Is this plausible? <FX-55 bows to the Pentium M?>

[yeha]

the turions are identical in terms of architecture, amd just chose to physically fabricate the cpu cores differently (in terms of transistor choices, etc.)

oc detective made a nice post about the differences here.

keep in mind though, that's the same process intel performed for their pentium-m line, and they overclock like champs. all we can do is wait and see what happens in the field.

 

[Super Nade]

The biggest attraction about the Turion is also potentially its biggest drawback. The Turion is not just an iteration of the A64 similar to previous mobile A64s but is a complete redesign (not of the architecture but what is used in the cpu). The transistor structure used in the Turion is completely different from normal A64s (e.g. removal of low Vt transisitors, use of low power transistors i.e high Vt transistors, no bottom capacitor in source/drain as well as a broader voltage spectrum). A consequence of this is that it is entirely possible that they wont run on current 754 mobos. This scenario is somewhat differnt to the Venice scaremongering going on though as the mobile is not supported on desktop 754 and this big change in transistor structure doesnt help matters. The other downside to the removal of low Vt transistors is that they potentially dont overclock as well as high Vt transistors have lower performance due to a lower drive current.

How does he conclude the part in red? Can't it mean you need lower voltages to achieve an OC? If so, this is certainly more efficient, right? What is meant by "a broader voltage spectrum" ?

Maybe OC Detective ought to stop by as well

 

[deception``]

I don't really see what the whole fuss is here.

Yes, the Dothan is a much more powerful processor than the FX-55. But the two should not be compared, simply due to the fact than that the Dothan is designed as a mobile solution only. In other words, Intel never had/has any intentions of releasing this as a desktop product, and any current desktop motherboard solutions for the Dothan are very unattractive.

I'm sure it has been said many times, but let's try to compare apples to apples here.

deception``

 

[yeha]

well typically, low-voltage-threshold transistors are faster than high-voltage-threshold ones, at the expense of more leakage. since the high vt transistors allowed amd to achieve satisfactory speeds with less power usage, they went with them. the downside is that they're unlikely to reach speeds as high as the leaky low vt ones can.

current cpu designs already pepper the dies with high and low vt transistors at appropriate points - critical paths get low vt for faster speeds, less used paths get high vt for less power consumption. for the turions, amd increased the ratio of high to low vt transistors to reduce power consumption even further.

i mean, if you have two cpus that are otherwise identical apart from the ratio of high vt to low vt transistors, the cpu with more low vts is very likely to both clock higher and use more power. if that weren't the case, all of amd's cpus would have veered towards using high vt transistors.

 

[Super Nade]

Why do you say that? What people look for is performance. If I can get a Pentium-M to perform better than a $800 FX-55, I'll take that! The XPM's were not designed to be desktop chips, nore were they supposed to run on duallies, people made the necessary mods and used them nonetheless.

In one of the links I posted in post #1 (the ASUS solution), the Pentium-D really stands up to the FX-x. Bear in mind, this is on a non-native/adapted motherboard solution.

If ASUS can develop such a mod, don't you think they may try to develop an integrated solution supporting P-4's and P-M's? You satisfy the mainstream as well as the enthusiast. If Intel are doing nothing to stop them now (obviously because of expected low volume sales of such a mod), what makes you think they are going to stop ASUS later? Looks like they wont care for a miniscule enthusiast market.

 

[deception``]

Why do you say that? What people look for is performance. If I can get a Pentium-M to perform better than a $800 FX-55, I'll take that! The XPM's were not designed to be desktop chips, nore were they supposed to run on duallies, people made the necessary mods and used them nonetheless.

Sorry SN, but most enthusiasts simply will not be willing to shell out such cash for a Dothan or an FX-55. On the other hand, most of us here are more concerned with netting high clocks for fairly low-to-moderate end processors; getting a large return for our money. Moreover, the Dothan processors, even with the Asus mod and DFI 855 boards, will still be in limited quantities far less than those of AMD's high-end FX-55.

