AMD 35W/65W dual-core CPU review @ X-Bit Labs

[line]

URL: http://www.xbitlabs.com/articles/cpu/di ... cient.html

[autoboy]

And the best part is they don't overclock better than standard cores. This should keep the prices lower by keeping them out of the hands of overclockers.

[DaveSimmons]

Sadly, these CPUs don't really exist yet here in the US, even though they were paper-launched in May.

Major retailers like Newegg and ZipZoomFly haven't even added them to their product database since they have no estimated time of receiving any supplies.

I'm happy with my current (noisy) X2 gaming system but I may have to go with Core 2 for a quiet server and work system I'm planning.

[vfrex]

They seem to have concluded performance per watt for us, but I was hoping to see actual benchmark/performance numbers alongside the power consumption measures to determine just where the X2 3800 stands.

[Mikael]

They seem to have concluded performance per watt for us, but I was hoping to see actual benchmark/performance numbers alongside the power consumption measures to determine just where the X2 3800 stands.

Excuse me? Aren't there enough X2 versus Conroe benchmarks all over the web? Xbitlabs did their own performance testing of these CPUs a few weeks ago.

[vfrex]

There are no benchmark comparisons of the EE vs standard vs Conroe processors that I know of. The EE versions are cherry picked, but clocked lower. I wonder if perhaps they take less of a performance hit than their 85w TDP bretheren would when underclocked.

[=assassin=]

Interesting, I hadn't noticed this review, thanks for posting a link to it!

[ryboto]

Sadly, these CPUs don't really exist yet here in the US, even though they were paper-launched in May.

Major retailers like Newegg and ZipZoomFly haven't even added them to their product database since they have no estimated time of receiving any supplies.

zipzoom has the 65W chips in, its the 35W EESFF that no one seems to have.

[Le_Gritche]

The EE versions are cherry picked, but clocked lower. I wonder if perhaps they take less of a performance hit than their 85w TDP bretheren would when underclocked.

I'm under the impression you believe EE versions are clocked lower than their regular counterparts. That's wrong obviously. Contrary to Turions of which for exemple MT-32 and ML-32 weren't clocked the same IRC, an Athlon64 X2 3800+, an Athlon64 X2 3800+ EE and an Athlon64 X2 3800+ SFEE are at the same clock of 2.0 Ghz. Only the voltage differs : they are undervolted, not underclocked.

zipzoom has the 65W chips in, its the 35W EESFF that no one seems to have.

It shouldn't be long, they already appeared in Germany, and yesterday in France (although with a 5 day shipping delay and a single vendor currently).
For regular, EE and SFEE X2 3800+, prices are respectively 159€, 210€ and 390€.

[vfrex]

So the stock clock rate on standard and EE models of the X2 3800 is 2.0GHz. What was the article talking about here?

Note that there is nothing extraordinary that Energy Efficient AMD processors appeared in the market. It is known that there is a linear dependence of the CPU heat dissipation on processor clock speed and squared dependence on CPU Vcore. These ratios help the processor developer to create more economical choices. The CPUs with 65W maximum heat dissipation work at slightly lower clock speeds and support lover Vcore of 1.2V-1.25V. This is exactly how they manage to reduce the thermal power of the dual-core processors from 85W to 65W. As for the dual-core Athlon 64 X2 3800+ with the maximum heat dissipation of 35W, it is the 1.075V Vcore and only 2.0GHz clock rate. In other words, Energy Efficient AMD processors are nothing completely new. The manufacturer simply sorts out at the packaging stage those semiconductor dies that can work stably at lower Vcore. After that they simply lower their core voltage and limit the maximum clock frequencies and voila! – you get the desired wattage.

[Mikael]

So the stock clock rate on standard and EE models of the X2 3800 is 2.0GHz. So what was the article talking about here?

Note that there is nothing extraordinary that Energy Efficient AMD processors appeared in the market. It is known that there is a linear dependence of the CPU heat dissipation on processor clock speed and squared dependence on CPU Vcore. These ratios help the processor developer to create more economical choices. The CPUs with 65W maximum heat dissipation work at slightly lower clock speeds and support lover Vcore of 1.2V-1.25V. This is exactly how they manage to reduce the thermal power of the dual-core processors from 85W to 65W. As for the dual-core Athlon 64 X2 3800+ with the maximum heat dissipation of 35W, it is the 1.075V Vcore and only 2.0GHz clock rate. In other words, Energy Efficient AMD processors are nothing completely new. The manufacturer simply sorts out at the packaging stage those semiconductor dies that can work stably at lower Vcore. After that they simply lower their core voltage and limit the maximum clock frequencies and voila! – you get the desired wattage.

What the author means is that the frequency ceiling for CPUs in the 65W line is slightly lower than the 89W line. Therefore, there aren't any 65W 5000+ (2.6GHz) or FX-60/FX-62.

