Yonah article here on Anandtech

[Zap]

Originally posted by: dexvx
When both laptops are idle (doing NOTHING), the ML performs better. But at load, the Dothan fares much better. I buy a laptop to use, not to idle. If all you do is idle on battery with a laptop, then the ML will be the better choice for that particular brand.

Some "uses" are pretty much at idle. Web browsing, e-book reading, DVD watching, office work... all these can be done at minimum clock speeds with ease.

Originally posted by: dexvx
And ultimately, that is the benefit of designing a package from the ground up as mobile.

This is Intel's real secret weapon.

Originally posted by: dexvx
The only effective way to test Turion vs Pentium-M is when an OEM produces the SAME SERIES of laptops but differing platforms. However, such an example is typically rare.

It is rare. Not impossible to do - potentially you can have same battery, chassis, HDD, optical, network, wireless, RAM, cooling, LCD panel... you can also even use similar chipsets if you go with ATI or SiS. Intel's mobile chipsets really help with the low power draw so it is difficult to compare just the CPU.

Originally posted by: dexvx
Normally it wouldnt matter in a desktop environment

Nope, it wouldn't.

 

[Viditor]

Originally posted by: Accord99

Originally posted by: Viditor
You say nearly identical...could you please list the exact specifications of each system and exactly what settings they are at? "Nearly" can be such a big place...

Same chassis, same cooling system, same video card, same HD, same amount of memory, they even used the same battery.

Different memory type, different wifi, and the Turion was set to 2T when 1T was perfectly stable. I think what this proves is that the power consumption has very little to do with the Turion OR the P-M, it has far more to do with the rest of the system. That's the only thing that can explain the wildly different benchmarks found on the various review sites...

 

[Leper Messiah]

Originally posted by: stateofbeasley
I don't know how much you read the front page of AnandTech, but it was all over the front pages at the end of August that Intel is moving to a new microarchitecture in Q3 '06. There's a new mobile core called Merom that is x64. It taped out last June and is up and running.

There will be no wait for a die shrink. Merom replaces Yonah at the high end 6-9 months after Yonah launches.

I try not to be exasperated with people on this forum, but I really really really think people need to know the manufacturer roadmaps before making arguments.

Originally posted by: mamisano
Yonah is due out in Q1, when do you think they are going to add 64-bit? Its not like the P4 when all they had to do was give the vendors the okay to enable P4 EMT, the 64-bit registers, etc are not Physically on the Yohah. So that means that either they wait for a Die shrink (2 years?) or move to a larger core to accomodate... Either way, don't expect Yonah to support 64-bit any time soon. Meanwhile, the move to 64-bit marches on.

I guess in the same token, then the next gen desktop chips from Intel should also NOT have 64-bit, as it is not "necessary"? Pretty lame excuse.

Mere hearsay. Wasn't Tejas supposed to get us to 10GHz by now? Yonah is like 5 months late too. Until we actually see Merom and crap, I'm not saying a damn thing.

 

[stateofbeasley]

Yonah is not 5 months late. It's been an early 2006 product for a long time.

http://www.anandtech.com/mb/showdoc.aspx?i=2264&p=2, a November 2004 article, references Yonah as appearing in 2006.

mamisano argued that Intel would not get x64 in mobiles for 2 years until a new shrink process. That is just eroneous.

Originally posted by: Leper Messiah
Mere hearsay. Wasn't Tejas supposed to get us to 10GHz by now? Yonah is like 5 months late too. Until we actually see Merom and crap, I'm not saying a damn thing.

 

[mamisano]

Originally posted by: stateofbeasley
Yonah is not 5 months late. It's been an early 2006 product for a long time.

http://www.anandtech.com/mb/showdoc.aspx?i=2264&p=2, a November 2004 article, references Yonah as appearing in 2006.

mamisano argued that Intel would not get x64 in mobiles for 2 years until a new shrink process. That is just eroneous.

Originally posted by: Leper Messiah
Mere hearsay. Wasn't Tejas supposed to get us to 10GHz by now? Yonah is like 5 months late too. Until we actually see Merom and crap, I'm not saying a damn thing.

No, what I said was, wait for a Die-Shrink -OR- increase the size of the core. Merom, which is expected to be out 2H of 2006 is based on the Conroe desktop chip. Similar to Yonah with 4MB Cache and 64-bit support. With 2x the cache and 64-bit support, the core is going to increase in size and so will the power consumption.

