Should I upgrade to a Barton?

[zuperdee]

Currently, I have an AMD Athlon XP 2400+ (Thoroughbred) processor, running at 2.0 GHz, 266 MHz FSB. Will I be able to fold much faster with a 2800+ Barton? Is it worth upgrading? (I hate overclocking, and the 3200+ is a bit too expensive for my taste.)

[dukla2000]

Is it worth upgrading?

IMHO, No. The major bang/buck folding with Athlons comes from the CPU clock speed. Thereafter there is an occasional benefit from the Barton's larger cache. FSB itself (266/333/400) is worth nothing, as is memory size.

You are currently 2GHz (with SSE): a 2800 Barton is 2.083GHz so you will see about a 3% gain in general, and on some WU you will see another couple of % (say 3) because of the extra cache. So unless you get the Barton free or can justify it some other way it doesn't make sense. Then again this isn't always about making sense!

(My 2500 Barton - 1.833GHz is generally outpointed by my 2400 Tbred @ 2GHz.)

[Buddabing]

Currently, I have an AMD Athlon XP 2400+ (Thoroughbred) processor, running at 2.0 GHz, 266 MHz FSB. Will I be able to fold much faster with a 2800+ Barton? Is it worth upgrading? (I hate overclocking, and the 3200+ is a bit too expensive for my taste.)

If you were to get a Barton, get a Mobile Barton. They are great processors which run at lower voltages, and thus can be quieted more easily. I know you hate overclocking, but the Mobile Bartons are multiplier unlocked and are good overclockers.

But since you have a 2.0 Ghz processor already, and you don't want to overclock, it's probably not worth it to upgrade at this point.

[zuperdee]

Well, as I understand it, the mobile Barton has a few problems:

1) My motherboard does not allow the Vcore voltage to be adjusted, so at 1.65 volts, I guess I'd be overvolting the processor a bit.

2) The other problem is that the mobile Barton's FSB is only 266 MHz. Why would I want to be stuck with only a 266 MHz FSB?

3) My motherboard would probably not work with such a processor, since it does not have automatic FSB detection in the BIOS.

How about a Thoroughbred 2700+? Would that fold the fastest? It looks awfully enticing. Are these looks deceptive?

[dukla2000]

How about a Thoroughbred 2700+? Would that fold the fastest? It looks awfully enticing. Are these looks deceptive?

Well I run my (year old) TBred 2100 at 13*166 = 2.167MHZ = XP2700 and figure in the threads where Haysdb et al have posted Barton 3200 (2.2GHz) results I am not too far behind.

1 of my boxen is running 2GHz and (when not slugged by Tinkers or video processing) returns about 27/28000 points/annum: the 'XP2700' does about 32000. Comparing TBreds on the same proteins the perfomance is proportionately directly related to the CPU clock - fsb makes no difference. (In an earlier incarnation 1 system at 12*166 was doing the same as another at 15*133)

[vortex222]

i swiched from a desktop barton 2500+ to a mobile 2500+ simply so i could run the core at a lower voltage(board dosnt allow lowering more then 0.1 below stock). in folding i have not noticed any difference, tho i have not benchmarked it. however in the games i play like Lineage2 and unreal2. i noticed a drop in framerate, presumably from the drop in fsb speed. because the mobile is acualy slightly faster core clock.

And in geral usage my computer seemed to loose a little bit of steam in some things like DVD compressing and soforth. and dosnt seem quite as reponsive. I am wanting to figure out the bridges so i can change the multiplier on my board and run the high bus speed again.

considering 99% of what i use my computer for, it dosnt matter, and the cooler running operation is very welcome. it was worth it really. i sold the old chip and upgraded for a few dollers.

[zuperdee]

I don't understand how you can increase the FSB on a processor without making it super unreliable. My Thoroughbred 2400+ crashes before it gets off the ground if I try to do that. No way would I try to overclock the FSB.

