which CPU/motherboard most efficient for E@H

debugas
debugas
Joined: 11 Nov 04
Posts: 170
Credit: 77,331
RAC: 0
Topic 192578

i am planning to buy a new laptop and i want to ask what type of technology would be more efficient to use for E@H ?

i define two types of efficiency here:

1) most energy efficient - spends least electric energy per WU

2) calculates WU per least time period

I've heard AMD are faster since they have larger CPU cache but intel CPUs use less energy (especially with its new core 2 duo technology). Does that intel tech use less power when i push it to 100% usage or is it just due to power management savings when running at low usage % ?

for my laptop i want to go with 1) type of efficiency

thanx

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494,410
RAC: 0

which CPU/motherboard most efficient for E@H

That doesn't sound 100% correct to me. While Einstein tends to run traditionally well on AMD CPUs, it's simply not true that AMDs in general have more cache than Intels. My brother's Pentium M (Dothan) notebook CPU has a full 2 MB of cache for a single core, whereas my AMD Athlon 64 3500+ (Venice) has only 512 KB although it's about 20% or so faster. So the amount of cache depends much more on the actual model of CPU you use. Btw, I don't think you really need a large cache for running Einstein. That is imo more important on other projects (eg SETI).
I would, however, agree that Intel CPUs use less energy at the moment. I have recently gotten myself a really nice Core Duo notebook and it's quite cool and also easy on the battery, especially for a dual core. AMDs tend to need more power, both desktop and mobile CPUs. I'm not really an Intel fan but atm they seem to be leading the race.
Hope I could help.
Annika

EclipseHA
EclipseHA
Joined: 19 Feb 05
Posts: 41
Credit: 10,540,182
RAC: 0

Also, the OS might be an

Also, the OS might be an impact... You may want to stick with XP!

See this thread on the main BOINC site:

"BOINC and Vista"

http://boinc.berkeley.edu/dev/forum_thread.php?id=1689

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,305
Credit: 1,585,302,682
RAC: 997,191

At present, the C2D generally

At present, the C2D generally (it's possible to cook tests that AMD wins in, but these are very much the exception not the rule) outperforms the A64x2 in both metrics. The turion64x2 performs roughly on par with the c1d on a clock for clock basis, but Intel's power management is marginally better. Unless battery life is your overwelming priority for a laptop though the difference is small enough that other features should be the deciding factor in a lowend laptop purchase.

Where AMD really had the lead over intel was when it was competing against the p4, but except for really low end junkboxes the p4 is no longer on the market. AMD is claiming a mid 20% floating point and a double digit integer performance edge over intels c2d platform for thier new CPU revision due mid this year. AFAIK they haven't released any 3rd party samples for benchmarking yet so any marketing numbers should be taken with a grain of salt. IIRC Intel's timeline for the transition to a 45nm process has slipped to early next year, so AMD might have a reasonably long time at number one again. Intels long term lead in manufacturing technology however is very much a factor in thier favor. With each new process scale being sized to roughly double the transitor density it gives them the option of making twice as many CPUs at a time, doubling the size of the cache, or something in between. This is in addition to the gain from the smaller feature sizes allowing higher clock rates (denser hardware) and consuming less power to operate at a given speed (lower total power consumption for the same speed, or even faster while holding the existing power level constant).

history
history
Joined: 22 Jan 05
Posts: 127
Credit: 7,573,923
RAC: 0

Debugas; I am going to agree

Debugas; I am going to agree with Annika. I prefer AMD processors because they simply have more arithmetic horsepower per cpu cycle. Most of my experience is with single core processors. Having said that, I will defer to the "duallie brothers" for further education. I am running one dual core rig, an AMD Windsor core 2.2ghz pushed to 2.5ghz. It is certainly more efficient to run two cores in a single box, than have the same amount of waste heat generated by multiple power supplies on single core rigs. Regarding my opening statement about arithmetic horsepower, I experimented with an Intel Prescott P4 2.4ghz processor pushed to 3.25 ghz. At this speed it performed at the pace of a stock 2 ghz Athlon crunching Einstein WU's. You are limiting your crunching potential by confining a fast processor in a laptop computer. The constraints imposted by laptop architecture will handicap the processor as performance motherboards are not designed for laptops. I build my own rigs and prefer Nvidia chipset motherboards.

Thanks for the Question, good luck on your new rig.

