Anand,
Are you now allowing the really annoying pop over ads in flash (or whatever they are coded in) because i just saw a Circuit City one on the 7800GT article when i accessed it.
I know it's a very tricky area because of the variables and testing conditions required, but if you could give us as much information as possible and try to keep everything as close as you can between platforms (vis a vis power usage), it would be greatly appreciated. Also, if you could do some battery benchmark testing as well...
A good example of what I mean is that in looking through all of the reviews on the web that compare Dothan and Turion, I've noticed that they produce almost identical power results using the ATI graphic chipsets but wildly different ones using the Nvidia (possibly due to PowerMizer)...
I realise that the battery usage is more dependant on which specific laptop you choose than on which CPU, but I'd like to get a feel for why...
I think it' would be much better to just isolate the CPU's power draw and use that instead (sure you'll get inaccuracy due to the different motherboards' VRMs but it'd be a closer thing, that way everyone can draw his/her own conclusions).
This is sorta OT I know, but a while back there was an AMD64 overclocking guide, with a note in it about an upcoming X2 overclocking followup. Any news on that coming out anytime soon on the site?
What Anand fails to realize is Dothan system is using Asus CT-479 adapter. What I noticed with Anandtech's articles is that the new test systems use default memory, while the systems that are tested AGAINST it are using tweaked timings. So DDR2-533 used for Yonah would have not optimal timings(4-4-4-8), while Asus CT-479 adapter based Dothan would be using optimal DDR400 speeds. And looking at gaming benchmarks and memory latency, it benefits a LOT from low latency memory.
I am curious to see how a Yonah system would do with some basic server applications. If it is lower power, it could make a good platform for a small business server or (even though a long shot) a good platform for blade and high density server enviroments.
Since one of main features of the Pentium M as well as one of Intel's future goals is low power consumption I think it would be a good time to try to improve you're testing in this area. Showing AC draw while better then nothing doesn’t tell us a lot about what the CPU is specifically doing regarding power usage.
A good example to follow would be http://www.silentpcreview.com/article265-page1.htm...">SPCR's test with various different PC configurations. I do realize that different boards will have slightly different power requirements but it would be a far more accurate then what you are currently doing.
I'd like to know if the YONAH supports VT texhnology as rumored lately; i'm going to buy a mac mini yonah ( well, i hope it's coming out in january ), and to be able to double boot would be nice indeed !
i imagine you can check this by booting LINUX and WINDOWS on your mobo ?
Since you are looking at how a new architecture compares against another, I would like to see the variables kept as close as possible. To me this means Yonah should be compared to an X2 with 2MB total of L2 cache. You could do this with the standard clock speed for the X2 and with it underclocked to 2.0GHz to give everyone an idea of the performance of the Yonah arhitecture to the K8 architecture, as well as retail part compared to (close to) retail part.
Also, for the power consumption part of the test, there are many variables at work. For example, there are the memory voltage differences, the differences in periphial power consumptions, and the chipsets. In fact the memory voltage was not even given for the Yonah setup. Furthermore, a mobile version of an X2 would use different transistors that consume less power. It is misleading and unfair to compare the Yonah setup to the X2 setup without even mentioning these differences.
To do a test comparing the two setups without mentioning these facts to the reader isn't fair. Since the article is comparing CPU architectures, the reader will assume that this means the different architectures account for the power difference. If it were an article comparing two systems or two products, then it would be fair because you want to compare them in the exact way that they will be used. Otherwise, you should explicitly state that multiple factors are in play.
quote: Since you are looking at how a new architecture compares against another, I would like to see the variables kept as close as possible. To me this means Yonah should be compared to an X2 with 2MB total of L2 cache. You could do this with the standard clock speed for the X2 and with it underclocked to 2.0GHz to give everyone an idea of the performance of the Yonah arhitecture to the K8 architecture, as well as retail part compared to (close to) retail part.
What kind of a stupid comment is that?? Whatever is the reason, AMD has decided to put 1MB on X2, while Intel decided to put 2MB on Yonah.
The more fair comparison would be Yonah with IMC, that would kick the pants off X2 anyday.
What does "under load" mean for the power comparison between X2 and Yonah? If close to a TDP max, and idle power is low, then the added 16 Watts would make for a very cool running chip as you were suggesting of course. I just want to know how cool is cool. 27 Watts TDP?
Sorry I wasn't more clear, under load is the same as it has been in previous articles - the system was running through our WME HD encode test, with CPU utilization pegged at 100%, but disk accesses are minimal (although the drive is still spun up).
