Thanks for pointing this out, guys... does anyone know if it matters if I get a cheap PCI Express card or a cheaper AGP card? I'm not planning on doing any serious gaming on the computer, but I want to know if it matters for other things. Forgive my ignorance, some of my favorite computers don't have any graphics cards... or monitors, for that matter!

Since gaming isn't your thing, and if you have a slot open, a PCI card should do nicely.


kev

---

kev

X2 4400+,4200+ @2.75GHz, XP1800+ @1.65GHz, P4 @1.6GHz

"Integrated graphics" that use the CPU RAM effectively steal a large share of memory bandwidth away from the CPU. For most normal usage, this is crippling for CPU performance in that the CPU is left waiting idle while starved of memory accesses.

Thanks for pointing this out, guys... does anyone know if it matters if I get a cheap PCI Express card or a cheaper AGP card? I'm not planning on doing any serious gaming on the computer, but I want to know if it matters for other things. Forgive my ignorance, some of my favorite computers don't have any graphics cards... or monitors, for that matter!

---

IMHO Intel chips have always kicked butt over AMD for serious applications.

Except for the last 3-5 years.
Intel has been lagging in pretty much every application other than Seti crunching & video editing. And for the last 12 months or so AMD has actually had the lead in video editing performance, inspite of Intel's HyperThreading & 1GHz+ clock advantage.
And when it comes to web serving & database performance, Intel hasn't even been in the race.

...my own P4 Prescott 2.8 oc to 3.0Ghz vs my own AMD 64 X2 4200. The older P4 kicks butt on the 64 bit AMD in almost every arena that I have been able to measure.

Then there is something seriously wrong with your AMD system.

Comparison 1
Comparison 2
Comparison 3
Comparison 4
Comparison 5

In every benchmark, the AMD CPU is faster- even those where clock speed is an important factor the 800HMz slower AMD CPU beats the Intel one.
In other less clock speed dependant benchmarks, and where the software can make use of mutliple threads, the AMD CPU beats the Intel CPU to a plup.



Edit- make links clickable.

---

Grant
Darwin NT

IMHO Intel chips have always kicked butt over AMD for serious applications.
If you are a kid and just want to play as cheaply as possible, the I guess it's ok to go the AMD route. Case in point is my own P4 Prescott 2.8 oc to 3.0Ghz vs my own AMD 64 X2 4200. The older P4 kicks butt on the 64 bit AMD in almost every arena that I have been able to measure. The AMD is quieter, cooler, and of course more energy effecient. But you know, a VW is more energy effecient than a Ferrari.

Wow, did YOU fall for Intel's marketing! Pull your head out of the sand and look around. You've been missing things for the last 5 years.

---

Without love, breath is just a clock ... ticking.
Equilibrium

INTEL "CONROE" blows away AMD's finest:

http://www.hardwareanalysis.com/content/article/1830/

http://www.hothardware.com/viewarticle.aspx?articleid=794&cid=1

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2713&p=2

:)

* That's something to consider on several levels, not only for PC-gaming performance (& other areas)... but, also for investments purposes, IF you "play that game too"...

(IMO? Intel's going to be one HECK of a stock buy very soon (soon as in NOW or shortly because of this))

IMO @ least, this is a signal of it/for it (time is now to buy Intel stock), & now is the time to get a jump on that...

AMD's about to split from what I understand (stock split), but Intel's about to rock their world.

Yes, the 'conroe' chips ARE real, as well as those results I put up (but are between 9-12 months out as a release to mass production/consumer good markets most likely & probably will be even BETTER/FASTER @ that time!)

Something to keep in mind not only if you're a PC-enthusiast, but also if you are an investor...

APK

P.S.=> "Let the CHIPS (cpu's) fall where they may...", lol! apk

You seem to believe that AMD will be sitting still for the next year. Sure, everything points to the fact that Intel will be in the lead for a good 6-12 months after Conroe comes out. But AMD is also working on new chips (AM2 isn't going to compete with Conroe right out of the box, unless they're hiding something big).

Yes, I like AMD processors. Granted, I'm not foolish enough to think that AMD will certainly have something to compete with Conroe right away (although there is a very slim possibility of this), but I'm also not foolish enough to believe that AMD will be losing everything because of Conroe. I'm sure AMD has some things in development for the next 18 months.