Regards-Tweakster

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494,410
RAC: 0

While I think you're right

While I think you're right that desktops can more easily be configured to be really powerful, I think one shouldn't forget that most people don't get their boxes for BOINC only, and when you are, for example, a student, or have a job which involves traveling around a lot, a notebook can be really useful. Personally, I couldn't do without. Of course, crunching power is an extra benefit ;-) Which reminds me... I've GOT to check how well that little duallie (nice word, tweakster) does crunching Einstein...

FalconFly
FalconFly
Joined: 16 Feb 05
Posts: 191
Credit: 15,650,710
RAC: 0

As it stands now, this is how

Message 61994 in response to message 61993

As it stands now, this is how it actually looks like (generally, not necessarily to EAH!) :

The Intel Core 2 Duo has an approx. 25% advantage over the AMD Athlon64 X2 in terms of performance per Clock cycle, part contributed by its Core, part by its 2MB unified L2 Cache. (4MB Unified to come with the arriving E6x50 series, as far as I can see)

At the same time, under full CPU load it uses less energy than even a comparably performing Athlon64 X2 EE (which retains an advantage in low CPU load power consumption due to Cool&Quiet and apparently lower leak currents in the Cores).

The comparison becomes difficult/impossible looking at Core 2 Duo E6600 and above, as there's no matching Athlon64 X2 anymore (the X2 6000+ or FX62 are - depending on Benches used - at best scratching short of the performance of a E6600)

-------------
Bottom line :
In the lower cost segment, intel currently has little to nothing to offer against the well-performing and still quite energy efficient smaller X2's, which simply win there because they're dirt cheap.

Going up the performance levels, however, the E6400 and above generally have little competition when it comes to Energy Efficiency combined with High Performance levels. The only catch is the relatively high price, but lacking serious competition in this Performance/Energy regime, it is justified.

Comparing my own stats over time, I find 2400MHz Athlon64 X2 4600's having a really tough time against 2133MHz E6400's for example. I reckon I'd need an X2 5000+ to be on par with an E6400.

Another Problem beginning with X2 5200+ is their rapidly increasing power consumption, the X2 5000+ EE (65nm) is almost nowhere to get hold of and no EE Models exist beyond that at this time.
I could easily afford a 6000+, but would never buy it because of the excessive power consumption, which sends the TCO (Total Cost of Ownership) into the sky over time.

If I hadn't been with AMD for so long and hadn't any upgrade path anymore, I'd go Core 2 Duo myself without hesitation.
Can't really go wrong with them, unless your Profile (e.g. Projects attached to) happen to just run better on the X2's or you're short on cash to spend - then the EE model X2's are still a very good buy.

AMD folks are basically waiting for the K8L/K10 design, which is supposed to yield a 40% FPU performance advantage over the existing Core 2 Duo design. Until then, the Brisbane 65nm Core has to save the day for AMD as good as it can.
(speaking of which, the increased L2 Latency don't seem to affect crunching from what I see off my X2 4000+ EE, so they remain completely viable for BOINC'ers)
-----------
On the Chipset choice, NVidia Chipsets are by far the most power-hungry ones existing.
The AMD RS690V/690G series apparently consumes alot less power (despite having integrated VGA), but haven't personally tested one yet.
The least power consumption (so far) I experienced with the older but matured VIA K8T890, which despite its age has no performance penalties for BOINC crunching.

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2,208,434
RAC: 0

FalconFly has some strange

FalconFly has some strange info about Core 2 Duo with 4MB cache, they been available since August 2006 =) The E6x50 series will have 1333 FSB.

I had both 2 and 4MB versions of Core2Duo and i must say 4MB is worth it, when both cores are crunching the 4MB versions leave more resources for doing heavy stuff. Not really a big problem with 2MB versions since E@H runs at low priority but the WUs will take up some cache no matter how low the priority is. Theres a clear difference between the two when running benchmarks while E@H is working in the background. In the same benchmarks without E@H running the difference is smaller so no doubt 4MB will give more torque when the load gets heavy. For laptops the 4MB versions are a bit expensive and a little more power hungry so 2MB is better bang for the buck.

AMD is a bad choice for laptop IMO, the reason Intel hasnt reduced prices much for laptop CPUs is that they see no threath from AMD at the moment. Right now i would only recommend AMD for desktops, the X2 3600+ is very cheap and moderate in power consumption, but maybe only untill later this April when Intel cuts prices, lets wait and see whats the best bang for the buck after that. AMD will have a comeback late this year when their new architecture is out, the Kuma dualcore should easily match Core2Duo clock for clock.

