"Venice" Athlon 64 power & heat review

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

"Venice" Athlon 64 power & heat review

Post by NeilBlanchard » Mon May 02, 2005 8:59 am

Hello:

Over at Lost Circuits, there is an interesting review the the Venice Athlon 64 focusing on the excellent power and heat characteristics of this new core.

Green Shoes
*Lifetime Patron*
Posts: 477
Joined: Thu Feb 17, 2005 6:41 am
Location: Nashville, TN

Post by Green Shoes » Mon May 02, 2005 10:18 am

Great find! The only reviews I had seen so far were trying to overclock them as much as possible on Mach IIs, with little regard to heat dissipation.

8.4 W at idle! :shock: Hopefully the extra 512 MB cache on the San Diego won't hurt it's power draw too much; the reviewer didn't seem to think it made a huge difference between Newcastle and Clawhammer. My San Diego is coming soon!

StarfishChris
Posts: 968
Joined: Fri Jan 07, 2005 7:13 pm
Location: Bristol, UK
Contact:

Post by StarfishChris » Mon May 02, 2005 11:13 am

It's at odds with the Xbit review - although this looks like a better comparison.

Mertz
Posts: 35
Joined: Sat Mar 12, 2005 5:38 am
Location: Sweden
Contact:

Post by Mertz » Mon May 02, 2005 11:30 am

StarfishChris wrote:It's at odds with the Xbit review - although this looks like a better comparison.
Grrr, I was about to post that when I refresh and see that you've already done that -_-

Anyways, does this mean that the Venice is supperior to the Winchester in every way (don't care pricewise) ? Or is Xbit the more reliable source?

Green Shoes
*Lifetime Patron*
Posts: 477
Joined: Thu Feb 17, 2005 6:41 am
Location: Nashville, TN

Post by Green Shoes » Mon May 02, 2005 11:41 am

Mertz wrote:Anyways, does this mean that the Venice is supperior to the Winchester in every way (don't care pricewise) ? Or is Xbit the more reliable source?
Possibly, if only by a tiny margin. If you already own a Winny, unless you can sell it for a good bit of money (not likely now that Venice is available) it's probably not worth it to upgrade. If you're still using a 130-nm A64 or something older, it's quite worth it.

EDIT:

Don't forget too that the XBit article was comparing what was basically a first-week Venice to a very mature late-week Winny....perhaps if they had used one of their first Winchester engineering samples they would have gotten much different results.

capleton
Posts: 41
Joined: Tue Apr 26, 2005 2:32 pm
Contact:

Post by capleton » Mon May 02, 2005 12:00 pm

im building a new system, and had intended on using the winchester 3000, would upgrading to the venice be worth it? as in thermal performance/price?

sorry, this might not be the right place to post that question :?

Mertz
Posts: 35
Joined: Sat Mar 12, 2005 5:38 am
Location: Sweden
Contact:

Post by Mertz » Mon May 02, 2005 12:12 pm

Green Shoes wrote:
Mertz wrote:Anyways, does this mean that the Venice is supperior to the Winchester in every way (don't care pricewise) ? Or is Xbit the more reliable source?
Possibly, if only by a tiny margin. If you already own a Winny, unless you can sell it for a good bit of money (not likely now that Venice is available) it's probably not worth it to upgrade. If you're still using a 130-nm A64 or something older, it's quite worth it.

EDIT:

Don't forget too that the XBit article was comparing what was basically a first-week Venice to a very mature late-week Winny....perhaps if they had used one of their first Winchester engineering samples they would have gotten much different results.
Well, I don't got a Winny or any A64... got a very crappy comp atm which I won't even bother giving the specs of :P

Getting a new rig very soon though and considering that the Venice is almost available at my local stores I was wondering if it truly is superior in every way. Mostly concerning temps & power consumption ;) As far as CPU power goes, the winnie has more then enough of that (for my needs). :P

yeha
Posts: 292
Joined: Thu Jan 13, 2005 7:54 pm

Post by yeha » Mon May 02, 2005 12:40 pm

i don't like it, those numbers look too low. is it a written law that all cpu power is delivered from the aux connector? i know some a64 setups work fine (albiet sub-optimally) when it's not connected, so there must be a system in place to draw cpu power from the main atx plug instead of the aux.

