P4 .65nm Sneak Peak at Tom's

[phaxmohdem]

Hey just wanted to point out a semi interesting article over at Tom's Hardware. They have managed to get an early engineering sample of a P4 Cedar Mill Chip. Benchmarks show clock for clock no performance benefit over current Prescott chips (But did you really expect any?) However the CPU's finally run cooler! And as such may give Intel the ability to put out 4+ GHz CPU's if they so choose.

Anyway, interesting read: Linky

 

[Hacp]

Wow nice. Any hints of 65NM P-Ds comming soon? The P4s ceder mill should Oc better and cooler .

 

[jazzboy]

Originally posted by: phaxmohdem
And as such may give Intel the ability to put out 4+ GHz CPU's if they so choose.
[/L]

I don't think Intel has any plans to increase clockspeeds, at least for cedar mill chips, although they do have plans for 3.4 ghz and 3.6 ghz presler chips (9xx series cpus).

As far as i know intel is now fully dedicated to moving away from the "more mhz sells" marketing - which means that if they released a 4 ghz pentium 4, the press would once again be going around like wildfire saying, "intel has now produced 4ghz chips!" or something like that.

 

[Hacp]

But thats better for us. We buy the cheapo ceder mill chips, maybe 3 GHZ and OC them like helll. I bet people will be able to get 4.5+ on turbo air cooling.

 

[Maximilian]

Originally posted by: jazzboy

Originally posted by: phaxmohdem
And as such may give Intel the ability to put out 4+ GHz CPU's if they so choose.
[/L]

I don't think Intel has any plans to increase clockspeeds, at least for cedar mill chips, although they do have plans for 3.4 ghz and 3.6 ghz presler chips (9xx series cpus).

As far as i know intel is now fully dedicated to moving away from the "more mhz sells" marketing - which means that if they released a 4 ghz pentium 4, the press would once again be going around like wildfire saying, "intel has now produced 4ghz chips!" or something like that.

Nothing wrong with the press doing that. I think they should push it as far as itll go before the new architecture chips are released. Lets see a 4+ ghz P4! Before clock speed goes out the window completely.

 

[phaxmohdem]

I'm almost certain that .65 Pentium D's will be released about the same time as single core. Intel would be foolish not to do so as that is their current "Hot" product. Roadmaps don't really reveal much except that the current smithfield core will be "replaced by Cedar Mill" so I'm assuming that Q1 2006 we might see a nice refresh of the Intel lineup, assuming .65 gets pulled off a lot easier than .90 did.

 

[RichUK]

This difference is large enough to bring Intel back into the performance per watt game when competing with AMD

LoL .. what is the point in die shrinking the old architecture, this will only reduce heat dissipation and current consumption as stated above, how does that benefit anyone except with regards to HTPC's.

This is just Intel making more money off of the old architecture yet again, its performance we are looking for, with the respect to "performance = IPC x frequency" with IPC being the dominant word, it seems that they still want to milk the Ghz race, which is quite clearly pointless on the old netburst architecture. What are they at, a 33 stage pipe now?

They mention with their new strained Silicon process they have accomplished better resistance to voltage leakage, but this would be needed anyway on a smaller manufacturing process as with die shrinks voltage leakage increases, and as you can se from the benchmarks their is no improvement, so this will give intel the ability to reach 4Ghz plus Woopty Dooo .. AMD will then release the FX59, for some more Intel ownage yet again

 

[Markfw]

I don;t see anything to be excited about. So they are cooler, and use a little less power. AMD still uses even less, and less heat, and are faster, and cheaper (except the 820d POC)

 

[stevty2889]

Cedarmill and Presler are pretty much hold over chips until Merom/Conroe are ready. Might have some nice OC potential, but still nothing to get excited about. At least they can have dual cores above 3.2ghz now, that will have some good potential overclockability as well. But pressler is the same as smithfield, 2 cedarmill stuffed in one package. Just now with 2x2mb of cache, and should run a good bit cooler. I wouldn't mind getting one to play with, it's almost time for me to get a new ES chip, hopefuly I'll get my hands on a cedar mill, or even better a presler.

 

[Pabster]

200W to 166W under full load is a drop, but not as much as I'd expected. Still going to be a hot chip.

 

[dmens]

Originally posted by: RichUK
LoL .. what is the point in die shrinking the old architecture

I guess 90nm K8 was a waste of time?

1) An optical shrink is a short, low-risk project that will make a ton of money. Financial ownage, if you will. Such a design will allow tapeout to coincide with fab ramp for the best usage of the fab.
2) Cedarmill silicon provides excellent feedback to the Merom designers.
3) Taping out a low-risk database on a new process will allow the fabs to ramp with high confidence.

Cedarmill's goal was to faciliate 65nm ramps, not performance for gamers. By the way, as a metric, IPC is about as useless as MIPS.

 

[jiffylube1024]

Yay, Intel is doing something... Nothing really exciting here, except that a company is close to releasing a 65 nm chip, which gets me excited for 65nm Athlon X2 and 65nm whatever Intel releases that is actually competitive (is it called 'Yonah' ?) . Until Intel releases their low-power, high IPC desktop sucessor to P4 based on the Pentium M architecture, this 65nm release wins them brownie points, and brownie points alone.