Quote:
(K10).. which is supposed to yield a 40% FPU performance advantage over the existing Core 2 Duo design


Thats plain wrong, the benchmark they used to show 42% advantage over C2Q was "throughput", specFP_rate. K8 already beat Core2 by approx 8% in that benchmark, so its quite lame actaully, unless K8 really is a better CPU, LoL. specFP is the benchmark that would have been interesting, add _rate and AMDs IMC makes all the difference. Bandwidth is good for memory benching but when used on CPU it has proven to have little to no impact on the real world raw processing performance. Anandtech had an article about this specFP_rate, and for my self i must add that i cant believe AMD thought they could get away with using such a biased benchmark. Compare it to when intel gave performance previews of Conroe, they were so sure of superior performance that they didnt care about handpicking benchmarks like they did in the past with Netburst Xeons. I was enthusiastic about K10 myself untill i found out what specFP_rate really is. =(

FalconFly
FalconFly
Joined: 16 Feb 05
Posts: 191
Credit: 15,650,710
RAC: 0

Oops, mixed up the E6x50 with

Message 61996 in response to message 61995

Oops, mixed up the E6x50 with the E6x20 series...

I think we'll have to wait until the K10 is actually released and we get real-world benches from neutral testers. Considering the delays, however, you might be right (and them trying to squeeze more performance out of it before release).

If AMD doesn't get it right by that time, they're in serious trouble and the improved 45nm Intel chips will already start knocking on the door.

I looked closer at what AMD is doing right now :
Seems that besides the 65nm Brisbane G1 Cores they've also started pushing out the Windsor F3 Stepping which brings previous 90nm 89/125W Windsor Cores downto 65/89W TDP respectively. At least they're working hard to improve the Energy levels, which is good...

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2,208,434
RAC: 0

RE: Oops, mixed up the

Message 61997 in response to message 61996

Quote:

Oops, mixed up the E6x50 with the E6x20 series...

I think we'll have to wait until the K10 is actually released and we get real-world benches from neutral testers. Considering the delays, however, you might be right (and them trying to squeeze more performance out of it before release).

If AMD doesn't get it right by that time, they're in serious trouble and the improved 45nm Intel chips will already start knocking on the door.

I looked closer at what AMD is doing right now :
Seems that besides the 65nm Brisbane G1 Cores they've also started pushing out the Windsor F3 Stepping which brings previous 90nm 89/125W Windsor Cores downto 65/89W TDP respectively. At least they're working hard to improve the Energy levels, which is good...

Yeah, sorry, i should have known you were talking about 6x20 series =)

I dont think its something wrong with K10s performance, my problem is with the selcted benchmarks, but for sure we need to see real benches of various performance scenarios first and that better happen soon if Mr Ruiz is serious about end of summer launch. Another thing is that so many people seems to think k10 has to beat Intels offerings to be a success, i believe even performance will be enough to make it successful, AMD has a much stronger market position and reputation now compared to when K8 launched. Im afraid K10 will be insanely expensive if the performance is superior, same story as X2 with premium prices for as long as they have the performance crown and only high end model offerings at first. Anyone remember how long we had to wait for X2 3800+ back in 2005? AMD really wanted to sqeeze the big money from buyers with only high end offerings at first and they probably would do it again if they get the chance. AMD can do it because they can sell all they can make anyway because of low production capacity compared to Intel.

Personally i will most likely go for Intel 45nm even if AMD gets a clear performance advantage, still havent seen any great overclocks on 65nm AMD so i doubt they can give the same overclocked performance per dollar.

Btw, Intels Kirk Skaugen was talking about something like constant thermal envelope on the Penryn 45nm processors (dont remember the terms he used). One core can clock itself up above standard frequency when the other core(s) has low workload, AMD has said they will more or less shut down idle cores, its like AMD will save that power and Intel will redirect it like in "Scotty, divert more power to main processor", and "(scottys accent) Aay Captan, diveerting from secondaary processors naww.". For me Intels solution sounds more promising, i think AMD should do the same to get to the performance geeks/gamers.

debugas
debugas
Joined: 11 Nov 04
Posts: 170
Credit: 77,331
RAC: 0

Thanx for replies. And yes

Thanx for replies.
And yes indeed the battery life for laptop is of highest priority.
I just was afraid that when C2D running at 100% its power consumption will be very much the same as AMD

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.