I would like to add my thanks to Senator as well. I have both a Celeron2.4 and AMD3000 with the same clock and the AMD is twice as fast. I'm really impressed with the knowledge that people on this project (like Senator) have! It's great to get answers like this when you need them. Cheers.
Stealing your words, aside from minor edits, is my sincerest form of being lazy ...
It was a very good explanation ... check it out in my FAQ in the performance section.
B52,
You may want to also take a quick look at the Glossary to see what I have there under cache, and logical and physical processors. I also have an old lecture set on processors through the ages ... a bit dated, and I don't have a "good running"
list of individual processors of recent history ... Though Tom's Hardware page does ...
It's more of a "right tool for the job" issue. The P4s are very good at streaming data (e.g. encoding and decoding media files) and other tasks where sheer bandwidth and high clockspeeds rule, and Hypertheading does a good job over covering for the weaknesses of the design in most common cases (E@H benefits somewhat). If you're running most scientific and engineering applications, AMD is the way to go, but for something like video editting P4s are probably a better choice.
> even my sempron 2200 is faster than a standard P4
Nothing "only" about a Semperon, it's the same core as the regular Athlon XP and E@H seems almost as happy with 256KB of L2 Cache as 512K with the Barton core.
> > AMD has surpassed the intel guys in many ways..
>
> It's more of a "right tool for the job" issue. The P4s
> are very good at streaming data (e.g. encoding and decoding media files) and
> other tasks where sheer bandwidth and high clockspeeds rule, and Hypertheading
> does a good job over covering for the weaknesses of the design in most common
> cases (E@H benefits somewhat). If you're running most scientific and
> engineering applications, AMD is the way to go, but for something like video
> editting P4s are probably a better choice.
Yes, but here is the question ...
For total throughput, are 2 AMD "Whatever" going to beat a pair of Xeon from the same price point.
I mean, lets say you budget $500 per CPU, get a dual system with all other things being equal ... would the Xeon beat the AMD in throughput because you effectively have a 3 CPU system when you get done for the day...
THAT is *MY* question. The problem is that the only way I know to find out is to buy them both and test fly them. When I was working I could do that ... and would not lose sleep over it. But now I am on a budget and can't buy a new computer a quarter (yes, go ahead be annoyed with me, but I was for a time doing a complete new computer or a major upgrade to one in the farm every 3 months).
Has anyone done this?
I know when I look at the numbers the Xeons seem to do work pretty fast ...
> For total throughput, are 2 AMD "Whatever" going to beat a pair of Xeon from
> the same price point.
>
> I mean, lets say you budget $500 per CPU, get a dual system with all other
> things being equal ... would the Xeon beat the AMD in throughput because you
> effectively have a 3 CPU system when you get done for the day...
Well, if you're talking about a single dual processor system the AMD is even further ahead since the Opertons (the server multiprocessor version of the Athlon64) each have their own build in dual channel DDR (The Xeons share a common bus) and Hypertransport interconnect which allows them to share bandwidth (Non-Uniform Memory Access). Not comparing the most recent hardware and it's as a server rather than a cruncher, but it gives you an idea of the comparison: http://www.aceshardware.com/read.jsp?id=60000275
> THAT is *MY* question. The problem is that the only way I know to find out is
> to buy them both and test fly them. When I was working I could do that ...
> and would not lose sleep over it. But now I am on a budget and can't buy a
> new computer a quarter (yes, go ahead be annoyed with me, but I was for a time
> doing a complete new computer or a major upgrade to one in the farm every 3
> months).
>
> Has anyone done this?
>
> I know when I look at the numbers the Xeons seem to do work pretty fast ...
Well, the speed for 2 Xeons will be less than 2x the speed of one, due to the shared bus and message passing for cache coherency, the speed for Dual Opertons should be very close to 2x the speed of a single Athlon64/Opteron. Of course then the OS adds to the confusion (32-bit or 64-bit, NUMA aware or not, Windows or Linux)
> Well, if you're talking about a single dual processor system the AMD is
> even further ahead since the Opertons (the server multiprocessor version of
> the Athlon64) each have their own build in dual channel DDR (The Xeons share a
> common bus) and Hypertransport interconnect which allows them to share
> bandwidth (Non-Uniform Memory Access).
And memory access is a prime factor in speed. I bought two Fry's specials which bundle a CPU with MB for loss leader. I could only afford one, but I bought one high end MB. So, at the end of the day I had two Intel P4 4.2 GHz processors installed and running, one on a high end MB with dual channel memory access the other without. The one did SETI@Home in roughly 3 hours and change (10K Seconds) and the other roughly 4 hours (15K seconds).
