How about we compare similar priced chips, thats the ONLY method we can use to be fair, you cant do clock per clock, because the p4 gives up ipc for top clockspeed.
Also you cant do ipc, because the athlon has more ipc and lower clock.
You could do the fastest from each, but that changes week to week month to month, and is not a good indicator of static performance.
My suggestion is we do price/price, because while its not static, it is the only way to logically determine which chips to compare imo.
The Cash Left In My Pocket,The BEST Benchmark
Also you cant do ipc, because the athlon has more ipc and lower clock.
You could do the fastest from each, but that changes week to week month to month, and is not a good indicator of static performance.
My suggestion is we do price/price, because while its not static, it is the only way to logically determine which chips to compare imo.
The Cash Left In My Pocket,The BEST Benchmark Problem is, we're talking about optimizations, not which CPU would be the best choice for a specific person (or overall).
That's my point exactly. I'm trying to get people to see IPC from the same perspective as they see clockspeed.
Agreed.
I agree for general use, but we're talking about programs being more optimized for Athlon than P4, or vice versa.
<font color=blue>Hi mom!</font color=blue>
Ah, I see what you're saying. You're saying which CPUs to compare using my method. The reason I don't think that's needed is because we're comparing with the Sandra benchmarks, using that as the baseline (as opposed to price, or clock speed, or PR ratings, or...). Assumedly if someone was to try and do this and actually use the results, they would do it on several different clockspeeds of each processor being compared.
<font color=blue>Hi mom!</font color=blue>
<font color=blue>Hi mom!</font color=blue>
eden
Champion
- Oct 3, 2001
- 26,541
- 0
- 50,780
I agree with you. It's hard to find the truth here. But again, I did not and will not put in IPC vs clock speed. You were right since a long time ago, both have to be in harmony to compare well, and not one better than other.
BUT, like you said to Mat, we're comparing optimizations. Look you may not see it my way, but the only reason I am going clock for clock, is only to level the truth and find it. There may be no concrete truth, but I will continue standing by my point, that in order for something to trully be optimized on a PC home system, it HAS to have been optimized to run higher per clock than the other, to be considered well optimized. I would not compare clock per clock for other reasons than just to show you the real point of optim. I mean ok Q3 runs better on Athlons, P4s have been optim. for Q3, but they do not outrun Athlons still, however at 2.4GHZ they do. So now we can say it's optimized to run better on P4? No way man, that'd be hyprocrisy, denial. If it had run better per clock by the optimization, at 2.4GHZ I'd fully agree, the system is more than optimal for Q3 gaming, it's the best out there. However, NADA, it isn't the case. Same thing applies to the Mpeg 4. I won't lie to you, P4s rock at it. I haven't seen updated Flask benchs, I do have seen the ones from december as they compared the new version's "P4 optimizations" vs AthlonXP's SSE. In most benches, a 2GHZ P4 was better than AthlonXP, however either by a frame or 3. (the scale of the bench stats show as if it's a lot, since the scale is huge and by increments of almost 30 only on a whole line of measure.)
This leads to me that a 1.66GHZ AthlonXP benefitted hugely from SSE optimizations, and that it runs much better than P4's optimizations, as per clock, in one bench the 2GHZ at say 1.66GHZ would have been near or equal AXP 2000's result, and in other benchs, the 2 at 1.66GHZ would be below. This means the optimizations can go further, until it is really optimized to RUN HIGHER.
Again I am debating here, so if someone, I swear, if someone dares to scream at me, calling me some idiot for stating an opinion again, please get the hell out. I am sick of anyone criticizing each and every comment I make, as if I am wrong all the time.
Now do you get what I mean man? Please stop referring to G4 anymore!
So to, me that is the true meaning of running optimally better on X cpu, otherwise why buy one that might have the same clock as the other (although rarely the case anymore with P4's high speeds, but again, for truth purposes) when the other CPU is better optimized?