I have read now also that the Thoroughbred 2700+ benchmarks with almost exactly the same results as the Barton 3000+. Methinks it is looking more and more attractive.

[haysdb]

I don't understand how you can increase the FSB on a processor without making it super unreliable. My Thoroughbred 2400+ crashes before it gets off the ground if I try to do that. No way would I try to overclock the FSB.

Depends on the motherboard and the cpu. I have had motherboards I could not overclock by 1 MHz, and on others I have been able to take a 166 MHz FSB processor directly to 200, without even increasing vcore.

David

[vortex222]

i would think my ultra400 and mobile chip could go higher then 266mhz. and should be ok if i dont increase the core clock. if i even had to reduce the core clock slightly to go from 266 to 400, then that would be acceptable.

[zuperdee]

Sounds like it's a real game of chance with it, eh? But it isn't so much the unreliability I worry about, as much as the fact that if it overheats, it could permanently damage the processor.

From what I've been able to find out, it sounds like the Thoroughbred core is absolutely 100% identical to the Barton, with the exception of the extra 256k L2 cache in the Barton. So should I do this upgrade? I do have a hidden motive, of course--I want to see if I can crush the rest of you in the folding contest. I want to be the best folder on this team.

[haysdb]

Zuperdee, I think you should just ADD the Barton, not replace the T'bred.

David

[zuperdee]

Zuperdee, I think you should just ADD the Barton, not replace the T'bred.

Hehe!! I would do that, were it not for the price, and the fact that I can't afford to build another complete system, motherboard and all. If I had the cash, I'd even try to build a dual Athlon MP system, and fold for 2CPU.com as well as SPCR. But for now, it is only a dream.

I am building a Barton 2500+ system for my mom though--is there any hope for this system to fold? (I just wanna make sure my system out-folds hers though.)

[ColdFlame]

I don't understand how you can increase the FSB on a processor without making it super unreliable. My Thoroughbred 2400+ crashes before it gets off the ground if I try to do that. No way would I try to overclock the FSB.

A CPU does not have a notion of the FSB. Motherboard does.

My Barton 2.4 runs pretty much equal to my Thorton 2.4 on most WUs.

[haysdb]

A CPU does not have a notion of the FSB. Motherboard does.

Would you clarify what you mean? FSB is most certainly a characteristic of a cpu.

My Barton 2.4 runs pretty much equal to my Thorton 2.4 on most WUs.

You mean both clocked to 2.4GHz? I would more or less agree. The Barton should be faster, but it is ultimately the clock speed that determines folding performance. L2 cache plays a lesser, but not insignificant, role. Memory speed plays only a minor role.

David

[zuperdee]

What exactly is a Thorton, anyway? So near as I can tell, it is basically just a Barton core with half the L2 cache disabled, effectively making it like a Thoroughbred, but with a larger die size. I assume they are probably doing this so they can sell those chips even if some of the L2 cache is found defective during testing. And why not do that? I'd be willing to bet it's a lot more economical simply to disable the defective portion of the cache and sell it, rather than throw the whole chip away.

[bcassell]

Would you clarify what you mean? FSB is most certainly a characteristic of a cpu.

Well, I didn't make the original statement, but I'll try to clear things up.

FSB isn't technically a charateristic of a cpu. A cpu has a rated speed, and possibly a fixed multiplier. IF a cpu has a fixed multiplier, then, in order to run the cpu at it's rated speed, you would have to run at a certain FSB speed. In this way, FSB can be thought of as an indirect characteristic of the cpu, but that is only by virtue of the fact that it has a fixed multiplier. A cpu rated for 2ghz could not care less whether it was being run at 20x100 or 10x200, it makes no difference to the cpu.

I beleive this whole part of the discussion started with the statement:

2) The other problem is that the mobile Barton's FSB is only 266 MHz. Why would I want to be stuck with only a 266 MHz FSB?