Earlier this week I upgraded the slow machine with a new A-Bit IC7-G MB and now that machine is running about as fast as the other one ...
> Well, the speed for 2 Xeons will be less than 2x the speed of one, due to
> the shared bus and message passing for cache coherency, the speed for Dual
> Opertons should be very close to 2x the speed of a single Athlon64/Opteron.
> Of course then the OS adds to the confusion (32-bit or 64-bit, NUMA aware or
> not, Windows or Linux)
Yeah, I know. The point is that though we don't see the full benefit of 2 CPUs in almost all systems is from the memory access issues and cache coherency and the like. But with the Opteron/Athlon I would get 2 CPUs with the ability to run 2 models and with the Xeon I would have 4 in-flight. So, though the AMD set-up processes faster, I would, in theory, get higher throughput on the Xeon.
> that brings us to the question :
>
> should i go for an upgrade to dual core or dual system... meaning another
> setup with a plain athlon, cheapo style..
>
> In the end I think I might just add another system to get more work done..
I just replaced a cheap MB that caused all the models to run slow. I knew it would likely do that as it was a "loss-leader" bundle at Frys. But I could not get the CPUs cheaper naked. So, I got one good MB, two cheapo (wife bought one, one per customer) bundles and sure enough the good MB did SETI@Home in about 3 hours something, the cheap one was more like 4 hours something.
So, that should help my processing ... I will be getting more help when I buy new systems later this year ... Though I am still unsure which is the best way to go, though it still looks to me like the Xeon would still produce the most work, and with an Intel MB be the most stable.
I bought a dual AMD a couple years ago and had nothing but trouble with it ... :(
I would like to add my thanks
)
I would like to add my thanks to Senator as well. I have both a Celeron2.4 and AMD3000 with the same clock and the AMD is twice as fast. I'm really impressed with the knowledge that people on this project (like Senator) have! It's great to get answers like this when you need them. Cheers.
Senator, Stealing your
)
Senator,
Stealing your words, aside from minor edits, is my sincerest form of being lazy ...
It was a very good explanation ... check it out in my FAQ in the performance section.
B52,
You may want to also take a quick look at the Glossary to see what I have there under cache, and logical and physical processors. I also have an old lecture set on processors through the ages ... a bit dated, and I don't have a "good running"
list of individual processors of recent history ... Though Tom's Hardware page does ...
AMD has surpassed the intel
)
AMD has surpassed the intel guys in many ways..
even my sempron 2200 is faster than a standard P4
senator seems to have hit it right on tha head there..
> AMD has surpassed the intel
)
> AMD has surpassed the intel guys in many ways..
It's more of a "right tool for the job" issue. The P4s are very good at streaming data (e.g. encoding and decoding media files) and other tasks where sheer bandwidth and high clockspeeds rule, and Hypertheading does a good job over covering for the weaknesses of the design in most common cases (E@H benefits somewhat). If you're running most scientific and engineering applications, AMD is the way to go, but for something like video editting P4s are probably a better choice.
> even my sempron 2200 is faster than a standard P4
Nothing "only" about a Semperon, it's the same core as the regular Athlon XP and E@H seems almost as happy with 256KB of L2 Cache as 512K with the Barton core.
> > AMD has surpassed the
)
> > AMD has surpassed the intel guys in many ways..
>
> It's more of a "right tool for the job" issue. The P4s
> are very good at streaming data (e.g. encoding and decoding media files) and
> other tasks where sheer bandwidth and high clockspeeds rule, and Hypertheading
> does a good job over covering for the weaknesses of the design in most common
> cases (E@H benefits somewhat). If you're running most scientific and
> engineering applications, AMD is the way to go, but for something like video
> editting P4s are probably a better choice.
Yes, but here is the question ...
For total throughput, are 2 AMD "Whatever" going to beat a pair of Xeon from the same price point.
I mean, lets say you budget $500 per CPU, get a dual system with all other things being equal ... would the Xeon beat the AMD in throughput because you effectively have a 3 CPU system when you get done for the day...
THAT is *MY* question. The problem is that the only way I know to find out is to buy them both and test fly them. When I was working I could do that ... and would not lose sleep over it. But now I am on a budget and can't buy a new computer a quarter (yes, go ahead be annoyed with me, but I was for a time doing a complete new computer or a major upgrade to one in the farm every 3 months).
Has anyone done this?