 

[hytek369]

dual core is the way to go

 

[dexvx]

Originally posted by: Viditor
Different memory type, different wifi, and the Turion was set to 2T when 1T was perfectly stable. I think what this proves is that the power consumption has very little to do with the Turion OR the P-M, it has far more to do with the rest of the system. That's the only thing that can explain the wildly different benchmarks found on the various review sites...

I think the statement in bold was pretty obvious beforehand.

However, many laptops do not have BIOS ability to tweak settings, such as memory timings, memory voltage, core voltage, etc etc. Laptops in general have this restriction. The majority of the times you will be running at SPD.

DDR-II uses about 50% less power than DDR, saving a few watts. However, you have to note that if the Pentium-M used DDR, it would be higher performing due to the inate higher latency of DDR-II.

 

[TuxDave]

Originally posted by: Leper Messiah
Mere hearsay. Wasn't Tejas supposed to get us to 10GHz by now? Yonah is like 5 months late too. Until we actually see Merom and crap, I'm not saying a damn thing.

If we want to be ENTIRELY accurate.

1) If Yonah is out early 2006, it's not late.
2) Tejas was originally a 5Ghz chip on the roadmap (now removed)

 

[stateofbeasley]

You should have acknowledged that Intel would have a mobile x64 part before. Your post implies that Intel will not have a mobile x64 part.

Merom is not based on Conroe. It is the reverse. Conroe is a Merom with relaxed thermal standards and more cache. Merom is also not similar to Yonah. The execution cores are completely different (Yonah still being P6 related while Merom is completely new).

I agree with you that die size will probably increase. Intel has already said that the TDP on the regular Meroms would be higher than Yonah, so no disagreement on higher power consumption either.

Originally posted by: mamisano
No, what I said was, wait for a Die-Shrink -OR- increase the size of the core. Merom, which is expected to be out 2H of 2006 is based on the Conroe desktop chip. Similar to Yonah with 4MB Cache and 64-bit support. With 2x the cache and 64-bit support, the core is going to increase in size and so will the power consumption.

 

[carlosd]

Originally posted by: dexvx

Originally posted by: carlosd
Here you again
Turion MT CPUs: 25W max TDP
Dothan with similar performance: 27W intel TDP at 75% load = 34W max TDP

Turion has a higher TDP than dothan??? Specify, you are talking about the ML line, but you ignore the MTs.
Pentium M doesn't dominate anything except the market sharing.

Ok, answer me this question. If the "real" Dothan TDP is 35W, then how come the Dothan wins by over 50% on the battery life at load when comparing a similarily equipped Turion-ML (with a TDP of 30W)?

http://www.laptoplogic.com/resources/detail.php?id=17&page=13

Your numbers are nice and daddy, but they are just synthetic marks of performance. There is far more to attribute to battery life than just raw TDP.

Turion ML TDP is 35W , look at specs, Pentium M is 34W max, this has something to do with the plataform and DDR2 use, but I am talking about Turion MT line MT not ML those 2 letters are very important.

 

[carlosd]

Originally posted by: dexvx

Originally posted by: Accord99

Originally posted by: Viditor
You say nearly identical...could you please list the exact specifications of each system and exactly what settings they are at? "Nearly" can be such a big place...

Same chassis, same cooling system, same video card, same HD, same amount of memory, they even used the same battery.

System Setup Link

CPU: Turion ML-40 (2.2Ghz) on a ?? vs P-M 760 (2Ghz) on a 915m / ich6m
Memory: Turion uses DDR while the P-M uses DDR-II, all 2x 512MB
HDD : same
Video : same
Screen: same
LAN: Turion uses 10/100 Realtek versus a Marvel 10/100/1000
Audio: Turion uses standard AC'97 versus HD Audio
Optical: same

Here you again, look for Turion MT comparisons , I will help you look for MSI M635 laptop benchies with Turion MT (MT remember MT not ML) against similarly configured Dothans and let's see which one has longer battery life.
Again Turion MT (YES the MT not the ML!!) is 25W it consumes less power than similar performance dothan. I am talking about a pure CPU comparison

 

[carlosd]

Originally posted by: dexvx

Originally posted by: Viditor
Different memory type, different wifi, and the Turion was set to 2T when 1T was perfectly stable. I think what this proves is that the power consumption has very little to do with the Turion OR the P-M, it has far more to do with the rest of the system. That's the only thing that can explain the wildly different benchmarks found on the various review sites...