Not really any closer to competition at the moment, but at least power consumption is down... A bit .
----------------------------


This part really cracked me up:

Those of you who may be disappointed by the lack of performance gains should keep in mind that Intel might easily be able to finally add faster Pentium 4 processors at 4+ GHz if there is market demand.

LOL! If there's market demand!!! Hmm, Intel is getting slaughtered in dual core processors; they don't even really compete with the higest end Athlon 64 X2's. I think the market for faster Intel chips is there alright, it's all up to Intel Inside to release some competitive parts!





Edit: wow, after checking through their Cedar Mill preview, I can conclude that THG has some serious zeal for Intel (not like that should be a surprise to anyone).

They ran enough tests only to prove that the old and new 3.4 Ghz chips are exactly the same. It's a freaking preview, yet they bench the crap out of it, probably soiling themselves over the massive 2 second difference in 3DSMax 7.0 . Yikes!

 

[Pabster]

Originally posted by: jiffylube1024
Edit: wow, after checking through their Cedar Mill preview, I can conclude that THG has some serious zeal for Intel (not like that should be a surprise to anyone)

Bah. Get off the THG-basher bandwagon.

65nm process chips are new technology, it is exciting, and it ought to be covered.

I suspect Tom would do a similar write-up if AMD provided 65nm samples as well.

 

[jiffylube1024]

Originally posted by: Pabster

Originally posted by: jiffylube1024
Edit: wow, after checking through their Cedar Mill preview, I can conclude that THG has some serious zeal for Intel (not like that should be a surprise to anyone)

Bah. Get off the THG-basher bandwagon.

65nm process chips are new technology, it is exciting, and it ought to be covered.

Covered is one thing, running so many tests getting the same result is another. Besides, THG's bias is just so funny... It makes me laugh every time.

I suspect Tom would do a similar write-up if AMD provided 65nm samples as well.

I doubt it, but if they did, I'd call them stupid again for running so many tests if they get the exact same results. A few benchies followed by "every test we ran was within a 1% margin of error, showing that there are no architectural improvements in this die shrink" would have sufficed, I think.

I think a more pro-active preview with overclocking results (they did none, which really irked me because I'd love to see how these suckers o/c), throttle watching after agressive 100% CPU load for 2+ days straight, etc. would have been more pertinent. Plus how about some CPU temperatures (idle/load) using a stock cooler and any reference cooler they like (XP-90, TT Typhoon, whatever)? I would have found that information very interesting. Instead it's the same old routine: benchmark ad nauseum. Gets kind of boring when every graph has two identical bars .

I won't bitch any more about it - we still need competition in the CPU market. Still, Intel isn't doing anything to knock our socks off (even with this), and considering AMD has (at absolute maximum) 20% of the market, I'm enjoying cheering for the little guy at ths point in time .

 

[phaxmohdem]

Originally posted by: jiffylube1024


I think a more pro-active preview with overclocking results (they did none, which really irked me because I'd love to see how these suckers o/c), throttle watching after agressive 100% CPU load for 2+ days straight, etc. would have been more pertinent. Plus how about some CPU temperatures (idle/load) using a stock cooler and any reference cooler they like (XP-90, TT Typhoon, whatever)? I would have found that information very interesting. Instead it's the same old routine: benchmark ad nauseum. Gets kind of boring when every graph has two identical bars .

I wondered about that at first too, but if you read carefully you'll catch the part where they "borrowed" it from an annonymous third party. Therefore they probably didn't want to risk toasting a chip that wasn't theres to toast to begin with. Wouldn't reflect well on Tom's or the poor annonymous guy who would have to then explain to Intel how he fried his engineering sample.

 

[MBrown]

No performance gain. But there could be some serious ocing potential. Am I wrong?

 

[jiffylube1024]

Originally posted by: phaxmohdem

Originally posted by: jiffylube1024


I think a more pro-active preview with overclocking results (they did none, which really irked me because I'd love to see how these suckers o/c), throttle watching after agressive 100% CPU load for 2+ days straight, etc. would have been more pertinent. Plus how about some CPU temperatures (idle/load) using a stock cooler and any reference cooler they like (XP-90, TT Typhoon, whatever)? I would have found that information very interesting. Instead it's the same old routine: benchmark ad nauseum. Gets kind of boring when every graph has two identical bars .

I wondered about that at first too, but if you read carefully you'll catch the part where they "borrowed" it from an annonymous third party. Therefore they probably didn't want to risk toasting a chip that wasn't theres to toast to begin with. Wouldn't reflect well on Tom's or the poor annonymous guy who would have to then explain to Intel how he fried his engineering sample.

Bah, you shouldn't toast a chip as long as you're overclocking it within reason. I don't mean using a Promie Mach II, just on air. Even a total doorknob like Tom & co. shouldn't be able to cook a CPU on air. It's pretty easy to keep it alive, just don't put a ridiculous amount of voltage in there (+15% tops), and keep an eye on temperatures. Heck the P4's throttle when they get too hot anyways, so there's very little risk.

 

[Socrilles]

The decrease in power really doesn't seem that significant or am i missing something......... ie they still gonna run hot

 

[Furen]

There could be some increase in overclocking but since these will pack about 70% of the heat of a Prescott onto roughly half the die area, I wouldnt expect them to be much better overclockers under normal conditions. If you sub-zero or water cool them then you might get better overclocking, since it'll be easier to remove the heat from the hotspots faster.