Conroe(Core 2) is the new Ultimate Silent and Cool Platform

[merlin]

It's scary how good Conroe is now that we have full test data. The one test that amazed me the most was when they enabled speedstep and got a 25C temperature at 1.6ghz and .9 volts at Tomshardware. That's completely insane how little heat it's producing at that point. I'm suspecting we'll have a lot of passively cooled undervolted cpu systems running around. I think I'll be recommending a 775 shuttle for my friend now.

[jaganath]

It's scary how good Conroe is now that we have full test data. The one test that amazed me the most was when they enabled speedstep and got a 25C temperature at 1.6ghz and .9 volts at Tomshardware. That's completely insane how little heat it's producing at that point

I think you're exaggerating the amazingness of that thermal result. My undervolted Sempron 3000 rarely breaks 30C with minimal airflow, running 1.6GHz @ 1.1V (can't go any lower with E6 chips). Also, it would be useful to know the ambient air temp in that test, as anyone can get a 25C CPU temp when the ambient is 10C. (ie what is the temperature delta over ambient)

[scorp]

The Conroe doen's really seem to be that amazing if we look only from the power consumption angle. It's TDP is far from the A64 X2 3800+ EE (35W), which still provides more than enough power for most users. If we are talking about performance / power consumption, than the Conroe is truly amazing.

[Aris]

it may be a godsend for all you die hard intel guys out there, but as for the rest of us, its just a step in the right direction. IMO, i'd still take a Pentium-M or a Core Duo over the Conroe (Core 2 Duo)

[jmke]

I think you're exaggerating the amazingness of that thermal result. My undervolted Sempron 3000 rarely breaks 30C with minimal airflow, running 1.6GHz @ 1.1V

you fail to notice that clock for clock that Core 2 will be a lot faster than that Sempron

IMO, i'd still take a Pentium-M or a Core Duo over the Conroe (Core 2 Duo)

Core Duo is quite a bit slower.

[QuietOC]

(deleted)

Maybe AMD will fix that 1.1V problem now....

But you can't directly compare power usage with core voltage between two vastly different chips. The Conroe + northbridge have considerable more transistors to keep charged than a K8.

It is ironic that it took Intel this long to slightly exceed the 90nm K8 with a rather complex (huge cache) 65nm chip. Maybe the Merom will be able to run on 0.7V or so, but I think we are going to hit a point where there just won't be enough voltage.

[Aris]

IMO, i'd still take a Pentium-M or a Core Duo over the Conroe (Core 2 Duo)

Core Duo is quite a bit slower.

cpu performance at this level has become fairly trivial. for people like me who only need a high performance computer for gamming, a good video card is all that is needed.

I'll take the small fps hit and go with a processor with a minimal power draw.

[smilingcrow]

The Conroe doen's really seem to be that amazing if we look only from the power consumption angle. It's TDP is far from the A64 X2 3800+ EE (35W), which still provides more than enough power for most users. If we are talking about performance / power consumption, than the Conroe is truly amazing.

You’re forgetting that the 65W TDP for Conroe is for 2.67 GHz at stock voltage. If you under-volt the entry level Core 2 Duo (1.86), it should easily match the 35W AM2 in power consumption but with better performance. AMD don’t seem to have plans to lower pricing on the EE parts and the 35W X2 3800 has a large price premium, so Intel should have the advantage there. It’s hard to say how available either CPU will be in the coming months.

The difference between these two is more down to the pricing of the CPUs and the motherboards rather than power consumption. I suggest that the power consumption of their respective motherboards is more of an issue than the CPU power consumption, with ATI chipsets usually being more frugal than Nvidia.

[smilingcrow]

IMO, i'd still take a Pentium-M or a Core Duo over the Conroe (Core 2 Duo)

Judging by preview benchmarks Merom offers a slightly better performance per watt than Yonah, which in turn has a significantly better performance per watt than Conroe. I wouldn’t write of Core 2 Duo yet.

[smilingcrow]

It's scary how good Conroe is now that we have full test data. The one test that amazed me the most was when they enabled speedstep and got a 25C temperature at 1.6ghz and .9 volts at Tomshardware. That's completely insane how little heat it's producing at that point. I'm suspecting we'll have a lot of passively cooled undervolted cpu systems running around. I think I'll be recommending a 775 shuttle for my friend now.

I wouldn’t get too excited by that temperature figure as who knows how accurate the temp sensor on that board is. Also, you have to consider what speed the CPU and case fans were running at to obtain that temperature. Considering that the minimum Speedstep frequency is 1.6 Ghz at stock FSB, this temperature reading seems suspiciously low.