Also, try to remember that the reasons for the Dothan's limited quantities are more than just performance. Without going into the matter fully, Intel cannot push the Dothan as a desktop solution, as it would take away from their efforts to turn the Prescott into a manageable and powerful processor. If they were to do such a thing (especially this soon), it would be essential an "admit of defeat" concerning their 90nm process, and could potentially raise skepticism amongst certain enthusiast (largely business-end) channels. As we already know, Intel has spent millions on R&D, marketing, and chipset development for their Prescott processors. One would have to imagine that the introduction of a newer single-core solution this soon could only be seen as a lack of confidence in their current (and relatively new) product. Basically, the leakage of this Processor could have much more damaging affects to Intel that far outweigh the benefits of a few pioneer sales.

Lastly, let's not forget that the reviewers are not stupid, either. Don't you think that others know of the Dothan's power? Sure they do. Guys from Anandtech, Toms, [H]ard, and all the other sites out there are more than aware of the Dothan's efficiency. However, they also know that, despite it's power, it is very unlikely to ever see this chip as a reasonable desktop solution. Hence, you'll see few mentions of this core, even when describing the FX and EE.

In one of the links I posted in post #1 (the ASUS solution), the Pentium-D really stands up to the FX-x. Bear in mind, this is on a non-native/adapted motherboard solution.

Nobody doubts the processors power, but the main point is that, while it is truly and efficient chip, it will be a cold day in hell before this thing is found in the average enthusiast's system.

If ASUS can develop such a mod, don't you think they may try to develop an integrated solution supporting P-4's and P-M's? You satisfy the mainstream as well as the enthusiast. If Intel are doing nothing to stop them now (obviously because of expected low volume sales of such a mod), what makes you think they are going to stop ASUS later? Looks like they wont care for a miniscule enthusiast market.

I don't think that Intel is too worried about Asus (and others) and their attempts to accomodate this processor. Unlike the Mobile Barton's (which were relatively inexpensive), few of these chips are passed throughout OEM channels; hence, Intel has no pressing threat concerning this processor.

deception``

 

[Sentential]

Here's the deal with Pentium-M. Its a great chip but you guys are overlooking a vast amount of information.

First and foremost, you are comparing a CPU (the FX) against an Intel CPU with 2X the cache. Dothan is NOT the most effcient x86 CPU. The Hammer core is more efficient than Dothan. Only reason it beats the 55 is because of cache.

Second Intel is aggresivly pushing the Pentium 5 onto desktop. Dothan IS the grandpappy of Pentium 5. You will see *very* quickly see Intel make such steps with the canning of the 5xx series CPUs in June.

In 2006 we will see a slew of changes on Intel's part. Here's whats gonna happen.

S479 --> S480. Yonah and later the dual-cored Memnon replaces Dothan.

LGA775 ---> LGA771 (I belive thats the pin count)
Smithsfield will be replaced with the dual core Prestler (start of Desktop Dothans) Prestler will later be replaced with the 1st Pentium5... the mighty Conroe.

Conroe is scheduled to be a quad core Dothan. It will support SSE3 and possibly 4 if they make it in time. It will also support Hyper-threading2 (HT for Pentium-M) as will Memnon.

Not alot of info available on HT2. No one really knows if they will improve the logical CPU count 4X or simply update it to support the Dothan cores. At the bare minimum Conroe will have 8 logcal CPUs....which is a **** load

It will be made on the 65nm process in early 2006 and should feature a whopping 8-16MB of cache (dont know yet). It will also support Vanderpool tech like Smithfield.

Interesting times are ahead

 

[HyperYagami]

Some people say P-M has a weaker FPU compared to A64.

 

[Sentential]

Some people say P-M has a weaker FPU compared to A64.

Because its a fact. It is inbetween netburst and the hammer design

 

[Gnerma]

First and foremost, you are comparing a CPU (the FX) against an Intel CPU with 2X the cache. Dothan is NOT the most effcient x86 CPU. The Hammer core is more efficient than Dothan. Only reason it beats the 55 is because of cache.

If it benches higher clock for clock, then its more powerful. End of story. It does not matter if one CPU has more cache then another, what matters are what numbers come out.