So, all of the energy efficient versions perform exactly as the more power hungry ones.

[autoboy]

159€, 210€ and 390€?

Whoa, i'm not paying double for something I could do myself with CrystalCPUID. My X2 3800+ runs stable at 2Ghz and 1.1V. So I can't quite get as low as the SFEE X2 3800+ but it will do for me. I was hoping they would be a few dollars more ($ 15 or so). The only real advantage is you don't have to have a bios that undervolts to get the CPU to run cooler without software like CrystalCPUID. I'm always a little afraid to leave my computer on 24/7 with barely enough cooling for the processor at 1.1V I don't have a bios that undervolts so this would have solved my problem.

[Mariner]

Hey, as a matter of interest autoboy, have you tried running the TCaseMax utility to find out the TDP for your X2?

If you knew this you could use CPU Power to calculate how low your X2 gets in comparison to these new SFEE parts. For example, if your X2 was one of the ones with around 60W TDP at stock voltage, you'll be down to around 40W at 1.1V anyway - not very far off the new low power parts!

(Assuming I'm working this out correctly!)

[JazzJackRabbit]

I glanced at the review (don't have much time) and unless I missed something I'm not particularly impressed. Conroe 6300 which is rated 65W easily beats AMD 65W offering, and is not that far behind 35W version.

The difference is even smaller under load which is where it matters. I don't really care how energy efficient it is when idle because it's easy to cool anyway, but I do care if it stays cool under load so that my fans don't ramp up. The air coming out of my exhaust nexus at 7V is only slightly warm, and that's with 3.4GHz P4 rated 90W inside, so like I said I don't really care if it runs cool idle, they all do, what I care about is load. And under load there is not enough difference in power consumption to justify paying that much extra money especially considering performance differences. Intel is looking better and better.


PS of course, that's taking everything for the face value, i.e. that Intel as well as AMD energy ratings are correct and that xbitlabs review also can be trusted.

[Mats]

And the best part is they don't overclock better than standard cores. This should keep the prices lower by keeping them out of the hands of overclockers.

Not so sure about that, look what amdzone did in their test, and that's with stock Vcore.

[Le_Gritche]

159€, 210€ and 390€?

Whoa, i'm not paying double for something I could do myself with CrystalCPUID. My X2 3800+ runs stable at 2Ghz and 1.1V. So I can't quite get as low as the SFEE X2 3800+ but it will do for me. I was hoping they would be a few dollars more ($ 15 or so). The only real advantage is you don't have to have a bios that undervolts to get the CPU to run cooler without software like CrystalCPUID.

I do wonder about the different C'n'Q steps on these EE CPU : do they undervolt even lower (like 0.6V or something ?) or are they limited by the motherboard VRMs and can only use the same steps than regular CPUs ?
At idle the X2 3800+ SFF EE used 11W with C'nQ and 14W without it ( page 6 of the article) It could be due to underclocking alone. Anyway, it hardly matters at such a low power consumption.

Sorry about the wrong trend I did set, but the 35W AMD's CPU are in fact called SFF EE not SFEE, got me confused with all the looking-alike double letters.

About the price, as I said there's currently only one French vendor having the SFF EE CPUs, so the lack of competition has probably an effect on the price.
In Germany the lower prices available are around 135, 167 and 347 € (according to thisAustrian price comparing website referencing German shops)

But the SFF EE X2 3800+ is still around double the price of the EE one.

If it remains like that, I would be content with an EE, which price premium is worth the savings in the trouble about manual undervolting and scavenging the web to find what are the capabilities of the BIOS of each and every motherboard I would consider buying.

[vitaminc]

Not going to pay for a ridiculously high price EE or SFF version when AMD's 65nm A64 X2 3800 has the same TDP as A64 X2 3800 EE.

[cAPSLOCK]

I don't really care if it runs cool idle.

Perhaps from a cooling standpoint that is true, but from a power usage standpoint idle consumption is the most important. I hadn't seen any core2 vs. X2 power comparisons yet, and this is quite surprising. The core2 has a better performance/watt, but in real life the X2 will probably use less juice.

Back on topic, I feel that the SFF EE is ridiculously priced at the moment, but the EE is probably worth it if you max out the CPU often enough.

I wonder how well the EE's undervolt, probably not much better than their standard counterparts I would imagine...

[widman]

I think intel eist didn't work in xbitlab test. Enable and disable eist still the same power consumption, 26W.

here from xbitlab with eist and cnq disabled .

[JazzJackRabbit]

Perhaps from a cooling standpoint that is true, but from a power usage standpoint idle consumption is the most important. I hadn't seen any core2 vs. X2 power comparisons yet, and this is quite surprising. The core2 has a better performance/watt, but in real life the X2 will probably use less juice.