[QuietOC]

I wouldn’t get too excited by that temperature figure as who knows how accurate the temp sensor on that board is. Also, you have to consider what speed the CPU and case fans were running at to obtain that temperature. Considering that the minimum Speedstep frequency is 1.6 Ghz at stock FSB, this temperature reading seems suspiciously low.

My 0.88V 1.4GHz Sempron basically reports ambient temperature at idle (the Ninja no doubt accounts for that). That Conroe temperature sounds reasonable.

[smilingcrow]

I wouldn’t get too excited by that temperature figure as who knows how accurate the temp sensor on that board is. Also, you have to consider what speed the CPU and case fans were running at to obtain that temperature. Considering that the minimum Speedstep frequency is 1.6 Ghz at stock FSB, this temperature reading seems suspiciously low.

My 0.88V 1.4GHz Sempron basically reports ambient temperature at idle (the Ninja no doubt accounts for that). That Conroe temperature sounds reasonable.

My mistake, the temperature isn’t the issue here, but the fan speed (& noise) needed to achieve that temp. Without that data, the temperature value isn’t of much use to silent people. I checked the review and the CPU fan was @ 1,584 & the chassis fan @ 18,750. Obviously the later is an error on their part; the other figures showed the chassis fan @ ~ 3,200. That puts a whole different context on the temp figure.

[QuietOC]

My mistake, the temperature isn’t the issue here, but the fan speed (& noise) needed to achieve that temp. Without that data, the temperature value isn’t of much use to silent people.

I'd like to see current/power draw.

FWIW: At 0.88V the Sempron is at ambient with no fans. Heat leaking from the video card and even IGP heats up the Ninja more than the cpu.

[merlin]

It's scary how good Conroe is now that we have full test data. The one test that amazed me the most was when they enabled speedstep and got a 25C temperature at 1.6ghz and .9 volts at Tomshardware. That's completely insane how little heat it's producing at that point. I'm suspecting we'll have a lot of passively cooled undervolted cpu systems running around. I think I'll be recommending a 775 shuttle for my friend now.

I wouldn’t get too excited by that temperature figure as who knows how accurate the temp sensor on that board is. Also, you have to consider what speed the CPU and case fans were running at to obtain that temperature. Considering that the minimum Speedstep frequency is 1.6 Ghz at stock FSB, this temperature reading seems suspiciously low.

I certainly don't consider it a perfect measurement of temperature, but the real issue at hand isn't regarding just power usage. It's the combination of low power and very good performance. A conroe will have the best power/performance efficiency of any processor out there, especially if undervolted. I'd figure if we can pull off like 2.5ghz at 1.1-1.2v, that would make for an awesome high performance, low power machine. If there were any "performance per watt" comparisons, I think conroe would win hands down. The best performance at the lowest power usage with a reasonable price would be the target of most people here I'd assume. Also even the energy efficient versions of the A64 lose when you compare a combination of both performance and power usage. They do win in terms of overall power usage, but I think that advantage could be reduced with underclocking/undervolting.

It's a scary paradigm shift that's happening here, I've almost always recommended AMD processors until now, but I can't recommend any of them now(in good conscience) unless someone is on an extremely small budget. I hope K8L is everything promised...and not delayed, because Core 2 looks like a one stop shop for all desktops and laptops for now. The only holdback is availability and pricing... I bet demand will cause trouble at least in the short term.

[merlin]

My mistake, the temperature isn’t the issue here, but the fan speed (& noise) needed to achieve that temp. Without that data, the temperature value isn’t of much use to silent people.

I'd like to see current/power draw.

FWIW: At 0.88V the Sempron is at ambient with no fans. Heat leaking from the video card and even IGP heats up the Ninja more than the cpu.

I'd love to see some more undervolting tests where we can see how far Conroe/Merom can undervolt and how much power it uses. I hope you got something in the pipeline like updating the Desktop CPU Power Survey with the new processors MikeC.

[smilingcrow]

My mistake, the temperature isn’t the issue here, but the fan speed (& noise) needed to achieve that temp. Without that data, the temperature value isn’t of much use to silent people.

I'd like to see current/power draw.

Dailytech has a list of Core 2 Duo reviews. I quickly browsed a few for power consumption figures and my initial reaction is that at idle AMD still has the advantage, due possibly to the lower idle clock speed of CnQ versus Speedstep. It’s also possible that the differences are as much to with the different motherboards as CPUs. Without wading through all the reviews and collating all the data it’s hard to say.

I’m tempted to say that for people building a low power system, the choice of motherboard is becoming as important as the choice of CPU. Unfortunately, nobody seems to have done a comprehensive review of this area yet.

