Dothan screams like scalded CAT!!!

[Flib]

when you use Ricky Tweak with P4 your Performace will go DOWN!
I think it's the same with the Dothan.

[STEvil]

anyone bothered trying jumping straight to 200 or 201fsb to see if something kicks in there?

[drunkenmaster]

Charlie, sorry m8, think you misunderstood a little, right now a granite bay type chipset wouldn't do squat for dothan. Its a case of p4 being designed for dual channel from the beginning and being starved of bandwidth till dual channel, while dothan is single channel and designed that way. Dual channel, just throwing up same kinda bandwidth a p4 can get now won't give that big a boost, and would probably start to cause the same issues we saw with ath xp's, with 133/166Mhz we saw the little extra memory speed compensate for mem controller not being fully efficient(100% is impossible, the extra 33Mhz just helped fill the gap) but when we had 133/200 performance went down, it was a case of memory running a touch to fast and it doesn't work as effectively.

Dothan will need a redesign to use dual channel type bandwidth at all effectively, but that will be very difficult as low power designs have to be as basic as possible. Most of the heat/wattage wasted comes from complex designs causing electron leakage.

[dippyskoodlez]

Sorry I can't get it connected to the internet, it's really a PITA to run a bench, grab a screenie and burn a disc.

Seriously....

http://www.crucial.com/store/listmfg...drives+storage

They pwn cd's.

I have a 128mb PNY drive. it is SO much easier than floppys or cd's.

[IvanAndreevich]

charlie
Try setting the lan from autonegotiation to different modes like 10 Half Duplex. See if that helps

[macci]

Charlie, sorry m8, think you misunderstood a little, right now a granite bay type chipset wouldn't do squat for dothan. Its a case of p4 being designed for dual channel from the beginning and being starved of bandwidth till dual channel, while dothan is single channel and designed that way.

Dothans need and love bandwidht just like the P4s did. Dothan has QDR bus (like P4 has) so it wont act similar to AXP's when dualchannel chipset is used. Just like Charlie mentioned runnign Dothan w/ i855 is pretty much like running P4 w/ i845PE. RAMBUS is the next step? (it is clock to clock faster than DualDDR at 180FSB)

but reconfiguring the cpu die to accept higher fsb, more bandwidth and so on isn't something that can happen instantly.

Since when has CPU die been limiting the FSB speed? fsb, bandwidth etc is pretty much chipset specific.

There are documented issues that stopped the release of speeds of dothans initially as they couldn't get certain fsb's working, i can't remember exactly but i think the dothan was supposed to be a 200Mhz fsb part, but it simply wouldn't work at that, in single channel, so was released at a lower fsb. SO there are issues there.

This 'issue' has nothing to do with the fact that if they had released Dothan w/ 200FSB DC mem it would have easily outperformed every desktop Intel system?
Can you post a link to these 'documented issues'?

Also the amount of engineers working on dothan compared to P4 projects are minimal.

Looks like the Dothan design team is superior to the P4 team
But I think that its fairly safe to assume that Intel uses quite a bit of resources for there Centrino technology and Dothan is (IMO) the most important part of it.

[IvanAndreevich]

macci
>>Since when has CPU die been limiting the FSB speed?
Athlon XP - very serious issue for some chips. My buddy's XP-M can run 200x14 semi-stable on air but is nowhere near stable @ 11x250, even though his DFI can EASILY do 250.

[charlie]

Seriously....

http://www.crucial.com/store/listmfg...drives+storage

They pwn cd's.

I have a 128mb PNY drive. it is SO much easier than floppys or cd's.

that's a GREAT idea, i even have one laying around
In addition, I picked up another NIC, so if I have probs with next board, I'll use NIC.

[Kunaak]

macci
>>Since when has CPU die been limiting the FSB speed?
Athlon XP - very serious issue for some chips. My buddy's XP-M can run 200x14 semi-stable on air but is nowhere near stable @ 11x250, even though his DFI can EASILY do 250.

250 has never been easy for AMD Xp's of any sort.
if you have a CPU. Motherboard and ram that can do 250 stable, and easy, then good for you. you were one of the few that found the magic combination of hardware that could do it stable and easy.

while I reached 269 on the DFI NF2, and could pass 3dmark at 255.
nothing above about 230 was 100% stable.

hell, I was just impressed it did 230 100% stable.
no other NF2 I ever used did near that, 100% stable.

[Perc]

@charlie

what the "ricky trick" ? ive been seeing that around but no one realy explains it...

thx perc,

[IvanAndreevich]

Kunaak
The point being it's HARD to do high FSB & high speeds at the same time. At least on Athlon XP, and maybe on Pentium-M's.

[Kunaak]

his arguement was that the core size, determined the FSB limit.
which I agree with macci on this... the core size has little to do with the FSB limit.