X2 35W is only 12W more efficient in idle, I don't think I'm even going to notice the difference in my electricity bill assuming computer runs 10-12 hours a day on average (including weekends). The fluctuations in my overall electricity draw from month to month are more than that.

[cienislaw]

For regular, EE and SFEE X2 3800+, prices are respectively 159€, 210€ and 390€.

yup, almost the same in Poland (one reseller put them to stock today), with few euros in plus or minus. anyway its CRAZY to pay so much. i know sometimes its worth, but as examples shows e6300 isnt so power hungry, or you can undervolt regular one.

[jaganath]

as examples shows e6300 isnt so power hungry, or you can undervolt regular one.

Has anyone done a study of the undervoltability of Conroe? I know it has dynamic EIST for each individual core, which is pretty cool (hopefully literally), so has anyone tried it w/ RMClock or similar?

[smilingcrow]

Has anyone done a study of the undervoltability of Conroe? I know it has dynamic EIST for each individual core, which is pretty cool (hopefully literally), so has anyone tried it w/ RMClock or similar?

I have an E6300 that I'm evaluating at the moment; It has a stock Vcore of 1.325V and a frequency of 1.86 GHz.
I'm currently testing it for stability at 2.8 GHz at the stock Vcore and so far it's running dual Prime95 with no errors.
I briefly tried undervolting and it seemed fine at 1.164V, but I didn't run dual Prime95 for long; I will later and report back.

This thing runs so cool; I have a Nexus 120 at 620rpm on a Ninja and it's 45C at full load at 2.8 GHz
The only other fans are two other Nexus 120s at the same speed; one is an exhaust fan parallel to the CPU and the other is located in the power supply compartment of the P180 next to a Silverstone fanless.

I'm going to try lowering the FSB in the BIOS to give an underclocked CPU and see how low I can go with the voltage. Clockgen seems to work okay with raising the FSB in Windows. I'm trying to get a reasonable idle power consumption with the abilty to ramp up the FSB/Vcore manually when I need the power.

[ziphnor]

I notice most of the systems using close to 200W in idle, even though most of CPU's have a reasonably low power consumption, why is that? My A64 3000+ based HTPC can only be pushed to ~100W when running at max load(using CPU burnin or whatever its called and the equivalent GPU program). Is it the X1900XT eating all that power?

[Mariner]

Is it the X1900XT eating all that power?

Yep - high-end graphics cards are very big, hot chips - the R580 chip in the X1900XT contains 384 million transistors, not far off double the number in an Athlon 64 X2!

[ziphnor]

Yep - high-end graphics cards are very big, hot chips - the R580 chip in the X1900XT contains 384 million transistors, not far off double the number in an Athlon 64 X2!

I know this is the CPU/mobo forum, but i got to ask: Doesnt ATi has any kind of power saving mode or similar, how can it need to use that much power in IDLE?! Are the nvidia 7900 series just as bad?

And here i was hoping to upgrade my HTPC to a Core 2 system(which is why i was reading this thread ) with a mid/high-endish graphics card without my electricity bill going completely ballistic

[jaganath]

Are the nvidia 7900 series just as bad?

IIRC the stock 7900GT uses 64W, not sure whether this is the max consumption. ATi has a power-saving tech called PowerPlay, I think? But only for laptop GPU's.

[Mats]

Are the nvidia 7900 series just as bad?

None of them are really low power in idle, but the 7900 GT uses only 48 W under load and 23 W in idle.

[ziphnor]

Thanks for the info/link. I guess the 7900GT isnt that bad, but its still a jump up from 7600GT. I just dont think the 7600GT is just the thing for gaming at 1920x1080, which i was hoping to try out The passively cooled MSI 7900 sounds more like it.

Anyway, dont let me pollute this thread anymore, and thanks again for the info.

To get back on topic, i must say the Conroe has altered my HTPC upgrade plans. I had originally planned for the 3800+ 35W, but the price cuts i have seen locally have only applied to the 65+W versions. Furthermore, the pretty huge perfomance/watt increase supplied by the Conroe, combined with the ease of automatic standby in HTPC software has convinced me. But then im planning for a high-end HTPC that can, say, record 2 HD streams, and play a third one using all kinds of fancy filtering etc. My current A64 3000+ cant even handle 1080i MPEG-2 playback in MediaPortal.

This is going to be my first Intel CPU since my Pentium-II 450Mhz

[smilingcrow]

Furthermore, the pretty huge performance/watt increase supplied by the Conroe, combined with the ease of automatic standby in HTPC software has convinced me.

You’ve hit it on the head there; you need a Conroe system to go into standby as it can’t compete with an X2 at idle due to the higher minimum clock speed of C2D. For a typical desktop system which is at idle ~90% of the time, AMD still offer the lowest power consumption if you exclude Intel’s mobile parts, which outshine everything else.