In the article you said that the logic used to dynamically resize L2 cache could be hurting performance. Is it possible that single core yonahs could end up performing better than dual cores in some situations?
How the hell does anyone know that Conroe/Merom has more pipeline stages than Banias/Dothan/Yonah?? Intel has never stated the pipeline stages for Banias/Dothan/Yonah. They have said 14 for Conroe/Merom, but for all we know it may be 14 for Banias/Dothan/Yonah. Some people say they measured it, but they all are between 12-14, and tells ABSOLUTELY nothing about the pipeline stages EXCEPT its in that range.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
29 Comments
Back to Article
anandbrownsahib - Friday, December 16, 2005 - link
Why do we care if your South Indian a$$ went to the dentist?ksherman - Saturday, December 17, 2005 - link
dumb head...overclockingoodness - Saturday, December 17, 2005 - link
Stop reading the blog then. It's his site and he can write whatever he wants. Take it or leave it...overclockingoodness - Saturday, December 17, 2005 - link
Oh, and he's not South Indian - whoever gave you that ideamonsoon - Wednesday, December 14, 2005 - link
Anand, what's going on ? Is part 2 still in progress, or did you abandon it for various reasons ?Thanks for feedback and your great site
=)
h7o - Friday, December 9, 2005 - link
So want to tell us what happend with the article? Hope you intel didn't come down on you.monsoon - Thursday, December 8, 2005 - link
Should i be scared that i still don't see PART 2 online ( maybe you got new NDA clauses falling on your head ? )......or is it just that you're working hard ( as usual ) to give us the most thorough preview possible ?
=)
SocrPlyr - Thursday, December 8, 2005 - link
Anand,Are you now allowing the really annoying pop over ads in flash (or whatever they are coded in) because i just saw a Circuit City one on the 7800GT article when i accessed it.
Josh
Viditor - Tuesday, December 6, 2005 - link
I know it's a very tricky area because of the variables and testing conditions required, but if you could give us as much information as possible and try to keep everything as close as you can between platforms (vis a vis power usage), it would be greatly appreciated. Also, if you could do some battery benchmark testing as well...A good example of what I mean is that in looking through all of the reviews on the web that compare Dothan and Turion, I've noticed that they produce almost identical power results using the ATI graphic chipsets but wildly different ones using the Nvidia (possibly due to PowerMizer)...
I realise that the battery usage is more dependant on which specific laptop you choose than on which CPU, but I'd like to get a feel for why...
Cheers
Furen - Thursday, December 8, 2005 - link
I think it' would be much better to just isolate the CPU's power draw and use that instead (sure you'll get inaccuracy due to the different motherboards' VRMs but it'd be a closer thing, that way everyone can draw his/her own conclusions).Viditor - Friday, December 9, 2005 - link
I agree completely...but that's tough to do (not to mention dangerous for the chip) on a multilayered laptop mobo.
ksherman - Monday, December 5, 2005 - link
Why do articles keep disapearing??!?! I think it has happend more times theis year than I can remember happening EVER...pdoconnell - Saturday, December 3, 2005 - link
This is sorta OT I know, but a while back there was an AMD64 overclocking guide, with a note in it about an upcoming X2 overclocking followup. Any news on that coming out anytime soon on the site?IntelUser2000 - Monday, December 5, 2005 - link
What Anand fails to realize is Dothan system is using Asus CT-479 adapter. What I noticed with Anandtech's articles is that the new test systems use default memory, while the systems that are tested AGAINST it are using tweaked timings. So DDR2-533 used for Yonah would have not optimal timings(4-4-4-8), while Asus CT-479 adapter based Dothan would be using optimal DDR400 speeds. And looking at gaming benchmarks and memory latency, it benefits a LOT from low latency memory.coldpower27 - Monday, December 12, 2005 - link
Maybe Anandtech could alleviate this problem by testing all systems with JEDEC standard latencies for memory.If DDR400 is used 3-4-4-8 timings.
If DDR2-533 is used 4-4-4-12 timings.
If DDR2-667 is used 5-5-5-15 timings.
Or we could use the best possible timings available for both platforms.
If DDR400 is used 2-2-2-5 timings.
If DDR2-533 is used 3-2-2-8 timings.