tay
Friend of SPCR
Posts: 793
Joined: Sat Dec 06, 2003 5:56 pm
Location: Boston, MA
Contact:

Post by tay » Mon May 02, 2005 1:10 pm

StarfishChris wrote:It's at odds with the Xbit review - although this looks like a better comparison.
I mentioned this on slashdot and someone replied saying the device used was not accurate. Check this out http://slashdot.org/comments.pl?sid=148112&cid=12412655

Anyway its about the same as winchester.

Mikael
Posts: 206
Joined: Mon Dec 06, 2004 3:12 am
Location: Gothenburg, Sweden

Post by Mikael » Mon May 02, 2005 1:14 pm

yeha wrote:i don't like it, those numbers look too low. is it a written law that all cpu power is delivered from the aux connector? i know some a64 setups work fine (albiet sub-optimally) when it's not connected, so there must be a system in place to draw cpu power from the main atx plug instead of the aux.
I thought about that too, but then again, the numbers for the other CPUs look fairly accurate. The drop in power usage when going to 90nm is nothing short of amazing.

Another thing to think about is that it would almost be too easy of an error if he actually missed something like that. :P It should be the first thing to investigate thoroughly.

ilh
Posts: 313
Joined: Sat Jan 08, 2005 3:31 pm
Location: Acton, MA, USA -- Folding for SPCR

Post by ilh » Mon May 02, 2005 1:24 pm

They did do go through the probe three times. While that doesn't help the %error of the measurement, it does help the floor. Effectively, they can measure down to 0.8A through the 12V, which is down to about 10W. As to the %error, that would likely be the same across all the CPUs, so comparisons would still be meaningful.

The single sample is a more serious issue, especially since some other single sample tests have gone the other way (Venice consuming more power than Winchester). Eventually as more tests come out we'll get a clearer picture. In the mean time, I am assuming that Winchester and Venice are fairly comparable from a power perspective. Maybe Venice will overclock and/or underclock better, but there are significant sample-to-sample differences for overclocking and undervolting within Winchester, so I think it is too early to tell.

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Mon May 02, 2005 2:41 pm

StarfishChris wrote:It's at odds with the Xbit review - although this looks like a better comparison.
Old news! :lol:

dan
Posts: 1243
Joined: Sun Jan 25, 2004 2:01 am
Contact:

Post by dan » Mon May 02, 2005 2:58 pm

is it possible to make any statements about venice's undervolting potential, versus winchester.

if we assume that winnie and venice at teh same frequency and vcore puts out the same amount of heat/power consumption (say b/c the sample size of 1 is too small) but venice can undervolt more than winnie, then i would say venice is a winner.

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Post by NeilBlanchard » Mon May 02, 2005 3:24 pm

Hello:

Here's a better place to have this thread:

http://forums.silentpcreview.com/viewtopic.php?t=21743

frostedflakes
Posts: 1608
Joined: Tue Jan 04, 2005 4:02 pm
Location: United States

Post by frostedflakes » Mon May 02, 2005 6:25 pm

Venice almost seems to perform too well, raping even the 1mb 2.4GHz Claw in most every test (not just the ones involving power consumption). But who knows, maybe this new revision is just that good. I personally can't wait to get my hands on a San Diego.

egghat
Posts: 73
Joined: Fri Mar 07, 2003 5:38 am

Post by egghat » Tue May 03, 2005 12:03 am

The number are lower than the 35watts TDP of the new Turion. If these were normal, the Turion wouldn't make any sense. I'm sceptical ... though if correct, these Venices would be a killer.

Bye egghat.