Do you have sources behind all of that other stuff or is it just your interpretation of Intels latest roadmaps? A quad core desktop Dothan in early 06? Give me a break.

 

[Super Nade]

Sorry SN, but most enthusiasts simply will not be willing to shell out such cash for a Dothan or an FX-55. On the other hand, most of us here are more concerned with netting high clocks for fairly low-to-moderate end processors; getting a large return for our money. Moreover, the Dothan processors, even with the Asus mod and DFI 855 boards, will still be in limited quantities far less than those of AMD's high-end FX-55.

Fully agree after looking up the latest prices that its not cost effective ATM for the average enthusiast. Surely this would change as volumes increase, right? If my low end P-M kills the high end 6xx wouldn't I go for that?

Also, try to remember that the reasons for the Dothan's limited quantities are more than just performance. Without going into the matter fully, Intel cannot push the Dothan as a desktop solution, as it would take away from their efforts to turn the Prescott into a manageable and powerful processor. If they were to do such a thing (especially this soon), it would be essential an "admit of defeat" concerning their 90nm process, and could potentially raise skepticism amongst certain enthusiast (largely business-end) channels. As we already know, Intel has spent millions on R&D, marketing, and chipset development for their Prescott processors. One would have to imagine that the introduction of a newer single-core solution this soon could only be seen as a lack of confidence in their current (and relatively new) product. Basically, the leakage of this Processor could have much more damaging affects to Intel that far outweigh the benefits of a few pioneer sales.

I'm not saying Intel are going to project the P-M as a DTR solution, it is the MoBo vendors who will try this to appeal to the enthusiast community.

Lastly, let's not forget that the reviewers are not stupid, either. Don't you think that others know of the Dothan's power? Sure they do. Guys from Anandtech, Toms, [H]ard, and all the other sites out there are more than aware of the Dothan's efficiency. However, they also know that, despite it's power, it is very unlikely to ever see this chip as a reasonable desktop solution. Hence, you'll see few mentions of this core, even when describing the FX and EE.

Nobody doubts the processors power, but the main point is that, while it is truly and efficient chip, it will be a cold day in hell before this thing is found in the average enthusiast's system.

Youa are basing your conclusions on current pricing. I fully agree with you regarding 0 effort from Intel to even hint that this could be a DTR. When did the enthusiast ever care about the actual intent of the manufacturer .

All I'm saying is if the P-M's are a lot cheaper than the FX-x, if they are easily available through OEM channels and if we have motherboard support. The avg 6xx user may consider this as a better option. Lot of IF's in there, yes I'm just indulging in a bit of idle speculating

Well, looks like I've beaten this one to death. Thanks for answering my original questions!

 

[Super Nade]

Second Intel is aggresivly pushing the Pentium 5 onto desktop. Dothan IS the grandpappy of Pentium 5. You will see *very* quickly see Intel make such steps with the canning of the 5xx series CPUs in June.

Interesting times are ahead

Yes Sir

I can't wait to see this new battle unfold!

 

[Sentential]

If it benches higher clock for clock, then its more powerful. End of story.

I *realize* that dothan is more powerful..but that is ONLY because of the cache which is what I said. The CPU design is NOT faster clock per clock just like a winchester isnt as fast as a claw.

 

[deception``]

I *realize* that dothan is more powerful..but that is ONLY because of the cache which is what I said. The CPU design is NOT faster clock per clock just like a winchester isnt as fast as a claw.

Clock-for-clock a 1MB Clawhammer DTR/Mobile desktop chip IS more powerful than a Winchester. Winchesters are mainly attractive for the 90nm die shrink (everyone loves cutting-edge), and the fact that it made s939 affordable.

deception``

 

[Gnerma]

Yea ok that makes alot of sense So lets compare how much more efficent and super-awesome the Clawhammer core is to say venice or winchester. That sorta logic is flawed if you want to make a direct comparison which you just tried to do

I don't understand what you are trying to say with this. What I am saying is that theoretical reasons why one CPU or another is better or worse than another is meaningless. Performance is the only thing that matters. If Intel puts 16mb of cache onto a dothan and it destroys everything else out there because if it, it is legitimately faster because the numbers say so. It is not cheating.