If DDR2-667 is used 3-2-2-8 timings.
toelovell - Friday, December 2, 2005 - link
I am curious to see how a Yonah system would do with some basic server applications. If it is lower power, it could make a good platform for a small business server or (even though a long shot) a good platform for blade and high density server enviroments.Operandi - Friday, December 2, 2005 - link
Since one of main features of the Pentium M as well as one of Intel's future goals is low power consumption I think it would be a good time to try to improve you're testing in this area. Showing AC draw while better then nothing doesn’t tell us a lot about what the CPU is specifically doing regarding power usage.A good example to follow would be http://www.silentpcreview.com/article265-page1.htm...">SPCR's test with various different PC configurations. I do realize that different boards will have slightly different power requirements but it would be a far more accurate then what you are currently doing.
Live - Thursday, December 1, 2005 - link
I think an Opteron 170 939 is a better match for this one then the x2 3800. The 170 has the same speed and cache.Live - Thursday, December 1, 2005 - link
As always I forgot the positive. Very interesting reading! I'm really looking forward to reading more.monsoon - Thursday, December 1, 2005 - link
Hello Anand,...and thank you for another great preview !=)
I'd like to know if the YONAH supports VT texhnology as rumored lately; i'm going to buy a mac mini yonah ( well, i hope it's coming out in january ), and to be able to double boot would be nice indeed !
i imagine you can check this by booting LINUX and WINDOWS on your mobo ?
THANKS
huges84 - Thursday, December 1, 2005 - link
Since you are looking at how a new architecture compares against another, I would like to see the variables kept as close as possible. To me this means Yonah should be compared to an X2 with 2MB total of L2 cache. You could do this with the standard clock speed for the X2 and with it underclocked to 2.0GHz to give everyone an idea of the performance of the Yonah arhitecture to the K8 architecture, as well as retail part compared to (close to) retail part.Also, for the power consumption part of the test, there are many variables at work. For example, there are the memory voltage differences, the differences in periphial power consumptions, and the chipsets. In fact the memory voltage was not even given for the Yonah setup. Furthermore, a mobile version of an X2 would use different transistors that consume less power. It is misleading and unfair to compare the Yonah setup to the X2 setup without even mentioning these differences.
To do a test comparing the two setups without mentioning these facts to the reader isn't fair. Since the article is comparing CPU architectures, the reader will assume that this means the different architectures account for the power difference. If it were an article comparing two systems or two products, then it would be fair because you want to compare them in the exact way that they will be used. Otherwise, you should explicitly state that multiple factors are in play.
IntelUser2000 - Monday, December 5, 2005 - link
What kind of a stupid comment is that?? Whatever is the reason, AMD has decided to put 1MB on X2, while Intel decided to put 2MB on Yonah.
The more fair comparison would be Yonah with IMC, that would kick the pants off X2 anyday.
Eug - Thursday, December 1, 2005 - link
What does "under load" mean for the power comparison between X2 and Yonah? If close to a TDP max, and idle power is low, then the added 16 Watts would make for a very cool running chip as you were suggesting of course. I just want to know how cool is cool. 27 Watts TDP?Anand Lal Shimpi - Thursday, December 1, 2005 - link
Sorry I wasn't more clear, under load is the same as it has been in previous articles - the system was running through our WME HD encode test, with CPU utilization pegged at 100%, but disk accesses are minimal (although the drive is still spun up).Take care,
Anand
h7o - Thursday, December 1, 2005 - link
Nice to hear more yonah vs dothan benchmarks.In the article you said that the logic used to dynamically resize L2 cache could be hurting performance. Is it possible that single core yonahs could end up performing better than dual cores in some situations?
IntelUser2000 - Monday, December 5, 2005 - link
How the hell does anyone know that Conroe/Merom has more pipeline stages than Banias/Dothan/Yonah?? Intel has never stated the pipeline stages for Banias/Dothan/Yonah. They have said 14 for Conroe/Merom, but for all we know it may be 14 for Banias/Dothan/Yonah. Some people say they measured it, but they all are between 12-14, and tells ABSOLUTELY nothing about the pipeline stages EXCEPT its in that range.Quiksel - Wednesday, November 30, 2005 - link
This is what I get for casual browsing anandtech on a slow wednesday night:as sad as this sounds, check out the "About" page with the staff bio's.... they are mega-old!
Vinney is still your fiancee in that blurb! :)
cyberpt - Wednesday, November 30, 2005 - link
I would love to see how Yonah is compared to AMD Turion 64 CPU?? Can you run a performance test for it?vsridhar420 - Friday, December 2, 2005 - link
Yes. The real comparision would be with Turion.