Ideal Hi-end CPU to cool down - Winchester Athlons

Tobias · Post by **Tobias** » Tue Nov 23, 2004 3:59 pm

What was the relationship between k7 desktop and mobile CPU:s regarding TDP? Was the desktop chip about twice as hot or was it less? AMD today released a new mobile sempron 3000+, made with 90nm SOI and 128k L2 which to me sounds as a winchester sempron. The TDP of this chip is supposed to be 25W. If one would conclude that the same relationship between desktop winchesters and mobile winchesters apply as between desktop bartons and mobile ones, that would give a ballpark figure of the "actual" TDP of desktop winchesters.

On the other hand, such an assumption may be to optimistic. I am sure AMD has found a way to on a separate line churn out mobile CPU:s that whacks the livving daylights out of all their other processors heatwise without added cost (it's price is 134$ compared to 135 for k7 3000+ and 144$ for k8 2800+).

Post by **MikeC** » Tue Nov 23, 2004 5:21 pm

Here are some prelim results from SPCR's lab. Sean & I experimented with two A64s: a 3800+ (2.4G 130nm core) and a 3500+ (2.2Ghz 90nm core). We kept things as consistent as we could, and used minimal power draw components. The clock speed of the 2 CPUs was controlled by changing the multiplier in the BIOS and keeping CnQ turned off.

The system:
Soltek K8TPro VIA chipset board (??20W??)
Matrox G550 VGA (??W??)
IBM 5400/8mb cache 80G 2.5" drive (1W idle, 5W peak)
512mb DDR400 memory (??W??)
Silverstone Fanless PSU (~78% efficiency)

DC system power draw
@ 2.2G (200x12 -- max speed of the Winchester core CPU with bus at 200MHz)
130nm core: 115W
90nm core: 81W
difference: 34W!

@1.8G (200x9 -- The 3000+ A64s run at this speed)
130nm core: 100W
90nm core: 73W
difference: 27W!

With CnQ enabled, the system with Winchester at idle @ 1.8G (3000+) drew only 49W AC!

There isn't much more to say except that even at 1.8G, this is one fast cool system, esp w/ dual ch mem. The temp monitoring system in these boards was not working properly but at no point during the testing did the Winchester HS ever get too hot to touch. It barely seemed to get warm. This is with a pretty low airflow very stock type of HS.

Sean is now hatching all manner of plots to dump his current system onto his poor unsuspecting brother so that he can get his hands on a cool Winchester system.

Edward Ng · Post by **Edward Ng** » Tue Nov 23, 2004 7:20 pm

MikeC wrote:Silverstone Fanless PSU (~78% efficiency)

What's ths story on the piece that puked its guts?

GOODFELLA · Post by **GOODFELLA** » Tue Nov 23, 2004 7:47 pm

I hate to post it here, but for anyone interested in a cheap Winchester and your in the Toronto area. Athlon 3200+ s939, $241 and 3500+, $335.
Those are Canadian prices.
www.canadacomputers.com
I went there and verified, indeed 9nm.
Ive bought alot of stuff there and they are always well stocked.
Check it out.

Oh yeah, they do not do online orders, only walk in customers.

kie · Post by **kie** » Wed Nov 24, 2004 12:00 am

MikeC wrote:
Sean is now hatching all manner of plots to dump his current system onto his poor unsuspecting brother so that he can get his hands on a cool Winchester system.

I don't blame him - I just built a new "work" orientated PC just to use this chip

Jan Kivar · Post by **Jan Kivar** » Wed Nov 24, 2004 4:50 am

MikeC wrote:Here are some prelim results from SPCR's lab. Sean & I experimented with two A64s: a 3800+ (2.4G 130nm core) and a 3500+ (2.2Ghz 90nm core). We kept things as consistent as we could, and used minimal power draw components. The clock speed of the 2 CPUs was controlled by changing the multiplier in the BIOS and keeping CnQ turned off.
...

What about the voltages used with the CPUs?

Cheers,

Jan

Post by **MikeC** » Wed Nov 24, 2004 8:38 am

Jan Kivar wrote:What about the voltages used with the CPUs?

It was kept at default throughout. 1.4? The exact # escapes me at this point, will check later.

Jan Kivar · Post by **Jan Kivar** » Wed Nov 24, 2004 11:46 am

MikeC wrote:
Jan Kivar wrote:What about the voltages used with the CPUs?
It was kept at default throughout. 1.4? The exact # escapes me at this point, will check later.

The 130mm parts have 1.5V as default, whereas the 90mm parts run with 1.4V. Are You saying 1.4V was used for both or what?

Cheers,

Jan

Post by **MikeC** » Wed Nov 24, 2004 12:49 pm

Jan Kivar -- what's with all the

???

Did you not see that this is a prelim report? (And here is a

of my own.) It is not a review, it is not anything but comments after an afternoon of initial messing about. If you know that they are 1.4V and 1.5V, then obviously you can subtract a couple watts from the 130mm CPU figures. It is not going to change the huge power dissipation difference.

