ASUSTeK Adapter Pentium M = > Pentium 4

Printable View

Show 100 post(s) from this thread on one page

03-06-2005, 11:22 AM
STEvil

Dothan are starved for bandwidth, thats why it came up and thats why everyone wants them on 478..
03-06-2005, 11:24 AM
uclajd

Quote:

Originally Posted by alpha0ne

i mean, lets be honest, what is the preshat good for apart from heating in winter :p:

Well, if you'd like to compare video encoding times, BRING IT ON! :stick:
03-06-2005, 11:54 AM
saratoga

Quote:

Originally Posted by Xassius

At the end of the day, it is bandwidth that counts. Whether it is high latency or low latency. It depends on the memory controller whether it likes high clocks or low latencies..etc..etc..etc

In current Intel desktops, DDR2 rules. If AMD adopted DDR2, no doubt you'll see 10Gb/sec Sandra buffered shots easily.

DDR2 is more suited to those with high FSB's (more than 300mhz FSB) on current desktops as for maximum performance, a 1:1 divider for DDR2 needs to be used. Since Dothans mostly stay below the 200mhz mark, we may not see a substantial gain over normal SO-DIMM DDR.

I have seen somewhere that CAS latencies on a laptop when reduced from 3 to 2 failed to procure any performance gains (was that LegitReviews.com ?)

DDR2 also produce a lot less heat - a more crucial point than performance in a laptop.

No latency is more significant then bandwidth in most applications. Very few programs really stream GBs worth of data. Virtually all have to load lots of small variables. In this case latency kills you. Particularly when loads take 100+ clock cycles (Dothan)!

One of the arguments in favor of dual core is that it decreases the impact of latency (which is hard as hell to lower) and shift the impact to bandwidth (which is easy to increase).

Quote:

bandwidth didn't double. Theorhetical bandwidth did "double" but not actual bandwidth.

What I define bandwidth is = how much data throughoutput you can access the memory at.

Actually it more or less did double. Theres some inefciency, but its small.

Quote:

I was not talking about theorhetical bandwidth...but actual bandwidth. So AMD processors gained 3% more bandwidth going to dual channel (for example)

Err do you not understand the difference between bandwidth and performance?
03-06-2005, 03:17 PM
cirthix

if you want to have a discussion about ddr2 vrs ddr, amd vrs intel, go somewhere else, leave it out of this thread.
03-06-2005, 05:02 PM
Xassius

Quote:

Originally Posted by STEvil

we've had 10gb/sec+ using DDR1.

Why?

Because latency decreases due to speed of the CPU (memory controller built in) allows it to access data faster.

Between 754 and 939/940 bandwidth doubled, but useable bandwidth didnt change much because of the efficiency of AMD's setup. DDR2 might bring more increases, but its all going to be latency which is going to take some work.. like DDR2 800 @ 3-2-2-x for a starting point.

10Gb/sec. Yep OPB. That's not everyone. I said that 9Gb/sec+ scores will be achieved fairly easily.
03-06-2005, 05:09 PM
Xassius

Quote:

Originally Posted by saratoga

No latency is more significant then bandwidth in most applications. Very few programs really stream GBs worth of data. Virtually all have to load lots of small variables. In this case latency kills you. Particularly when loads take 100+ clock cycles (Dothan)!

One of the arguments in favor of dual core is that it decreases the impact of latency (which is hard as hell to lower) and shift the impact to bandwidth (which is easy to increase).

Actually it more or less did double. Theres some inefciency, but its small.

Err do you not understand the difference between bandwidth and performance?

Why do you decrease latency? To increase bandwidth.

memory with latencies of 5-5-5-12 at 1000mhz will transfer small packets of data faster than 2-2-2-5 at 200mhz.
03-06-2005, 05:16 PM
saratoga

Quote:

Originally Posted by Xassius

Why do you decrease latency? To increase bandwidth.

No. THe two are unrelated. Increaseing or decreaseing one does not change the other.

You're thinking of throughput which is something very different, and is specific to each and every piece of code (there is no throughput for a processor, just for a given piece of code on a given processor). I suggest you check google, or a textbook. THis sort of thing is covered in the intro to many networking and engineering texts. I won't argue this any further because its not relevent to the thread.
03-06-2005, 07:57 PM
iddqd

Quote:

Originally Posted by Xassius

Why do you decrease latency? To increase bandwidth.

memory with latencies of 5-5-5-12 at 1000mhz will transfer small packets of data faster than 2-2-2-5 at 200mhz.

By the same token, I could argue that I scale the clock to decrease the latency. Why, at 275MHz 2-2-2, the latency is much lower than at 200MHz 2-2-2.