I think the statement in bold was pretty obvious beforehand.

However, many laptops do not have BIOS ability to tweak settings, such as memory timings, memory voltage, core voltage, etc etc. Laptops in general have this restriction. The majority of the times you will be running at SPD.

DDR-II uses about 50% less power than DDR, saving a few watts. However, you have to note that if the Pentium-M used DDR, it would be higher performing due to the inate higher latency of DDR-II.

Oh I see finally you got it.

 

[Viditor]

Originally posted by: dexvx

Originally posted by: Viditor
Different memory type, different wifi, and the Turion was set to 2T when 1T was perfectly stable. I think what this proves is that the power consumption has very little to do with the Turion OR the P-M, it has far more to do with the rest of the system. That's the only thing that can explain the wildly different benchmarks found on the various review sites...

I think the statement in bold was pretty obvious beforehand.

However, many laptops do not have BIOS ability to tweak settings, such as memory timings, memory voltage, core voltage, etc etc. Laptops in general have this restriction. The majority of the times you will be running at SPD.

DDR-II uses about 50% less power than DDR, saving a few watts. However, you have to note that if the Pentium-M used DDR, it would be higher performing due to the inate higher latency of DDR-II.

We shall see...the Turion X2 will be using DDR-II, so we can make a better performance comparison when these all come out. My point is that having a Turion or a P-M based laptop will have very little difference on power consumption compared to who has manufactured the laptop. What they WILL make a difference on will be performance...and I'm still seeing the Turion having a very slight edge there (though certainly nowhere near the edge that AMD has over Intel in all other lines).
The other thing will be 64 bit.
So, to reiterate on the differences:

1. Power consumption - It doesn't matter. Both Turion and P-M will be the same. What matters is who has made the other components and how it's assembled.
2. Performance - Turion has a very slight lead here, though it's close enough to make very little difference.
3. 64 bit - Turion only. If you ever plan on using 64 bit software, then the P-M is not the CPU for you.


Edit: I forgot one very important metric, price! It looks like Intel has the edge here...not on the CPU, but certainly on system price (which is all that really counts right now). IMHO, this will change after the beginning of the year as new Turion based systems are announced...we shall see!

 

[Accord99]

Originally posted by: Viditor
Different memory type, different wifi, and the Turion was set to 2T when 1T was perfectly stable. I think what this proves is that the power consumption has very little to do with the Turion OR the P-M, it has far more to do with the rest of the system. That's the only thing that can explain the wildly different benchmarks found on the various review sites...

Only in idle conditions, can it sometimes matter. Under full load, the 50% edge that the P-M has over the Turion ML makes the small differences from such things as a wifi card unimportant.

 

[Viditor]

Originally posted by: Accord99

Originally posted by: Viditor
Different memory type, different wifi, and the Turion was set to 2T when 1T was perfectly stable. I think what this proves is that the power consumption has very little to do with the Turion OR the P-M, it has far more to do with the rest of the system. That's the only thing that can explain the wildly different benchmarks found on the various review sites...

Only in idle conditions, can it sometimes matter. Under full load, the 50% edge that the P-M has over the Turion ML makes the small differences from such things as a wifi card unimportant.

I'm sorry but absolutely nothing I've seen either here or on the web has shown the P-M to have even close to a 50% power consumption edge under full load over either Turion...

In the Presence PC test you posted, they state that the P-M clocked down both the CPU and the FSB (hence reducing performance), while the Turion didn't...this was a function of PowerMizer, and not speedstep or CNQ. That's why there was such a disparity between the tests done on the Nvidia (Presence PC) system and the ATI (Laptop Logic) system...

From Presence PC (translated)

In detail, one notes that this last manages to reduce the frequency of its processor to 507 MHz, against 800 MHz for Turion which manages to go down low on the level from the multiplying coefficient (x4 against x6), but which cannot modulate the frequency of its bus HyperTransport. However, at INTEL, the fall of the FSB not only allows to arrive to a weaker frequency CPU, but also to gain on the consumption of the memory (266 MHz - > 169 MHz), and probably also on the frequency of bond NCV Express train. It is as on this level as one measures the relevance to set up a mobile platform, and not just a processor

 

[Accord99]

Originally posted by: Viditor
In the Presence PC test you posted, they state that the P-M clocked down both the CPU and the FSB (hence reducing performance), while the Turion didn't...this was a function of PowerMizer, and not speedstep or CNQ. That's why there was such a disparity between the tests done on the Nvidia (Presence PC) system and the ATI (Laptop Logic) system...