I *realize* that dothan is more powerful..but that is ONLY because of the cache which is what I said. The CPU design is NOT faster clock per clock just like a winchester isnt as fast as a claw.

Edited post reply: The CPU design clock per clock makes no difference to the end user. What matters is performance. Although, I understand why you'd think this way. You're the man who ditched a better performing AMD setup for an Intel rig because you were not satisfied with the overclock of the AMD.

*Gnerma gets personal*

 

[Sentential]

Clock-for-clock a 1MB Clawhammer DTR/Mobile desktop chip IS more powerful than a Winchester. Winchesters are mainly attractive for the 90nm die shrink (everyone loves cutting-edge), and the fact that it made s939 affordable.

deception``

The point is that the IPC is the same between a claw and a winch, yet the claw is faster. Why? Because of cache.

If you were to give Dothan 1MB of cache it would lose by a large margin to the FX. Much like if you were to give a clawhammer 2MB of cache.

 

[deception``]

The point is that the IPC is the same between a claw and a winch, yet the claw is faster. Why? Because of cache.

If you were to give Dothan 1MB of cache it would lose by a large margin to the FX. Much like if you were to give a clawhammer 2MB of cache.

I disagree slightly. The increased cache did not bring about a significant boost for the 6xx stepping processors over their older brethren (5xx Prescotts). However, I also believe that cache is probably a little more important to a Dothan, as it operates in a high-efficiency manner ala AMD. Despite this fact, one cannot simply say that cache alone is the determining factor in the Dothan's superiority.

Personally, I simply think that Intel made great developmental strides in an effort to provide a healthy, low-power mobile solution. The clocks and enthusiast attractiveness of the Dothan should be considered nothing more than a byproduct of this success.

deception``

 

[yeha]

well as far as the ipc goes, the only comparisons i've found at equal clock speeds are here and here, however they don't take into account the difference in memory bandwidth.

the only way core ipc could be settled would be comparing 2 chips with equal cache and memory bandwidth - say a 1.6ghz 90nm celeron-m and an fx-55 clocked down to 1.6ghz with a 1:2 memory divider. sure that's crippling the amd's bandwidth, but we're after the true ipc here, not any advantages that cache or fsb width gives you

truth be told if i had equally-inexpensive motherboards to choose from, i would build a system around a celeron-m or pentium-m today. it'd be just as fast and much, much easier to cool silently than an a64 system of the same speed, if i was aiming for ~2.5ghz with either platform. plus i could get away with value ram for the intel setup. i just think it's a pity the boards aren't there yet. they're coming though - did anyone else notice how much intel was hawking small footprint / low energy desktops at the last idf? even mocking up a mac mini clone? it won't be prescotts in those small boxes, that's for darn sure.

 

[caboob]

First and foremost, you are comparing a CPU (the FX) against an Intel CPU with 2X the cache. Dothan is NOT the most effcient x86 CPU. The Hammer core is more efficient than Dothan. Only reason it beats the 55 is because of cache.

I disagree. If anything, increasing the cache results in more latency because of more registers to address. The Dothan cache is unique in that it is of much lower latency than that used for 2MB Prescott and even for A64. It would've been interesting to see what performance this Dothan cache would have done for the Prescott.

 

[Sentential]

I disagree slightly. The increased cache did not bring about a significant boost for the 6xx stepping processors over their older brethren (5xx Prescotts).

What you are omiting is that they increased the latency of the L2 cache . They are equal because the latency on the 6xx is higher . Its not because they see no benefit from more cache. It would be like saying that Newcastle sees no benefit from having 1MB of cache (clawhammer).

With more cache it does see a sizable increase in peformance if all other variables (in pipelines / latency) remain the same. THis was not the case with the 600 Prescotts.

In an effort to combat thermal protocals they added EIST and kicked back the latency on the cache (as well as added some to compensate for this) This is why they see no benefit.

____________________

If you want a fair comparison lets see what Banias gets vs say the FX55. If I recall correctally Banias only had 1 Meg.