PS: I just remembered, the Newtek Lightwave 7B test which to me is a cheat, will be, for the sake of this argument, used as an example where the P4 optimizations REALLY show. In this context, yes by all means, a P4 is your best friend, it is OPTIMIZED at its best for L7B, and Athlons per clock cannot even compete, so it is optimized, go for it.
--
Thunderbirds in wintertime, Northwoods in summertime! :lol:
BUT, like you said to Mat, we're comparing optimizations. Look you may not see it my way, but the only reason I am going clock for clock, is only to level the truth and find it. There may be no concrete truth, but I will continue standing by my point, that in order for something to trully be optimized on a PC home system, it HAS to have been optimized to run higher per clock than the other, to be considered well optimized. I would not compare clock per clock for other reasons than just to show you the real point of optim. I mean ok Q3 runs better on Athlons, P4s have been optim. for Q3, but they do not outrun Athlons still, however at 2.4GHZ they do. So now we can say it's optimized to run better on P4? No way man, that'd be hyprocrisy, denial. If it had run better per clock by the optimization, at 2.4GHZ I'd fully agree, the system is more than optimal for Q3 gaming, it's the best out there. However, NADA, it isn't the case. Same thing applies to the Mpeg 4. I won't lie to you, P4s rock at it. I haven't seen updated Flask benchs, I do have seen the ones from december as they compared the new version's "P4 optimizations" vs AthlonXP's SSE. In most benches, a 2GHZ P4 was better than AthlonXP, however either by a frame or 3. (the scale of the bench stats show as if it's a lot, since the scale is huge and by increments of almost 30 only on a whole line of measure.)
This leads to me that a 1.66GHZ AthlonXP benefitted hugely from SSE optimizations, and that it runs much better than P4's optimizations, as per clock, in one bench the 2GHZ at say 1.66GHZ would have been near or equal AXP 2000's result, and in other benchs, the 2 at 1.66GHZ would be below. This means the optimizations can go further, until it is really optimized to RUN HIGHER.
Again I am debating here, so if someone, I swear, if someone dares to scream at me, calling me some idiot for stating an opinion again, please get the hell out. I am sick of anyone criticizing each and every comment I make, as if I am wrong all the time.
Now do you get what I mean man? Please stop referring to G4 anymore!
So to, me that is the true meaning of running optimally better on X cpu, otherwise why buy one that might have the same clock as the other (although rarely the case anymore with P4's high speeds, but again, for truth purposes) when the other CPU is better optimized?
PS: I just remembered, the Newtek Lightwave 7B test which to me is a cheat, will be, for the sake of this argument, used as an example where the P4 optimizations REALLY show. In this context, yes by all means, a P4 is your best friend, it is OPTIMIZED at its best for L7B, and Athlons per clock cannot even compete, so it is optimized, go for it.
--
Thunderbirds in wintertime, Northwoods in summertime! :lol:
Is it even possible to optimize for something as synthetic as Sandra? Sandra was of course just an example, something as synthetic as possible would be best.
Eden, I still think you're missing my point. Let's say you have two CPUs, one is 4 IPC and 1GHz. The other is 2 IPC and 2GHz. Now, the first will always have the highest IPC and the second will always have the highest clockspeed, because that's how they're designed.
Take a program and run it on each. With no optimizations, the program will run better per clock than the second. Add a little bit of optimization for the second CPU, and the first still runs faster per clock. Suddenly, you have a program that runs faster per clock on the first processor, but is more optimized for the second. Make sense?
<font color=blue>Hi mom!</font color=blue>
Eden, I still think you're missing my point. Let's say you have two CPUs, one is 4 IPC and 1GHz. The other is 2 IPC and 2GHz. Now, the first will always have the highest IPC and the second will always have the highest clockspeed, because that's how they're designed.
Take a program and run it on each. With no optimizations, the program will run better per clock than the second. Add a little bit of optimization for the second CPU, and the first still runs faster per clock. Suddenly, you have a program that runs faster per clock on the first processor, but is more optimized for the second. Make sense?