[/quote]
That's for idle load, all clock optimizations were disabled for the cpuburn and Doom 3 test and both laptops were at full speed. The P-M laptop has a 30W edge in the Doom 3 test. The numbers match well with the other review:

http://babelfish.altavista.com/babelfis.../le-turion-dans-les-desktops/page4.php
Where a 2.13GHz P-M has a difference of 17W between idle with Speedstep and full speed with cpuburn. The 2.2GHz Turion ML is 42W and the MT is 26W.

 

[Sniper in a Diaper]

Originally posted by: michaelpatrick33
I am a little disappointed by the Yonah numbers. At 65nm and all the additions I thought it would give a little better showing than that, but it does at least compete with AMD. I love the smell of competition in the early morning ... err late evening.

Edit: The power numbers do look very good but not exactly overwhelming considering the 65nm advantage. If AMD matches their power numbers with their dualcore Turions at 90nm in Q1 '06 than Intel should be worried considering AMD still has 65nm.

Hint: If Yonah is consuming less power at 65nm than AMD's 90nm chip, that doesn't mean it will still hold that advantage once AMD switches over to 65nm.

 

[dexvx]

Originally posted by: carlosd
Turion ML TDP is 35W , look at specs, Pentium M is 34W max, this has something to do with the plataform and DDR2 use, but I am talking about Turion MT line MT not ML those 2 letters are very important.

The TDP of the Dothan is 27W. You're under the mistaken impression of desktop TDP's being at 75% of maximum, or whatever that is.

Moreover, the MT max'es at 1.8Ghz. It simply cannot compete with a 2.26 Dothan or a 2.2 ML in that sense. You could compare it to LV Pentium-M(which caps at 1.6Ghz / 2MB having a TDP of 10W).

Originally posted by: carlosd
I will help you look for MSI M635 laptop benchies with Turion MT (MT remember MT not ML) against similarly configured Dothans and let's see which one has longer battery life.

I'm waiting for the URL.

Originally posted by: Sniper in a Diaper
Hint: If Yonah is consuming less power at 65nm than AMD's 90nm chip, that doesn't mean it will still hold that advantage once AMD switches over to 65nm.

Process doesn't mean anything. When Prescott was taping out, people thought the transition would be similar as Willamette to Northwood; higher performance per clock, more cache, less heat per Mhz. Instead it was a disaster.

 

[Viditor]

Originally posted by: Accord99

Originally posted by: Viditor
In the Presence PC test you posted, they state that the P-M clocked down both the CPU and the FSB (hence reducing performance), while the Turion didn't...this was a function of PowerMizer, and not speedstep or CNQ. That's why there was such a disparity between the tests done on the Nvidia (Presence PC) system and the ATI (Laptop Logic) system...

That's for idle load, all clock optimizations were disabled for the cpuburn and Doom 3 test and both laptops were at full speed. The P-M laptop has a 30W edge in the Doom 3 test. The numbers match well with the other review:

http://babelfish.altavista.com/babelfis.../le-turion-dans-les-desktops/page4.php
Where a 2.13GHz P-M has a difference of 17W between idle with Speedstep and full speed with cpuburn. The 2.2GHz Turion ML is 42W and the MT is 26W.[/quote]

Again, you're quoting a SYSTEM POWER DIFFERENCE between being under load and at idle, and attributing all of it to the CPU. I just don't follow how you can do that...it's like saying that the difference between 2 golfer's scores in a tournament are strictly due to their choice of club manufacturers...

 

[Viditor]

Originally posted by: dexvx

Originally posted by: carlosd
Turion ML TDP is 35W , look at specs, Pentium M is 34W max, this has something to do with the plataform and DDR2 use, but I am talking about Turion MT line MT not ML those 2 letters are very important.

The TDP of the Dothan is 27W. You're under the mistaken impression of desktop TDP's being at 75% of maximum, or whatever that is.

Moreover, the MT max'es at 1.8Ghz. It simply cannot compete with a 2.26 Dothan or a 2.2 ML in that sense. You could compare it to LV Pentium-M(which caps at 1.6Ghz / 2MB having a TDP of 10W).