I know when I look at the numbers the Xeons seem to do work pretty fast ...
> For total throughput, are 2
)
> For total throughput, are 2 AMD "Whatever" going to beat a pair of Xeon from
> the same price point.
>
> I mean, lets say you budget $500 per CPU, get a dual system with all other
> things being equal ... would the Xeon beat the AMD in throughput because you
> effectively have a 3 CPU system when you get done for the day...
Well, if you're talking about a single dual processor system the AMD is even further ahead since the Opertons (the server multiprocessor version of the Athlon64) each have their own build in dual channel DDR (The Xeons share a common bus) and Hypertransport interconnect which allows them to share bandwidth (Non-Uniform Memory Access). Not comparing the most recent hardware and it's as a server rather than a cruncher, but it gives you an idea of the comparison: http://www.aceshardware.com/read.jsp?id=60000275
> THAT is *MY* question. The problem is that the only way I know to find out is
> to buy them both and test fly them. When I was working I could do that ...
> and would not lose sleep over it. But now I am on a budget and can't buy a
> new computer a quarter (yes, go ahead be annoyed with me, but I was for a time
> doing a complete new computer or a major upgrade to one in the farm every 3
> months).
>
> Has anyone done this?
>
> I know when I look at the numbers the Xeons seem to do work pretty fast ...
Well, the speed for 2 Xeons will be less than 2x the speed of one, due to the shared bus and message passing for cache coherency, the speed for Dual Opertons should be very close to 2x the speed of a single Athlon64/Opteron. Of course then the OS adds to the confusion (32-bit or 64-bit, NUMA aware or not, Windows or Linux)
> Well, if you're talking
)
> Well, if you're talking about a single dual processor system the AMD is
> even further ahead since the Opertons (the server multiprocessor version of
> the Athlon64) each have their own build in dual channel DDR (The Xeons share a
> common bus) and Hypertransport interconnect which allows them to share
> bandwidth (Non-Uniform Memory Access).
And memory access is a prime factor in speed. I bought two Fry's specials which bundle a CPU with MB for loss leader. I could only afford one, but I bought one high end MB. So, at the end of the day I had two Intel P4 4.2 GHz processors installed and running, one on a high end MB with dual channel memory access the other without. The one did SETI@Home in roughly 3 hours and change (10K Seconds) and the other roughly 4 hours (15K seconds).
Earlier this week I upgraded the slow machine with a new A-Bit IC7-G MB and now that machine is running about as fast as the other one ...
> Not comparing the most recent hardware
> and it's as a server rather than a cruncher, but it gives you an idea of the
> comparison: http://www.aceshardware.com/read.jsp?id=60000275
I will look at that ...
> Well, the speed for 2 Xeons will be less than 2x the speed of one, due to
> the shared bus and message passing for cache coherency, the speed for Dual
> Opertons should be very close to 2x the speed of a single Athlon64/Opteron.
> Of course then the OS adds to the confusion (32-bit or 64-bit, NUMA aware or
> not, Windows or Linux)
Yeah, I know. The point is that though we don't see the full benefit of 2 CPUs in almost all systems is from the memory access issues and cache coherency and the like. But with the Opteron/Athlon I would get 2 CPUs with the ability to run 2 models and with the Xeon I would have 4 in-flight. So, though the AMD set-up processes faster, I would, in theory, get higher throughput on the Xeon.
that brings us to the
)
that brings us to the question :
should i go for an upgrade to dual core or dual system... meaning another setup with a plain athlon, cheapo style..
In the end I think I might just add another system to get more work done..
> that brings us to the
)
> that brings us to the question :
>
> should i go for an upgrade to dual core or dual system... meaning another
> setup with a plain athlon, cheapo style..
>
> In the end I think I might just add another system to get more work done..
I just replaced a cheap MB that caused all the models to run slow. I knew it would likely do that as it was a "loss-leader" bundle at Frys. But I could not get the CPUs cheaper naked. So, I got one good MB, two cheapo (wife bought one, one per customer) bundles and sure enough the good MB did SETI@Home in about 3 hours something, the cheap one was more like 4 hours something.
So, that should help my processing ... I will be getting more help when I buy new systems later this year ... Though I am still unsure which is the best way to go, though it still looks to me like the Xeon would still produce the most work, and with an Intel MB be the most stable.
I bought a dual AMD a couple years ago and had nothing but trouble with it ... :(
helpfull -> helpful effect ->
)
helpfull -> helpful
effect -> affect
"My other computer is a virus farm."