---

I think AMD's major problem is that they saw the current memory bus limitations coming and chose to go around it for now and let Intel fight the chipset and memory speed technology for them first, by first optimizing around the 200MHz PC3200 bus by putting the memory controller in the processor chip, going to a serial HT bus and placing their bets on the 64 bit programs giving them an early advantage in the server world and business world. And that hightech games would jump on the 64 bit bandwagon.

They can then add the faster FSB when the technology settles out, without having to spend the effort to drive and develop it.

Well, 64 bit mainstream programs have been slow to adapt to the advantages of the 64 bit world. And 64 bit programs with 128 bit memory access (dual channel) does not fit all all applications.

For Setiathome, 64 bit instructions and things like the dual chanel 128 bit wide memory bus are not much of an advantage. The Setiathome application does not run much faster on a dual channel bus than it does on a single channel AMD system. Other programs and applications might. And Setiathome has the same limitations on Intel processors in the long run.

The problem with Setiathome is the data it is processing and the way it is done. The "chunk" size of data accessed by setiathome during processing has to be either 32 bits for Standard precision variables or 64 bits (eight bytes) at a time for double precision variables. For some processing sequences where the data is read in strict sequence, 128 bit access or DDR2 type of memory accesses provide for a speed up in memory access of the data. But when setiathome starts to look at drift and other patterns in the data (and Enhanced will do more of this), as far as the computer hardware being able to access memory, each chunk of data comes from a random location and is read in a small chunk. So a wide bus (128 bits) or fast sequential data access has little advantage if the program code is compact enough that functional sections of it fits in the L1/L2 cache.

When AMD Athlon 64 processors are clocked above 2200MHz, there is little additional gain for Setiathome because it becomes bound by the slow memory access of the 200MHz bus (overclocking of the memory of course helps). Same limit for dual processors.

Because of this small chunk size data access by setiathome, the only way to decrease crunch times is to have faster access to the data. A larger L1 cache (like AMD has) helps some, but a larger L2 cache is required as well as faster memory bus (not wider). This is where Intel currently has an advantage, as they have been forging down the faster FSB path with their highend products, and have already started using the 2MB L2 caches. But there is also a physical limit on how fast memory can be if it is external to the processor. And there is a limit on how big and fast the L2 cache can be. The larger it is, the slower it will have to be clocked to keep from over heating.

I expect that now that AMD has its 128 bit dual channel bus in place and its HyperTransport bus in place, its next step up will be larger L2 caches and then increasing the memory FSB speed to match Intel's faster processors. The larger L2 caches would be the quickest "fix" for them in the race with Intel. That is just a matter of fab technology and production. The faster memory FSB will be more difficult as it requires new support chips and system boards, but Intel has already lead the way there and it is just a matter of now having the chipset vendors adapt the best of the faster FSB technology to AMD's bus. But given time, they can match anything Intel does, and have an advantage with the integrated memory controller and HT I/O bus for now.

IMHO Intel chips have always kicked butt over AMD for serious applications.
If you are a kid and just want to play as cheaply as possible, the I guess it's ok to go the AMD route.

Although I am the first to point out that Intel chips are superior as S@h crunching workstations due to their recent adoption of enormous CPU caches, AMD's HyperTransport and really fast bus speeds make Opterons better to run as serious servers than anything Intel has or will have anytime soon.

And servers are definitely for grown-ups! The basic difference right now and for the next 3-5 years is that AMD moves data better, and Intel crunches data better. They each have their lucrative niches (enterprise servers for Opteron, high-end workstations for Xeon).

---

IMHO Intel chips have always kicked butt over AMD for serious applications.
If you are a kid and just want to play as cheaply as possible, the I guess it's ok to go the AMD route. Case in point is my own P4 Prescott 2.8 oc to 3.0Ghz vs my own AMD 64 X2 4200. The older P4 kicks butt on the 64 bit AMD in almost every arena that I have been able to measure. The AMD is quieter, cooler, and of course more energy effecient. But you know, a VW is more energy effecient than a Ferrari.

Reference Pentium M's, my main machine is a 1.86GHz and completes units in 33 mins average.