Originally posted by: carlosd
I will help you look for MSI M635 laptop benchies with Turion MT (MT remember MT not ML) against similarly configured Dothans and let's see which one has longer battery life.

I'm waiting for the URL.

Originally posted by: Sniper in a Diaper
Hint: If Yonah is consuming less power at 65nm than AMD's 90nm chip, that doesn't mean it will still hold that advantage once AMD switches over to 65nm.

Process doesn't mean anything. When Prescott was taping out, people thought the transition would be similar as Willamette to Northwood; higher performance per clock, more cache, less heat per Mhz. Instead it was a disaster.

dexvx is both right and wrong here...
1. He's right in that TDP isn't a power measurement at all...in fact it's not a measurement of any kind! It's a guesstimate and guideline of heat dissapation given by the manufacturer to OEMs.
2. Process CAN mean quite a bit...the 65nm process is the only real difference between the Prescott and Cedar Mill, and the Cedar Mill has a much better power/performance ratio. It was also a difference for the A64 when they moved to 90nm. There are however limits to this...Netburst tends to push at the edge of those limits quite quickly.

 

[Accord99]

Originally posted by: Viditor
Again, you're quoting a SYSTEM POWER DIFFERENCE between being under load and at idle, and attributing all of it to the CPU. I just don't follow how you can do that...it's like saying that the difference between 2 golfer's scores in a tournament are strictly due to their choice of club manufacturers...

Cpuburn is a program that heavily taxes the CPU. It doesn't go to memory, it doesn't use the video card. Therefore the vast majority of the power difference comes from the CPU.

 

[Fox5]

Originally posted by: dmens
Not the high level uarch, but low level implementation. Various decisions were made all over yonah to sacrifice performance. If merom made the right plays in terms of features that yield good ROI, the 50% gain is believable.

What will the absolute power be like? If Intel can increase performance 50% over Yonah (without a die shrink or more cores)....well they'd probably say so. Anyhow, that'd be Intel with the kind of performance lead an Athlon has over a P4 per watt, or the kind of absolute lead Intel had over old AMD and Cyrix chips.

DDR-II uses about 50% less power than DDR, saving a few watts. However, you have to note that if the Pentium-M used DDR, it would be higher performing due to the inate higher latency of DDR-II.

No it wouldn't, the 915 chipset performs worse with DDR than DDR2.

 

[Viditor]

Originally posted by: Accord99

Originally posted by: Viditor
Again, you're quoting a SYSTEM POWER DIFFERENCE between being under load and at idle, and attributing all of it to the CPU. I just don't follow how you can do that...it's like saying that the difference between 2 golfer's scores in a tournament are strictly due to their choice of club manufacturers...

Cpuburn is a program that heavily taxes the CPU. It doesn't go to memory, it doesn't use the video card. Therefore the vast majority of the power difference comes from the CPU.

Except that Cpuburn is optimized differently for P5, P6, K7, etc...it isn't the same on the Turion as it is on the P-M. This means that using it as a benchmark for power usage is fairly useless...

Edit: I should add that most ALL of these highly synthetic benchmarks are very misleading. (How often to you run an assembler loop on your laptop?)
It's one of the reasons that I am more impressed with the Laptop Logic review, as they used a varity of tests and reran them under a variety of conditions. It gives a much better indication of the true relative power usage and battery life...plus, the benches they used weren't optimized for one CPU or the other...

 

[Sniper in a Diaper]

Originally posted by: dexvx

Process doesn't mean anything. When Prescott was taping out, people thought the transition would be similar as Willamette to Northwood; higher performance per clock, more cache, less heat per Mhz. Instead it was a disaster.

An exception does not negate the rule. When you shrink things, you can't possibly have voltage differences that are as large as those that were previously tolerable.

 

[Accord99]

Originally posted by: Viditor
Except that Cpuburn is optimized differently for P5, P6, K7, etc...it isn't the same on the Turion as it is on the P-M. This means that using it as a benchmark for power usage is fairly useless...

Then you still have the Doom 3 test, where the P-M laptop used 30W less power for the same performance.

It's one of the reasons that I am more impressed with the Laptop Logic review, as they used a varity of tests and reran them under a variety of conditions. It gives a much better indication of the true relative power usage and battery life...plus, the benches they used weren't optimized for one CPU or the other...

All of laptoplogic's battery tests were synthetic tests.