And the thing is, since the mobile Barton's multiplier is NOT locked, FSB is technically NOT a characteristic of that cpu. The above poster would not be limited to a 266mhz fsb, they could run any fsb they wanted, adjust the multiplier, and still run at stock speeds.

Now, obviously Intel and AMD consider the FSB as a characteristic of their cpus, and I'm sure AMD isn't going to give you any help if you have problems running your cpu at a higher fsb / lower multiplier, but my point is that technically, fsb is only a characteristic of the cpu by virtue of a fixed multiplier.

Bryan

[haysdb]

What exactly is a Thorton, anyway? So near as I can tell, it is basically just a Barton core with half the L2 cache disabled, effectively making it like a Thoroughbred, but with a larger die size. I assume they are probably doing this so they can sell those chips even if some of the L2 cache is found defective during testing. And why not do that? I'd be willing to bet it's a lot more economical simply to disable the defective portion of the cache and sell it, rather than throw the whole chip away.

Exactly right.

I'd be willing to bet the other half of the cache is just fine in most of those chips. It's cheaper to make just one die than two, and the chips will be configured based on market demand. If the market wants T'breds, they become T'breds. If the market wants Bartons, they become Bartons. I'm sure a certain percentage do have defective cache's and can ONLY be made into T'breds, but even if the yield reached 100%, you can be assured there would still be enough T'breds to fullfill market demand.

Let's take another example. VIA sells a variety of chipsets, for AMD, for Intel, full featured KT's and lower featured KM's. They are all made from the the same die, and are simply "configured" to be whatever is needed. Sure, a few million transistors are "wasted" that way, but it's cheaper to make just one wafer than to make one for each permutation of features.

David

[mpteach]

Would you clarify what you mean? FSB is most certainly a characteristic of a cpu.

Well, I didn't make the original statement, but I'll try to clear things up.

FSB isn't technically a charateristic of a cpu. A cpu has a rated speed, and possibly a fixed multiplier. IF a cpu has a fixed multiplier, then, in order to run the cpu at it's rated speed, you would have to run at a certain FSB speed. In this way, FSB can be thought of as an indirect characteristic of the cpu, but that is only by virtue of the fact that it has a fixed multiplier. A cpu rated for 2ghz could not care less whether it was being run at 20x100 or 10x200, it makes no difference to the cpu.

I beleive this whole part of the discussion started with the statement:

2) The other problem is that the mobile Barton's FSB is only 266 MHz. Why would I want to be stuck with only a 266 MHz FSB?

And the thing is, since the mobile Barton's multiplier is NOT locked, FSB is technically NOT a characteristic of that cpu. The above poster would not be limited to a 266mhz fsb, they could run any fsb they wanted, adjust the multiplier, and still run at stock speeds.

Now, obviously Intel and AMD consider the FSB as a characteristic of their cpus, and I'm sure AMD isn't going to give you any help if you have problems running your cpu at a higher fsb / lower multiplier, but my point is that technically, fsb is only a characteristic of the cpu by virtue of a fixed multiplier.

Bryan

I want to run a mobile barton 2600 @ 220 x 10.5 (2.32Ghz)

Obviously mobos have fsb limits. Is there any limit on the fsb of a cpu?

For isntance, if a mobo could handle high fsb speeds, could someone run a XP2600 @ 1155 x2 ?

[mpteach]

IMHO, No. The major bang/buck folding with Athlons comes from the CPU clock speed.

(My 2500 Barton - 1.833GHz is generally outpointed by my 2400 Tbred @ 2GHz.)

If nothing matters except clock speed why do people fold with athalons? Pentiums run at much higher clock speeds.

I thought that athalons excelled in applications were large efficient caches were neccesary.

[haysdb]

IMHO, No. The major bang/buck folding with Athlons comes from the CPU clock speed.

(My 2500 Barton - 1.833GHz is generally outpointed by my 2400 Tbred @ 2GHz.)