StarfishChris
Posts: 968
Joined: Fri Jan 07, 2005 7:13 pm
Location: Bristol, UK
Contact:

Post by StarfishChris » Tue May 03, 2005 12:31 am

Mats wrote:
StarfishChris wrote:It's at odds with the Xbit review - although this looks like a better comparison.
Old news! :lol:
I never said it was :!:

Mikael
Posts: 206
Joined: Mon Dec 06, 2004 3:12 am
Location: Gothenburg, Sweden

Post by Mikael » Tue May 03, 2005 2:00 am

egghat wrote:The number are lower than the 35watts TDP of the new Turion. If these were normal, the Turion wouldn't make any sense. I'm sceptical ... though if correct, these Venices would be a killer.
Well, they do run at the same default voltage as the Turion 35W. Also, it's not unusual that actual power consumption is lower than the TDP. As mentioned in the article, TDP is probably a simulated number for when loading every single transistor on the chip.

lm
Friend of SPCR
Posts: 1251
Joined: Wed Dec 17, 2003 6:14 am
Location: Finland

Post by lm » Tue May 03, 2005 5:36 am

Mikael wrote:
egghat wrote:The number are lower than the 35watts TDP of the new Turion. If these were normal, the Turion wouldn't make any sense. I'm sceptical ... though if correct, these Venices would be a killer.
Well, they do run at the same default voltage as the Turion 35W. Also, it's not unusual that actual power consumption is lower than the TDP. As mentioned in the article, TDP is probably a simulated number for when loading every single transistor on the chip.
Aren't AMD TDP's actually kind of future-proofed, so that they don't have to change the tdp when introducing new clockspeeds of the same core.

frostedflakes
Posts: 1608
Joined: Tue Jan 04, 2005 4:02 pm
Location: United States

Post by frostedflakes » Tue May 03, 2005 7:20 am

According to AMD, "Thermal Design Power (TDP) is measured under the conditions of Tcase Max, IDD Max, and VDD=VID_VDD, and include all power dissipated on-die from VDD, VDDIO, VLDT, VTT, and VDDA."

In other words, AMDs TDP numbers reflect worst-case scenarios, and should be taken with a grain of salt. It is unlikely that in a desktop environment, one will be running the processor at it max temp, where it draws the most power. Hence the real-life power consumption numbers are much lower.

It also explains why the TDP numbers for Turion seem so high. Tcase Max for the mobiles in somewhere in the range of 95*C, compared to 65-70*C (something like that) for the desktop Venice/San Diego. So Turion is tested under much more demanding conditions. In a desktop environment they would probably draw much less power than even Venice/San Diego, although they probably even out in a mobile environment where spaces are cramped and operating temps are higher.

HammerSandwich
*Lifetime Patron*
Posts: 1288
Joined: Sat Oct 25, 2003 3:21 pm
Location: 15143, USA
Contact:

Post by HammerSandwich » Tue May 03, 2005 11:10 pm

ilh wrote:As to the %error, that would likely be the same across all the CPUs, so comparisons would still be meaningful.
Doesn't the mobo VRM's efficiency vary with load?

Mertz
Posts: 35
Joined: Sat Mar 12, 2005 5:38 am
Location: Sweden
Contact:

Post by Mertz » Wed May 04, 2005 1:09 am

Another review of the Venice 3800+ from neoseeker.com

Not much of a heat/wattage comparison although it does mention that the vcore needed for the Venice is lower compared to the Winchester.

Im guessing that the Venice is preferable... if it is in stock by the time I get the rest of my comp together Im getting it although I wouldn't see any need to upgrade from the Winchester to Venice... the difference isn't that amazing.

egghat
Posts: 73
Joined: Fri Mar 07, 2003 5:38 am

Post by egghat » Wed May 04, 2005 4:44 am

If the Venices run on the same voltage as the Turion, the 30 watt measurement would be quite plausible. Strange, that xbitlabs measured something completly different.

But all I know is that the core voltage of the Venices is 1.40 volts (e.g. here), which is way higher than Turion's stock voltage (1.1 volt IIRC).