<font color=blue>Hi mom!</font color=blue>
to quote you from above...
"my head hurts"
lol
Intel vs AMD has gotten damn complicated...
there is no valid 100% scientific comparison, there are some generalizations that will give a ballpark figure on the differences...
hell, compare the Athlon vs P4 in integer performance only, its a decent fight.
lets compare pure FPU performance... ok, lets not...
price? well, in theoretical discussions about CPU architecture, who gives a rats ass about price?
<i>The early bird gets the worm but the second moust gets the cheese.</i>
"my head hurts"
lol
Intel vs AMD has gotten damn complicated...
there is no valid 100% scientific comparison, there are some generalizations that will give a ballpark figure on the differences...
hell, compare the Athlon vs P4 in integer performance only, its a decent fight.
lets compare pure FPU performance... ok, lets not...
price? well, in theoretical discussions about CPU architecture, who gives a rats ass about price?
<i>The early bird gets the worm but the second moust gets the cheese.</i>
eden
Champion
- Oct 3, 2001
- 26,541
- 0
- 50,780
Correct, that I would not defy, it's true and I agree.
However you are also missing my point, which comes after yours has been set up. My take on this, is no matter how optimized a processor gets, it is simply not optimized enough until it takes over the other processor's per clock performance (again in benchs, look at two processors at the same clock, not by clock cycle or something small, just how the benchs show A and B at the same clock). So it just isn't optimized well until it really overtakes the same clock speed the other has. Again you should refer to my Lightwave 7B to show you how very powerful optimization can do. My take on that test, is that Newtek either was in cahoots with Intel, although Ray said he doesn't remember any deals, or that they did not include the SSE optimizations for AthlonXP. So until Palomino is also OPTIMIZED for L7B, and that it does outrun the P4 at the same clock, the P4 remains the better buy, even if per clock there is ONE measily frame of difference, the P4 remains the better buy as it also has the higher clock speeds, which multiplies the little frame into a lot.
Now that this is discussed, this was only to see if an optimization is well used. The other side is the overall performance. Now if a 2.4GHZ is almost 20FPS ahead that an AXP 2400 in Q3, even if it lacks good optim. per clock, then here we HAVE to go for overall performance and go for P4.
So there are two sides to see a properly set up CPU: Optimization and Overall Performance. With the former being worth investigating if it is better per clock in the said needed program for you.
--
Thunderbirds in wintertime, Northwoods in summertime! :lol:
However you are also missing my point, which comes after yours has been set up. My take on this, is no matter how optimized a processor gets, it is simply not optimized enough until it takes over the other processor's per clock performance (again in benchs, look at two processors at the same clock, not by clock cycle or something small, just how the benchs show A and B at the same clock). So it just isn't optimized well until it really overtakes the same clock speed the other has. Again you should refer to my Lightwave 7B to show you how very powerful optimization can do. My take on that test, is that Newtek either was in cahoots with Intel, although Ray said he doesn't remember any deals, or that they did not include the SSE optimizations for AthlonXP. So until Palomino is also OPTIMIZED for L7B, and that it does outrun the P4 at the same clock, the P4 remains the better buy, even if per clock there is ONE measily frame of difference, the P4 remains the better buy as it also has the higher clock speeds, which multiplies the little frame into a lot.
Now that this is discussed, this was only to see if an optimization is well used. The other side is the overall performance. Now if a 2.4GHZ is almost 20FPS ahead that an AXP 2400 in Q3, even if it lacks good optim. per clock, then here we HAVE to go for overall performance and go for P4.
So there are two sides to see a properly set up CPU: Optimization and Overall Performance. With the former being worth investigating if it is better per clock in the said needed program for you.