Oh, sorry. I meant to say Pentium-M. That's riiight. Pentium-M. :banana:
03-06-2005, 10:38 PM
ZL1

and I was just reading the "p4 4ghz cancelled thread" :) (alot of p-m in there)
I hope (like everyone else) that ocing will be possible

guys lets keep this thread alive (and flameless), if anyone gets a sample, please oh please post

D

PS by the way how come all p-ms at newegg show s478 ?
03-06-2005, 11:48 PM
Xassius

Quote:

Originally Posted by iddqd

By the same token, I could argue that I scale the clock to decrease the latency. Why, at 275MHz 2-2-2, the latency is much lower than at 200MHz 2-2-2.

Oh, sorry. I meant to say Pentium-M. That's riiight. Pentium-M. :banana:

agreed.

Oh! sorry. I was defining bandwidth as throughput :p: Just like in sandra... I always referred to the scores as "bandwidth". Will have to learn my vocab.

I don't think the Dothan is too starved for memory "throughput" in my opinion. It's got the 2mb Cache to help it's pipeline along.

Also - it's sad that we won't see any adaptors for LGA775 :(
03-07-2005, 01:55 AM
Foxie3a

Quote:

Originally Posted by wimpie007

Have you ever heard of that little company, called AMD?
They happen to do that with Cool'n'Quiet :)

Sorry, I didn't mean that I wished they had speedstep in the motherboards, I meant to say that I wish that they were able to disable it if you would put a mobile processor in a desktop board, instead of the multiplier going down to 12.
03-07-2005, 02:17 AM
STEvil

12x200 is 2400mhz, and most 478 boards can scale well past that pretty easily.
03-07-2005, 06:33 AM
Revv23

yeah but how much FSB can the dothan handle?
03-07-2005, 07:00 AM
alexio

I think not much more than just above 200mhz, 210-220mhz in the best case I think.

But we'll see, I hope I'm wrong.
03-07-2005, 07:51 AM
Revv23

i guess no one has ever had it in a good chipset to find out eh?
03-07-2005, 11:06 AM
Shade00

Personally, I think the Dothan has been chipset limited. The i855 is more or less a rehash of the i845, which had trouble making it near 200 even in its PE incarnation. Sure, some would do it, but not all.
03-07-2005, 11:37 AM
alexio

I saw a review were a 133mhz Dothan hit 190mhz fsb on an i855 chipset while the 100mhz Dothan only did 160mhz or so.

Does the i855 has some sort of switch for both 100mhz and 133mhz, if so than they were both limited by the board failing to lockt the agp/pci frequenty.

In that case we'll just have to waith untill the adapter is for sale.
03-07-2005, 12:41 PM
Revv23

Quote:

Originally Posted by ZL1

PS by the way how come all p-ms at newegg show s478 ?

there is only 478 pins on dothan, but the pinout is different, hence it being incompatible with desktop mobo's, so they call it socket 479 in hopes of avioding confusion
03-07-2005, 12:49 PM
ZL1

Quote:

Originally Posted by Revv23

there is only 478 pins on dothan, but the pinout is different, hence it being incompatible with desktop mobo's, so they call it socket 479 in hopes of avioding confusion

oh, cool, thanks !

Dan
03-07-2005, 02:25 PM
iddqd

Quote:

Originally Posted by Revv23

there is only 478 pins on dothan, but the pinout is different, hence it being incompatible with desktop mobo's, so they call it socket 479 in hopes of avioding confusion

They could have called it stocket M. Or something.
03-07-2005, 04:32 PM
Tony

I have a 100fsb 1.6 that does 193fsb quite easy on 855, I have a feeling it will do well over 200 on an 875 board if the socket adapter allows for high fsb's.

its going to be a new ball game with dothan +ddr1 dual channel, a whole new breed of cool quiet gaming rigs will be born ;)
03-07-2005, 04:47 PM
Revv23

asus could revitalize DDR, socket 478, and intel's enthusiast populatirty all in one fell swoop if a device like this performes well.
03-07-2005, 04:49 PM
ZL1

Quote:

Originally Posted by Revv23

asus could revitalize DDR, socket 478, and intel's enthusiast populatirty all in one fell swoop if a device like this performes well.

:toast:
03-07-2005, 05:29 PM
Franky 4 Finger

I hope this works out. As much as I love my winnie I am sick and tired of the AMD fanboys.
03-07-2005, 06:33 PM
Northwood

personally i'm looking forward to "Yonah".
Theres too many problems trying to run Dothan in a desktop, just not enough bandwidth.

hopefully Yonah will be 800FSB and DDR1 dual channel, or even 1066FSB if poss :)

another chip i'm looking forward to is Intels "Pressler" can't remember the specs but hopefully it will kick much arse :D

Show 100 post(s) from this thread on one page