EDIT: Bios reports 1.49V for both.

Jan Kivar · Post by **Jan Kivar** » Thu Nov 25, 2004 7:56 am

Yeah.... Sorry.

There is still quite an improvement even if the effect of alleged voltage difference is reduced from the 130mm parts (the drop should be about 12% if I managed to use Kostik's CPU Power correctly).

That said, I can't wait for the real article. And I do hope that AMD wouldn't forget us S754 users yet.

Cheers,

Jan

Tobias · Post by **Tobias** » Fri Nov 26, 2004 3:16 pm

Using kostiks cpu power to extrapolate from those numbers shows that a Max TDP that would be consistent with a 8W drop at 1,4V for a 400MHz drop is about 45W. That would be about 13W with Cn'Q enabled or about 28W at full speed@1,1V. This would definately put passive cooling within reach. A onefan system should no longer be any challange... What do you think, would be possible with a top blowhole to use a Zalman 7700 fanless and an Antec Phantom? Perhaps an empty pci-bracket for some freeflowing cool air?

Am I nuts? Is it even possible to extrapolate like this? how much airmovement is realy needed to run such a setup? I am either way SO tempted to try, to bad only that I am a poor student:(

Post by **MikeC** » Fri Nov 26, 2004 3:44 pm

One very important caveat for all those who seek fanless operation: Both Intel and AMD tech docs refer specifically to the need for motherboard PS components near the CPU to be cooled by the fan on the CPU HS. Take away this fan and your risk of motherboard failure goes up significantly. That's a fact.

All the more reason to follow the path we most often suggest: Use very quiet fans and have at least a little airflow with massively powerful HS. A quiet enough fan makes no more noise than a super quiet HDD, which becomes your main limiting factor to silence if you get the fan noise low enough.

So fanless is an interesting target, but not really a holy one. In fact there may be more drawbacks than benefits.

burcakb · Post by **burcakb** » Fri Nov 26, 2004 11:08 pm

With all the low wattage values going around, a few days ago I was tempted to run the Zalman7000 on my Newcastle passively. I'd already undervolted it to 1.2V and with an EXTREMELY slow-turning fan (I have no idea how slow as the bios monitor says something on average of 200 rpm) and still had low temps - 44C full folding.

Within 1 minute of disconnecting the fan, temp was up to 68C. I didn't go further and reconnected the fan.

Frankly, there was absolutely no difference in noise, so why take the risk?

paperboy5 · Post by **paperboy5** » Sat Nov 27, 2004 8:21 pm

how do I differentiate between a newcastle and winchester cpu?
as in is there any markings on the retail boxes?

thanks.

sneaker · Post by **sneaker** » Sat Nov 27, 2004 10:00 pm

I was referring to the tomshardware article today and noticed that the table appears to have been updated since it was first published, adding power consumption figures with CnQ enabled. I'm pretty sure they weren't there before, because a few days after the article appeared I was almost going to send them an e-mail complaining about the lack of with-CnQ figures.

MikeC wrote:Here are some prelim results from SPCR's lab.

There is one question raised by the x86-secret article linked to earlier in this thread that I'd like answered if you have the hardware available. The conventional wisdom seems to be that motherboards are insignificant from a power consumption perspective, but the article suggests the socket 939 platform is significantly more power hungry than 754--enough to offset some of the Winchester's advantage--and it speculates the dual-channel RAM may account for the difference. First, their results are yet to be backed up--I haven't seen any other articles showing a significant difference betwen 754 and 939. Second, if there is a difference, is this due to the dual-channel operation? Perhaps testing with dual-channel disabled will answer this.

Another question is raised by this article Neil linked to a month ago. It suggests the single-chip nForce "consumes a whole lot more" power than chipsets from VIA and Intel. I suppose a large number of motherboards would need to be tested to come to a reliable conclusion. Nevertheless, it's something to keep in mind with any 754 vs 939 test; a comparison where one is VIA-based and the other nForce-based may be unfair.

lenny · Post by **lenny** » Sun Nov 28, 2004 1:12 am

paperboy5 wrote:how do I differentiate between a newcastle and winchester cpu?
as in is there any markings on the retail boxes?

This may get more complicated in the future, but for now:

- All 3000+ and 3200+ 939 are Winchester core
- Some 3500+ 939 are Winchester core. Others are Newcastle.
- For 3500+ OEM (and maybe retail?) part, chip # is ADA3500DIK4B
- For retail box, part # *may* be ADA3500BIBOX
- All 754 and other speeds not mentioned are 130nm

Probably more information than most people want to know available here.

halcyon · Post by **halcyon** » Sun Nov 28, 2004 7:33 am

Very informative review in many ways.