[IvanAndreevich]

Kunaak
As in surface area of the die?

[dippyskoodlez]

his arguement was that the core size, determined the FSB limit.
which I agree with macci on this... the core size has little to do with the FSB limit.

With the high IPC of the dothan, I dont think dual channel would help as much, kinda like A64's...Then again.. it does have a 4x fsb. :exclamati But, noone knows the effects until we see one in action

Even with 250fsb on an AXP we didnt see THAT much of an improvement. 1:1.

[STEvil]

Die size? I dont see a mention of die size, just that he thinks a CPU which doesnt actually contain a memory controller can limit FSB.

I can see the possibility since the CPU has to talk to the northbridge (or system bus depending on configuration(s)) at some point or another but overall its going to be the chipset..

[IvanAndreevich]

>>a CPU which doesnt actually contain a memory controller can limit FSB
Well, I think that this is certainly the case with Athlon XP's. Do you get what I mean from the post above?

[STEvil]

I have yet to experience it with any XP's or Bartons..

[NeGe0]

>>a CPU which doesnt actually contain a memory controller can limit FSB
Well, I think that this is certainly the case with Athlon XP's. Do you get what I mean from the post above?

It can limit FSB? Maybe I don't understand but I don't think the P4 has a memory controller and I don't see it limited....but that's just me, I could be wrong.

[IvanAndreevich]

NeGe0
STEvil
Ok let's look at a real situation..

1) CPU + board can easily run 250x9
2) CPU can run 200x14
3) NO WAY IN HELL to get it stable @ 250x11

Please comment.

[macci]

Isnt' that more of a 'NF2 feature'?
I haven't had much experience w/ NF2 boards but I do remember that it was / is? easier to hit high FSB w/ lower multiplier? At some point high Vcore was killing the FSB w/ those boards too right? So as far as I can tell its a platform problem. The chipset is designed for 200FSB..

Let's see here..
Celeron s370 a 66FSB chip => does 133+ FSB no problem
Pentium 3 s370 a 100FSB chip => does 200+ FSB no problem
Duron SocketA a 100FSB chip => does 200+ no problem
Pentium4 1.6 a 100FSB chip => does 200+ no problem
etc etc.
POint being that I have yet to see a chip that would limit the FSB. It has always been the platform that sets the limits.

With the high IPC of the dothan, I dont think dual channel would help as much, kinda like A64's...Then again.. it does have a 4x fsb.

A64s have the integrated low latency mem controller which is probably one of the reasons why dualchannel doesn't make much of a difference (cache is more important).
Dothans dont have similar mem controller and as far as I can tell the Dothan acts very similar to the P4s w/ single channel RAM and increasing the bandwidht really makes a difference.
Example:
2.4G w/ 162 RAM => 35sec SPi1M
2.2G w/ 196 RAM => 35sec Spi1M
2.4G w/ 214 RAM => 32sec Spi 1M

[dippyskoodlez]

I have yet to experience it with any XP's or Bartons..

i havent had a cpu that wouldnt do a fsb i knew was stable either.

Ive had around 15 AXP's that would all do any FSB I set at em. tbred, tbird, and bartons.

[dippyskoodlez]

Dothans dont have similar mem controller and as far as I can tell the Dothan acts very similar to the P4s w/ single channel RAM and increasing the bandwidht really makes a difference.
Example:
2.4G w/ 162 RAM => 35sec SPi1M
2.2G w/ 196 RAM => 35sec Spi1M
2.4G w/ 214 RAM => 32sec Spi 1M

Yes, dotans arent like an A64, but in comparing to a P4, the dothans are not bandwidth starved like a P4 is. P4 is long pipeline, needing a lot more bandwidth to perform like it should. Dothan however is based a lot more like an Athlon based core, and doesnt need nearly as much bandwidth to perform. As seen in the overclocked dothan benchies.

[Pejsen]

Yes, dotans arent like an A64, but in comparing to a P4, the dothans are not bandwidth starved like a P4 is. P4 is long pipeline, needing a lot more bandwidth to perform like it should. Dothan however is based a lot more like an Athlon based core, and doesnt need nearly as much bandwidth to perform. As seen in the overclocked dothan benchies.

I must admit that i haven't got as much experience as you guys probably do, but looking at the test results OC'ers here on XS and also Kyosen and the results on various test sites i think it is fair to say that the Dothan really loves bandwith, even though the pipeline is short.

[NT_Kid0]

I think a big problem with the argument that a long pipeline chip like p4, that needs the bandwaith to perform better vs a dothan which has shorter pipelines not benefiting as much from extra bandwith is that we dont truly know the limits of their designs to start with. Didnt someone on xs have an engineering sample of a prescott and manage to run the ram at 1:1 400fsb ? thats pretty impressive for a 200 fsb released chip.

[NT_Kid0]

please ignore the above ramblings if they dont make any sense