If nothing matters except clock speed why do people fold with athalons? Pentiums run at much higher clock speeds.

I thought that athalons excelled in applications were large efficient caches were neccesary.

This only applies within the same architecture. Clock for clock, an Athlon gets more work done than a P4, so the two cannot be compared directly.

Very generally speaking, a T'bred/Barton running at a given clock speed will perform about the same running Folding@Home, regardless of whether the cache is 256 or 512MB and regardless of whether the memory is running at 333 or 400 FSB. The additional cache and faster memory do have some influence over performance, but it varies from WU to WU and, in general, is not a major factor.

David

[zuperdee]

Now, obviously Intel and AMD consider the FSB as a characteristic of their cpus, and I'm sure AMD isn't going to give you any help if you have problems running your cpu at a higher fsb / lower multiplier, but my point is that technically, fsb is only a characteristic of the cpu by virtue of a fixed multiplier.

Except for one thing: you're implying that Intel and AMD are somehow ripping us off, which is incorrect. Speaking as someone who has taken a course in computer architecture, I can assure you that the same could technically be said of clock speed: It isn't technically a "characteristic of the CPU" either. Both the bus clock and the CPU core clock are merely pulsing signals fed to these parts, in order to drive the sequential logic circuits.

However, they ARE a "characteristic" in the sense that the rated FSB/clock speed of the chips is what AMD/Intel have certified them for through testing at their factories. Yes, they sometimes might be able to be set higher than what they were certified for, but that's merely an incidental thing, NOT the rule. The reason I hate overclocking either of them is because it voids the warranty, and for VERY good reasons: 1) It might cause the CPU to run hotter than otherwise, which can cause premature failure, and 2) it might cause the CPU to operate unreliably, and AMD/Intel would be WELL within their right to say that they are effectively absolved of any liability for computing failures, since YOU caused them. I would hate to see something like the Golden Gate Bridge designed on an AMD CPU, then fail because the computer user was negligent, and overclocked his CPU, causing incorrect calculations.

I'd be willing to bet the other half of the cache is just fine in most of those chips.

Again, that is quite possible, but what percentage? And even if it were true, you could never know for sure if enabling that extra cache would REALLY be okay, and AMD would be well within their right to say they assume no legal liabilities if things go haywire as a result.

[bcassell]

First mpteach:

I want to run a mobile barton 2600 @ 220 x 10.5 (2.32Ghz)

Obviously mobos have fsb limits. Is there any limit on the fsb of a cpu?

For isntance, if a mobo could handle high fsb speeds, could someone run a XP2600 @ 1155 x2 ?

As far as I know, you could run a cpu at 2200x1 or 220x10 or 22x100 and it wouldn't matter to the cpu. If your motherboard can support the fsb, then your cpu should be fine (as long as your lower the multiplier to maintain stable speeds).


Now zuperdee --

Except for one thing: you're implying that Intel and AMD are somehow ripping us off, which is incorrect.

I implied no such thing! AMD and Intel rate their cpus for a given multiplier/bus speed. Of course that is all they will support, it only makes sense. They're not ripping us off, they're applying good business practices. Yes, AMD could make their Barton 2500+ run on a 200mhz fsb instead of a 133, but that excludes the chip from being used in older chipsets, etc. My only point was that as far as the cpu is concerned, it doesn't care if it's running on 133mhz fsb or a 200mhz fsb, it only cares what it's clock speed is.

Speaking as someone who has taken a course in computer architecture,

Not to sound hostile, but you don't want to play the credentials game with me =).

I can assure you that the same could technically be said of clock speed: It isn't technically a "characteristic of the CPU" either. Both the bus clock and the CPU core clock are merely pulsing signals fed to these parts, in order to drive the sequential logic circuits.