Thanks for the link!

nbac · Post by **nbac** » Sun Nov 28, 2004 10:27 am

MikeC wrote:One very important caveat for all those who seek fanless operation: Both Intel and AMD tech docs refer specifically to the need for motherboard PS components near the CPU to be cooled by the fan on the CPU HS. Take away this fan and your risk of motherboard failure goes up significantly. That's a fact.

True, but with such low power consumption of the Winchester chip, the regulator will probably stay rather cool?

Tobias · Post by **Tobias** » Sun Nov 28, 2004 10:48 am

And on some boards with the socket aligned in a north-south direction those components will not get much cooling from many coolers anyway.

on the other hand perhaps A64 boards are more sensitive due to cnQ...

Post by **MikeC** » Sun Nov 28, 2004 11:20 am

True, but with such low power consumption of the Winchester chip, the regulator will probably stay rather cool?

And on some boards with the socket aligned in a north-south direction those components will not get much cooling from many coolers anyway.

on the other hand perhaps A64 boards are more sensitive due to cnQ...

Obviously, going fanless does not mean imminent failure of those components.

I am simply reporting information I believe is relevant and prudence is suggested. Personally, I much rather have the assurance of a little forced airflow over the CPU HS and MB components than whatever satisfaction I might derive from knowing that there is no fan on the CPU HS. If you can't hear the difference (and general you can't if the fan is spinning slow enough) then what's the point of fanless except some kind of ego trip?

YMMV.

Tobias · Post by **Tobias** » Sun Nov 28, 2004 2:03 pm

The lack of silent enough fans in sweden:)

okey, I know, if you want them bad enough, you may get them, but it is... messy...

paperboy5 · Post by **paperboy5** » Sun Nov 28, 2004 5:14 pm

This may get more complicated in the future, but for now:

- All 3000+ and 3200+ 939 are Winchester core
- Some 3500+ 939 are Winchester core. Others are Newcastle.
- For 3500+ OEM (and maybe retail?) part, chip # is ADA3500DIK4B
- For retail box, part # *may* be ADA3500BIBOX
- All 754 and other speeds not mentioned are 130nm

Probably more information than most people want to know available here.

thanks.

alexb · Post by **alexb** » Mon Nov 29, 2004 2:39 am

Cool... for some reason I didn't get notification of replies to this thread I started... didn't realize it went so far!

When I started this thread, some laughed at my findings and the fact that these CPUs are great candidates for quiet operations... I guess truth prevails at the end!

Great follow ups... cannot wait for offical review!

Storm · Post by **Storm** » Mon Nov 29, 2004 6:19 am

xbit labs 90mm review. I believe their power usage numbers are more accurate.

Mats · Post by **Mats** » Mon Nov 29, 2004 6:43 am

Storm wrote:xbit labs 90mm review. I believe their power usage numbers are more accurate.

Don't think so. Xbit says:

To get the value of the processor power consumption, which is the same as the processor heat dissipation, according to the energy conservation law, we actually measured the current going through the 12V processor circuit. In fact, this method is not very precise, because it doesn’t take into account the performance index of the processor voltage regulator circuit, however, it suits quite well for a preliminary rough comparison.

Tobias · Post by **Tobias** » Mon Nov 29, 2004 6:56 am

It may not be the most precise testing available, but I think it is precise enough to for relative testing (130nm against 90nm) and for absolute number, I'd say a worst case scenario of 51W overclocked (assuming 100% regulator effficiency) is very good, compared to some other CPU:s on the market.

It is also interessting to notice that the 130nm-chip stays below it's 89W TDP rating, as it should

edit: reworked for less drama

Mats · Post by **Mats** » Mon Nov 29, 2004 7:31 am

Wow, calm down...
Xbit says: "...not very precise..." and ".... a preliminary rough comparison....", those are their words, not mine.

Storm · Post by **Storm** » Mon Nov 29, 2004 3:34 pm

Yes these processors are lower powered than 130mm, but I think amd is not making a big fuss about their processors low power because its not that low. Its similar to other amd die shrinks in the past.
What xbit labs say is they tested the current on the 12v processor circuit if you look at the pictures from tomshardware they are testing the same circuit, im assuming from the x86-secrets translation they are testing the same way.

I believe all these tests are not precise and their tests are affected by the motherboard used. x86-secrets had a 10watt difference between a S754 and S939 processor, tomshardware shows a 3.2 watt difference between a S754 and S939.

I might be wrong though.

alexb · Post by **alexb** » Mon Nov 29, 2004 4:06 pm

Well, I guess it would be very hard to determine the exact wattages of these CPUs since it depends on many factors.

The fact remains though -> The Winchester Processors are currently THE BEST BET for building hi-end quiet/silent systems BY FAR. With performance close to the top of the line P4 the cooling/power requirement is much lower!

mb2 · Post by **mb2** » Mon Nov 29, 2004 7:20 pm

cpucity.co.uk also have winchesters in the UK if u dont want to buy from ebuyer..
i'm just waiting for that ATI chipset..