For clarity there are two versions of Pentium M's the older (non-Dothan) have 1MByte L2 cache and are slower than the current Dothan cpu's with 2 MByte L2 cache.
I suspect 1.8Ghz refered to by kev1701e is older version.

They are ideal for crunching Seti where cache is King. But to use them as Desktop is initially expensive, the best mobo is the AOpen i915GMm-HFS but it is expensive. Although there are some older AOpen and MSI mobo's using 855 chipsets that are slightly cheaper. And there is a ABit daughter card for use on some of their mainstream top end mobo's.

The main benefit is that even fitted with a 6600GT graphics card it uses less than 90 watts (measured a wall socket) when crunching BOINC projects. It is not very good at crunching Einstein and on CPDN does 3sec/TS.

For those that own them, did you know that you can reduce the cpu core voltage to 1.1 volts from nominal 1.35V to 1.5V and reduce heat further.

From the performance of my Pentium M, I am prepared to say that all the imdications in review linked by Grant are probably true and AMD is probably going to get there butts kicked again.

---

When we finally figure it all out, all the rules will change and we can start all over again.

Reference Pentium M's, my main machine is a 1.86GHz and completes units in 33 mins average.

For clarity there are two versions of Pentium M's the older (non-Dothan) have 1MByte L2 cache and are slower than the current Dothan cpu's with 2 MByte L2 cache.
I suspect 1.8Ghz refered to by kev1701e is older version.

They are ideal for crunching Seti where cache is King. But to use them as Desktop is initially expensive, the best mobo is the AOpen i915GMm-HFS but it is expensive. Although there are some older AOpen and MSI mobo's using 855 chipsets that are slightly cheaper. And there is a ABit daughter card for use on some of their mainstream top end mobo's.

The main benefit is that even fitted with a 6600GT graphics card it uses less than 90 watts (measured a wall socket) when crunching BOINC projects. It is not very good at crunching Einstein and on CPDN does 3sec/TS.

For those that own them, did you know that you can reduce the cpu core voltage to 1.1 volts from nominal 1.35V to 1.5V and reduce heat further.

From the performance of my Pentium M, I am prepared to say that all the imdications in review linked by Grant are probably true and AMD is probably going to get there butts kicked again.

---

See here for a note on another thread comparing caches for s@h-enhanced.

[edit] Also consider carefully Tetsuji's good comments here. It's all a very detailed balancing act for what might be wanted, when, and how... ;-) [/edit]

Happy crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

...i hope you didn't mix up the Pentium M and the Pentium 4-M. These have different cores and are not related. Pentium M is the grandson of Pentium III and Pentium 4-M is some kind of low voltage Pentium 4 (desktop).

The P-M is indeed a very nice design, especially so for the low power 2MByte L2 cache.

However, I consider the P-4 dash for deep pipelines and high GHz a very silly distraction just for the sake of Marketing leverage for the sake of simple meaningless numbers... ("My number is bigger than your number" silliness...!) You could argue that AMD gained its lead from that dead end...

Regards,
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

...
Here's mine compared to an apparently stock Pentium M 1.8 (could be a laptop or not)
WU.

On close inspection of the actual run times, the P4M completed the WU in 4477 seconds, while my OC'd 3000+ did it in 3117 seconds - 22 minutes, 40 seconds faster.
...

kev,
i hope you didn't mix up the Pentium M and the Pentium 4-M. These have different cores and are not related. Pentium M is the grandson of Pentium III and Pentium 4-M is some kind of low voltage Pentium 4 (desktop). A not overclocked Pentium M (2Mb L2) with 1.86 GHz can do a wu with optimized setiapp in 2100 secs., maybe less.

They're both mobile cpus, but you are right - the one I compared to was a 4-M, my bad :) However, I'm going to stick with the idea that the P-M 1.6 Hypnotoad was looking at is overclocked because it's doing WUs in the ~1700-1800 second range.


kev

[edit] typos

---

kev

X2 4400+,4200+ @2.75GHz, XP1800+ @1.65GHz, P4 @1.6GHz

... I saw no graphics card in your specs. Huge drain on your resources and cycles if that's the case. Get even a cheap card and you'll speed up a LOT.

