz FSB. It would have been nice to do something like 1440/120 with that processor but the board wouldn't allow it. I doubt my IDE devices would have liked that high a PCI speed.
Now I'm using a Northwood P4 1.6 on a P4B with 1GB of SDRAM (don't laugh ). Sure the Northwood would be a great overclocker but the board doesn't seem to be stable at all beyond 107 MHz FSB. At 100 though it's pretty solid.
With my luck I'd be happy to get one of these new Celerons up to 4.0 GHz.

It seems that Intel always sticks something into the Celerons to make them perform well below equally clocked P3/P4s. I see now that they have 512KB L2, would this be enough for "today's" applications? Will the FSB be the performance hinderer here or does it look like they will tweak the L2 cache latency and/or set associativity?

(As for overclocking things that aren't mine, there were those few 486 SX 25s back in high school (late 90's) which I pushed up to 50 MHz with no extra cooling whatsoever. )

 

[mwmorph]

Oc results are great. i mean 5ghz air stock cooler is phonemonal, but it also shows why netburst should die. ther is absolutely no reason you should have to co to 5 ghz to be competitive with a fx57 at 2.6. that's insaine. are yuo telling me the amd64 has 2x more ipc or 1/2 the pipeline length as a netburst, becaue if it;s true, the netburst formula is pretty weak.

 

[Falloutboy]

I thought intel was dumping netburst? and cedar mill was a Pentium M based proc. if intel is truly staying with netburst I think AMD is going to have a party