[MikeC]

I'd love to see some more undervolting tests where we can see how far Conroe/Merom can undervolt and how much power it uses. I hope you got something in the pipeline like updating the Desktop CPU Power Survey with the new processors MikeC.

Of course.

On the test bench right now are low power parts from AMD --

AMD X2 3800+ @ 35W-max (AM2)
AMD X2 4600+ @ 65W-max (AM2)

Conroe samples are supposedly on their way to me; I guess SPCR must be "second tier" in Intel's review list, as some other sites have obviously had them for a week or two.

I'm trying to get low voltage Core Duo U2500, 1300 and 1400 samples, but no luck so far. I really don't understand why Intel wouldn't want them in the hands of interested and qualified media like SPCR.

[jaganath]

I'm trying to get low voltage Core Duo U2500, 1300 and 1400 samples, but no luck so far. I really don't understand why Intel wouldn't want them in the hands of interested and qualified media like SPCR.

Maybe because we bashed their Netburst procs, especially Prescott? Even big corporations can be petty like that. I'd love to see a review of LV and ULV CD parts, we're looking at single-figure power consumption there.

jmke:you fail to notice that clock for clock that Core 2 will be a lot faster than that Sempron

You fail to notice that the Sempron is perfectly adequate for most people's everyday tasks (hence the name). I'm not bashing Core 2, but it is a bit galling that Intel finally shows up at the performance per watt game after so many years of ignoring it completely and we're all supposed to forget that Prescott ever happened.

[Brians256]

I really don't understand why Intel wouldn't want them in the hands of interested and qualified media like SPCR.

They get better PR from having the reviews spread out. It's not just the initial buzz; they want to keep the community talking.

[smilingcrow]

On the test bench right now are low power parts from AMD --

AMD X2 3800+ @ 35W-max (AM2)
AMD X2 4600+ @ 65W-max (AM2)

Conroe samples are supposedly on their way to me; I guess SPCR must be "second tier" in Intel's review list, as some other sites have obviously had them for a week or two.

I have no idea how big SPCR is but Core 2 Duo is relatively a bit of phenomenon, so I imagine everybody is after them for review purposes. If SPCR has made it to “second tierâ€

[merlin]

I'm trying to get low voltage Core Duo U2500, 1300 and 1400 samples, but no luck so far. I really don't understand why Intel wouldn't want them in the hands of interested and qualified media like SPCR.

Maybe because we bashed their Netburst procs, especially Prescott? Even big corporations can be petty like that. I'd love to see a review of LV and ULV CD parts, we're looking at single-figure power consumption there.

jmke:you fail to notice that clock for clock that Core 2 will be a lot faster than that Sempron

You fail to notice that the Sempron is perfectly adequate for most people's everyday tasks (hence the name). I'm not bashing Core 2, but it is a bit galling that Intel finally shows up at the performance per watt game after so many years of ignoring it completely and we're all supposed to forget that Prescott ever happened.

I don't think anyone is going to ever forget that. There's a reason we've been preaching Amd for years now as the price, performance, and power leader. But to also be fair, Intel's been in the performance per watt game for notebooks almost the entire time as the leader. They've only been completely failing in the desktop arena for about 5 years. What's being stated now is that Core 2 is the best option in most cases.

[Firetech]

Conroe samples are supposedly on their way to me; I guess SPCR must be "second tier" in Intel's review list, as some other sites have obviously had them for a week or two.

They've got the ES press kits from what I've seen in the review pics, so with any luck you'll get 'production' samples.

I'm trying to get low voltage Core Duo U2500, 1300 and 1400 samples,

Maybe that's another reason why you haven't got samples (not 'mainstream' enough?).

No hurry, they aren't available on shelves yet (or are they?)

[MikeC]

The low voltage core duos are already available.

As to why SPCR's not getting Conroe samples first, it's obviously because they want the sites that have the huge geek reach -- mainstream performance oriented sites.

Why they don't send us samples of the low voltage core duos -- bad marketing, imo.

[randytsuch]

How come no one is talking about performance/price? That, in combination with the low power consumption of the conroe, is where Intel is going to kill AMD. I know AMD is going to announce a price drop soon, but it would have to be a really big drop to get them even close to where Intel is, at the announced Conroe prices.

BTW, I use my machine for video rendering, and have a couple year old P2.4 OC'ed to 3G. I really want a faster machine, but may wait a little longer to upgrade, and see if prices come down even more. Probably won't be able to buy any Conroes for a little while anyway.