Is an on-board GPU really a drain on CPU cycles? SETI@home never uses more than 35 MB of memory, so I certainly don't need to free up more RAM. I could definitely get a cheap graphics card, but didn't think I needed one since I'm not a gamer (you blew my cover there, didn't you!).

Memory contention is the big draw on system resources. The video chip is using the same FSB as the CPUs. Having a seperate card will transfer some of that memory access to the card leaving more of the memory bandwidth for the CPU.

I very much agree.

"Integrated graphics" that use the CPU RAM effectively steal a large share of memory bandwidth away from the CPU. For most normal usage, this is crippling for CPU performance in that the CPU is left waiting idle while starved of memory accesses.

Confusingly, the s@h benchmarks are unaffected... (They run from the CPU cache.)

From what I've seen of recent systems, AMD look to have the much better set of bandwidth ratios going from the CPU core through the caches and then to system RAM. This means that you have good flexibility to choose your overall system performance by what speed RAM you use. (Cheaper less speedy RAM has less of an adverse effect than for such as a P4.)

Cool stuff!

Happy crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

It seems your mb has on-board video, correct? If so the "missing" RAM is used by Win/mb to run graphics. I saw no graphics card in your specs. Huge drain on your resources and cycles if that's the case. Get even a cheap card and you'll speed up a LOT.

Is an on-board GPU really a drain on CPU cycles? SETI@home never uses more than 35 MB of memory, so I certainly don't need to free up more RAM. I could definitely get a cheap graphics card, but didn't think I needed one since I'm not a gamer (you blew my cover there, didn't you!).

Memory contention is the big draw on system resources. The video chip is using the same FSB as the CPUs. Having a seperate card will transfer some of that memory access to the card leaving more of the memory bandwidth for the CPU.

---

BOINC WIKI

BOINCing since 2002/12/8

It seems your mb has on-board video, correct? If so the "missing" RAM is used by Win/mb to run graphics. I saw no graphics card in your specs. Huge drain on your resources and cycles if that's the case. Get even a cheap card and you'll speed up a LOT.

Is an on-board GPU really a drain on CPU cycles? SETI@home never uses more than 35 MB of memory, so I certainly don't need to free up more RAM. I could definitely get a cheap graphics card, but didn't think I needed one since I'm not a gamer (you blew my cover there, didn't you!).

---

your memory configuration appears to be odd (896 MB). What are your system specs?

I looked up my specs and I was mistaken on my own cache. I also have only 512kb, not 1 MB! So your 3000+ overclocked to 2.3GHz should be faster than my 2.2GHz by 5% or so. I don't know why the memory is reported down, however.

- processor: AMD Athlon 64 3500+, 2.2GHz, Venice S939
- cache: 128kb primary, 512kb secondary
- memory: Slots A0,A1 have 512MB each; A2,A3 are empty
- moboard: MSI RS480R2 (MS-7093)
- chipset: ATI Radeon Express 200
- BIOS: Phoenix LTD 6.00 PG 04/29/2005
- disk: 200.3GB useable storage
- CD/DVD: TSSTcorp CD/DVDW TS-H552B CDROM Drive
- network: Realtek RTL8139/810x Fast Ethernet NIC
- software: Windows XP Media Center, SP-2 (build 2600)
- TV tuner card and IR remote.

It seems your mb has on-board video, correct? If so the "missing" RAM is used by Win/mb to run graphics. I saw no graphics card in your specs. Huge drain on your resources and cycles if that's the case. Get even a cheap card and you'll speed up a LOT. Any room in your case? I assume so even in a small form factor setup.

Are you using the correct slots for dual channel? I would think that A0&A2 or A1&A3 would be the slots to use. I may be off for your board but all of mine require a "1st and 3rd" population for dual channel. Alternate configs are likely as well. Triple check your manual (they're sometimes good for something after all).

If both of these things are issues and you can correct them both, you should see some noteworthy improvements.

Let us know -

Jim

---

Without love, breath is just a clock ... ticking.
Equilibrium

...
Here's mine compared to an apparently stock Pentium M 1.8 (could be a laptop or not)
WU.