Yes, we can get really anal and talk about the only "characteristics" of the cpu being the trace lengths and transistor timings, but that doesn't have a whole lot of practical use. My whole point was to explain that the cpu didn't care what FSB is was being run at. Forget about the definition of "characteristics of the cpu", this is what I was tyring to get across -- for a given clock speed, the fsb does not matter to the cpu.

Bryan

[isp]

I didn't get to read the whole thread, I have to go in a few minutes, but I'm pretty sure you wouldn't see a big enough difference from 2400+ to 2800+ barton...

If you do upgrade to a barton though, you have an excuse to get the rest of the parts for another box...you already have a 2400+

[mpteach]

The reason I hate overclocking either of them is because it voids the warranty, and for VERY good reasons: 1) It might cause the CPU to run hotter than otherwise, which can cause premature failure, and 2) it might cause the CPU to operate unreliably, and AMD/Intel would be WELL within their right to say that they are effectively absolved of any liability for computing failures, since YOU caused them. I would hate to see something like the Golden Gate Bridge designed on an AMD CPU, then fail because the computer user was negligent, and overclocked his CPU, causing incorrect calculations.

In that logic, if the Petronas towers collapsed then it would be the screwy engineers fault who changed his heatsink and undervolted his fans. If he hadnt undervolted his fans then panaflo would be liable for the hundred billion dollar claim..

I hate to say it but pc hardware manafactures and software makers are only liable for their machines and not any data or designs or email viruses that are created with thier products.

We dontl live in a 1950's status qou reality, we build machines how we want them. Sometimes we have to do heavy modifications and increase/decrease cooling. Overcloking is techically viable. There are system stability test that can be run. The Cray supercomputerf a decade ago were alll watercooled and designed to be run on what we consider and overclock.

[vortex222]

bus speed and cache size do not matter so much for folding, but folding is not the only thing that matters on my computer. the high bus speed on my computer would give me a free proformance boost without increasing noise. and would make the computer faster in everything else.

i did find however that my athlon xp2200 took a nosedive when it got put on a KM266 board. however plugging in a video card into its agp slot helped somwhat. i wouldnt know why though, i guess the memory controler became more efficant when that happend.

[zuperdee]

I didn't get to read the whole thread, I have to go in a few minutes, but I'm pretty sure you wouldn't see a big enough difference from 2400+ to 2800+ barton...

I came to the same conclusion. I've decided to try the 2700+ thoroughbred; I've read reports that it is at LEAST as good as a 2800+ barton, and sometimes almost as good as a 3000+ barton.

If you do upgrade to a barton though, you have an excuse to get the rest of the parts for another box...you already have a 2400+

Hehe--I'll probably end up selling the 2400+, cause I can't afford to build another computer right now.

[vortex222]

i would sell your mom the old chip youhave now, and buy a barton to overclock

[unregistered]

I would just wait and get the 3200 in 6 months. I think you will be dissappionted in the performance increase from a 2400 to a 2800. Hardly worth it. With the 3200 at least you will have extended your MB as far as it will go (unless AMD adds another chip to the Barton line).

Thats what I plan on doing with my 2200. I have a 2800 barton in my other box that is better and faster but it's not worth it for me to jump up just to a 2800. The 3200 will come down to what I am going to pay soon enough.

[haysdb]

This is the most "spirited" discussion we've had around here for awhile.

David

[zuperdee]

I would just wait and get the 3200 in 6 months. I think you will be dissappionted in the performance increase from a 2400 to a 2800. Hardly worth it. With the 3200 at least you will have extended your MB as far as it will go (unless AMD adds another chip to the Barton line).

Actually, I believe AMD will soon be adding a whole slew of new chips to the Athlon XP line--after all, isn't Socket 754 ultimately supposed to become AMD's new XP platform?

(As I understand it, AMD was planning to make Socket 754 AthlonXP's that are 32-bit, but still have the on-die memory controller.)

This is the most "spirited" discussion we've had around here for awhile.

Hey, I'm hoping it will soon yield results for the SPCR team, when I'm able to fold faster!