Bye egghat

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Wed May 04, 2005 5:10 am

egghat wrote:But all I know is that the core voltage of the Venices is 1.40 volts (e.g. here), which is way higher than Turion's stock voltage (1.1 volt IIRC).
A64E: 1.4 V or 1.35 V, both available.

T64 ML: 1.35 V
T64 MT: 1.2 V

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Thu May 19, 2005 8:06 pm

Lostcircuits have reviewed the AMD Athlon 64 X2 4800+. It looks like going from 130 nm to X2 is not a big deal (63 W), but the X2 doesn't run at 100 % load with only one Prime95 running.... They get 86 W when running two instances of Prime95.

halcyon
Patron of SPCR
Posts: 1115
Joined: Wed Mar 26, 2003 3:52 am
Location: EU

Post by halcyon » Thu May 19, 2005 10:21 pm

A Finnish overclocker Sampsa Kurri has tested A64 X2 4800+ with S&M 1.6beta (the strongest heat generator, more efficient than CPUBurn or Prime95 or anything else known to man).

A64 X4800+ results
==============
CPU internal probe: 63 C
Underside of the sink (thermal probe): 45,4C
System power consumption (kill-a-watt reading): 236W

As a comparison, Intel Pentium Extreme Edition 840 settled the temps at the following (same loading with S&M):

Intel Pentium EE 840 results
====================
CPU internal probe: 91 C
Underside of the heatsink: 64,2 C
System power consumption (kill-a-watt reading): 315W
(system crahed after 8 minutes of running with the stock Intel cooler)

Above measure were at stock speeds / voltages.

Note, S&M does load the CPU more than anything out there (test it yourself, if you don't belive me). It can easily reach temp +5C higher or more compared to prime95 (any test) and even better than CPUBurn. You need to run it with safe mode disabled and using 100% load setting (both are internal options in S&M).

To me it looks like X2 4800+ can still be cooled although stretching it (in a silent air cooling scenario). P4 840 again... no.

StarfishChris
Posts: 968
Joined: Fri Jan 07, 2005 7:13 pm
Location: Bristol, UK
Contact:

Post by StarfishChris » Mon Jun 06, 2005 5:06 pm

Looking on this review and assuming their measurements are correct, my Venice 3200+ running at 1.1v consumes 15.4W at load according to CPU Power. With no cooling at all other than the passive heatsink it takes about two minutes to rise a degree, so it can't be producing much heat. I'm pleasantly surprised by it and can't wait for ASUS to answer my mail and tell me how to go below 1.1v ;)

I don't know how well P-M can undervolt, but this could be a faster and/or cheaper solution.

perplex
Posts: 298
Joined: Wed Apr 13, 2005 1:29 pm
Location: Club Tropicana

Post by perplex » Mon Jun 06, 2005 5:41 pm

wow! you sure thats an accurate reading?

pony-tail
Posts: 488
Joined: Sat Aug 23, 2003 4:39 pm
Location: Brisbane AU

Post by pony-tail » Mon Jun 06, 2005 8:32 pm

Those results put the 4800X2 at lower wattage and temps than my 3.4 gig northwood P4 wich means that a dozen or so good quality heatsinks on the market can cool it very quietly It looks to be a good way for me to go actually if I can rake the $$$

The I
Posts: 41
Joined: Tue Jun 07, 2005 1:19 am
Location: Denmark

Post by The I » Tue Jun 07, 2005 1:26 am

The Venice should be superior to Winchester, at least on paper. Apart from adding SSE3-instructions the other mayor news with Venice is that it employs stretched silicon, which is the same method that Intel uses in the Prescotts (don't let that scare you away, though). I believe IBM and AMD claims that their method to stretch silicon offers 24 % faster transistor switching at the same amount of power, which I believe should translate into an equally larger performance/watt-increase.

Actually I believe the lostcircuits-article comments on this and their readings seem to confirm that the Venice actually does offer this MHz/watt advantage.

This said it's probably possible that these early Venice-yields aren't consistently superior to Winchester. But at least I'm not in doubt about which processor I'll use when I build a rig in August, I think Venice will have matured by then.

Post Reply