On close inspection of the actual run times, the P4M completed the WU in 4477 seconds, while my OC'd 3000+ did it in 3117 seconds - 22 minutes, 40 seconds faster.
...

kev,
i hope you didn't mix up the Pentium M and the Pentium 4-M. These have different cores and are not related. Pentium M is the grandson of Pentium III and Pentium 4-M is some kind of low voltage Pentium 4 (desktop). A not overclocked Pentium M (2Mb L2) with 1.86 GHz can do a wu with optimized setiapp in 2100 secs., maybe less.

---

_\|/_
U r s

your memory configuration appears to be odd (896 MB). What are your system specs?

I looked up my specs and I was mistaken on my own cache. I also have only 512kb, not 1 MB! So your 3000+ overclocked to 2.3GHz should be faster than my 2.2GHz by 5% or so. I don't know why the memory is reported down, however.

- processor: AMD Athlon 64 3500+, 2.2GHz, Venice S939
- cache: 128kb primary, 512kb secondary
- memory: Slots A0,A1 have 512MB each; A2,A3 are empty
- moboard: MSI RS480R2 (MS-7093)
- chipset: ATI Radeon Express 200
- BIOS: Phoenix LTD 6.00 PG 04/29/2005
- disk: 200.3GB useable storage
- CD/DVD: TSSTcorp CD/DVDW TS-H552B CDROM Drive
- network: Realtek RTL8139/810x Fast Ethernet NIC
- software: Windows XP Media Center, SP-2 (build 2600)
- TV tuner card and IR remote.

---

I just want to point out that, these future products aside, Intel is already in the lead when it comes to SETI@home... just for using larger caches. AMD is clearly in the lead when it comes to gaming and server performance, but relatively cheap 1600 MHz Pentium M laptops outperform my 3500+ Athlon 64 (Venice, SSE3, 1mb cache) desktop on SETI@home simply because they have a 2mb cache.

Cache size doesn't much matter with gaming machines or for servers. But while gamers are better off with Athlon 64 (at least until these new Intel parts come out), crunchers are better off with Pentium D (or even Pentium M laptops!). It's all about the cache size, really.

Hypnotoad - my first impression is that the Pentium M you referenced is probably overclocked and is not a laptop. My second impression is that your 3500+ is a tad slow.

My Winchester core 3000+ (512k cache), clocked to 2.3GHz should be on par or slightly worse at crunching than yours, yet it seems to be a bit faster.

Here's mine compared to an apparently stock Pentium M 1.8 (could be a laptop or not)
WU.

On close inspection of the actual run times, the P4M completed the WU in 4477 seconds, while my OC'd 3000+ did it in 3117 seconds - 22 minutes, 40 seconds faster. At the default 1.8GHz clock, my 3000+ would have been 10-15 minutes slower, but still faster than the equivalently clocked P4M.

Also, your memory configuration appears to be odd (896 MB). What are your system specs?


kev

---

kev

X2 4400+,4200+ @2.75GHz, XP1800+ @1.65GHz, P4 @1.6GHz

INTEL "CONROE" blows away AMD's finest:

http://www.hardwareanalysis.com/content/article/1830/

http://www.hothardware.com/viewarticle.aspx?articleid=794&cid=1

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2713&p=2

:)

* That's something to consider on several levels, not only for PC-gaming performance (& other areas)... but, also for investments purposes, IF you "play that game too"...

(IMO? Intel's going to be one HECK of a stock buy very soon (soon as in NOW or shortly because of this))

IMO @ least, this is a signal of it/for it (time is now to buy Intel stock), & now is the time to get a jump on that...

AMD's about to split from what I understand (stock split), but Intel's about to rock their world.

Yes, the 'conroe' chips ARE real, as well as those results I put up (but are between 9-12 months out as a release to mass production/consumer good markets most likely & probably will be even BETTER/FASTER @ that time!)

Something to keep in mind not only if you're a PC-enthusiast, but also if you are an investor...

APK

P.S.=> "Let the CHIPS (cpu's) fall where they may...", lol! apk

---

http://torry.net/authorsmore.php?id=1781

"The object's hull is made of SOLID neutronium: A single StarShip cannot combat it!" quote Mr. Spock, Star Trek original series, episode title: "The Doomsday Machine"