Randy

[Firetech]

Price/Performance looks pretty impressive. Even if AMD drastically reduce prices, it's going to be a tough choice.
I had thought I'd just wait and get a nice X2 S939 for my current rig at reduced price but my local vendors have a shortage even before the discounts hit and the prices have actually risen!
It looks like even the lower end Core 2 Duo's are outperforming the best mainstream CPU's AMD has to offer. It's going to be an interesting end to the year

[|Romeo|]

The low voltage core duos are already available.

As to why SPCR's not getting Conroe samples first, it's obviously because they want the sites that have the huge geek reach -- mainstream performance oriented sites.

Why they don't send us samples of the low voltage core duos -- bad marketing, imo.

I fail to see how it is bad marketting -the LV & ULV parts are available in Micro-FCBGA i.e. they have to be attached to a board at manufacture. If you know anyone who can do the design and have adaptor boards made economically please let me know as I'd like to make them a job offer.

SPCR is essentially a consumber site, and the LV & ULV parts are not consumer products. The LV & ULV parts are not sold in retail channels, so why would Intel want to try & market them?

The only samples of LV or ULV that you are likely to get would be from motherboard makers (Aopen springs to mind) who have designed a board with one of the LV parts attached. Even then I cannot see it being a commercial success, as getting the (small) quantity of each speed to produce right would be bloody tricky for what would be a quite complex board.


I know this sounds kind of like a tirade; but seriously -how could you test a CPU that you couldn't install in a motherboard?

[smilingcrow]

The low voltage core duos are already available.

Why they don't send us samples of the low voltage core duos -- bad marketing, imo.

Assuming they are available in the same socket as the standard Core Duos, motherboard support might be an issue. It was an issue for LV & ULV Pentium Ms with some boards.
I’m not quite sure what you are going to gain by reviewing these! I can tell you in advance, they run cool

They seem expensive for chips that appear to just have had their FSB reduced to 533 and their multipliers tweaked downwards. You can make most if not all of these tweaks yourself with software and/or BIOS settings with a standard Core Duo.
I’d like to see you try that in your next motherboard & CPU showdown.

And while you are at it, maybe try the same with Conroe, reducing its FSB to 667 or even 533. Then there’s the difference in power consumption between the 2 & 4 MB cache versions. You can test that by reducing the max multiplier of the 4MB chip to the same as one of the 2MB chips.
Lastly, I’d like to see how over-clocking the FSB affects system power consumption. i.e. compare a stock E6600 @ 2.4 1066 FSB with an E6300 1.86 @ 2.4 ~1333 FSB.
I don’t ask for much, I know.

[RDaneel]

Not only is the performance per watt looking nice, the OC possibilities on the ~$300 part look amazing!

Why is it on SPCR I'm one of the only guys wondering about OCing and overvolting, and on HardOCP I'm one of the only guys worried about cooling and noise?

I think the bottom line today is that Intel is back in the game (both in terms of performance and moving thermals in the right direction), and AMD's price cuts will move us all towards cheaper performance that doesn't beg for a 550W PSU! We all win...

[merlin]

Not only is the performance per watt looking nice, the OC possibilities on the ~$300 part look amazing!

Why is it on SPCR I'm one of the only guys wondering about OCing and overvolting, and on HardOCP I'm one of the only guys worried about cooling and noise?

I think the bottom line today is that Intel is back in the game (both in terms of performance and moving thermals in the right direction), and AMD's price cuts will move us all towards cheaper performance that doesn't beg for a 550W PSU! We all win...

I'm more for overclocking and undervolting, the best of both worlds!

[smilingcrow]

Not only is the performance per watt looking nice, the OC possibilities on the ~$300 part look amazing!
Why is it on SPCR I'm one of the only guys wondering about OCing and overvolting, and on HardOCP I'm one of the only guys worried about cooling and noise?

Some of the Conroe’s are over-clocking very well at stock voltages, which suggests they may well be good under-volters as they have a lot of headroom at stock.
It will be interesting to see how high you can overclock an E6600 (or lower) and still cool it quietly. The power consumption is going to start rising quickly north of ~3.5 GHz I imagine.
A better option for serious power users would be Woodcrest. It’s got to be easier to cool two 65W chips than one 120W chip (Conroe O/C). You also get more performance to boot, as two 3 GHz Woodcrests will easily beat even a 4 GHz Conroe in heavy duty applications that are optimized for more than 2 cores. I’m assuming that Woodcrest can undervolt, which is why I rate it at 65W rather than 80W.
Now if only someone would release a chipset/motherboard for Woodcrest that supports RAM other than FB-DIMMs and can overclock. nVidia!