--
Thunderbirds in wintertime, Northwoods in summertime! :lol:
ok, lets say the max IPC for an Athlon is 14, the max IPC for a P4 is 10.
no matter how well it is coded, the P4 still cannot do more than 10 instructions per cycle, so the Athlon will win every time, no matter how well the program is optimized for the P4.
lets say instructions per second, it gets comparable then...
<i>The early bird gets the worm but the second moust gets the cheese.</i>
no matter how well it is coded, the P4 still cannot do more than 10 instructions per cycle, so the Athlon will win every time, no matter how well the program is optimized for the P4.
lets say instructions per second, it gets comparable then...
<i>The early bird gets the worm but the second moust gets the cheese.</i>
Yes, thank you. Instructions per [unit of time] would be the best way of putting it. See what I mean now, Eden? Processors are designed differently, and unless you have two processors with identical IPC, then you can't say that only better performance per clock means optimized.
<font color=blue>Hi mom!</font color=blue>
<font color=blue>Hi mom!</font color=blue>
eden
Champion
- Oct 3, 2001
- 26,541
- 0
- 50,780
So you're telling me that IPC limits the full optimization process? And that as much as you optimize a P4, it will never have better per clock than Athlon's? I hardly find that impressing at all, I mean look at the GF3 then vs GF2 Ultra's big clock speed! Who said it has a better IPC unless optimizations are in, or the Xbar is in? One a GF3 is optimized, it fully destroys any GF2 per clock, regardless of having the Xbar(which without GF3 optimized technology, translates into raw 10% more performance any time)or a lower clock speed.
Again why are you bringing IPC into this? I mean if optimizations are what defies and breaks the limits of a processor, then why bring IPC inside again.
--
Thunderbirds in wintertime, Northwoods in summertime! :lol:
Again why are you bringing IPC into this? I mean if optimizations are what defies and breaks the limits of a processor, then why bring IPC inside again.
--
Thunderbirds in wintertime, Northwoods in summertime! :lol:
your words:
"My take on this, is no matter how optimized a processor gets, it is simply not optimized enough until it takes over the other processor's per clock performance (again in benchs, look at two processors at the same clock, not by clock cycle or something small, just how the benchs show A and B at the same clock). So it just isn't optimized well until it really overtakes the same clock speed the other has. "
you brought up IPC.
IPC doesn't limit the optimization process, you are defining optimization as something else entirely.
No matter how well it is optimized, a processor with an IPC of 10 cannot do more work per cycle than a processor with an IPC of 14.
IPS is totally different, that would be an acceptible comparison, but the P4 will NEVER beat the Athlon at IPC, it wasn't designed to.
<i>The early bird gets the worm but the second moust gets the cheese.</i>
"My take on this, is no matter how optimized a processor gets, it is simply not optimized enough until it takes over the other processor's per clock performance (again in benchs, look at two processors at the same clock, not by clock cycle or something small, just how the benchs show A and B at the same clock). So it just isn't optimized well until it really overtakes the same clock speed the other has. "
you brought up IPC.
IPC doesn't limit the optimization process, you are defining optimization as something else entirely.
No matter how well it is optimized, a processor with an IPC of 10 cannot do more work per cycle than a processor with an IPC of 14.
IPS is totally different, that would be an acceptible comparison, but the P4 will NEVER beat the Athlon at IPC, it wasn't designed to.
<i>The early bird gets the worm but the second moust gets the cheese.</i>
very much so, sandra runs just like any other app, if you optimise your system to run sandra it can cloud judgement.
Then theres the matter of special features, if sandra runs the p4 with sse2, and quake3 does not the p4 would look worse by comparison, whereas if the benchmark didnt use sse2, any apps which did would artificially inflate the scores.
I think eden is having the same problem we are burger, there is NO way to determine WHICH cpus to benchmark appart.
You like the way you mentioned, I like the price/price method, and eden likes the =clock, ALL methods are valid, its not a function of clockspeed or ipc, its just a matter of which cpus do we run the bench's on.
The Cash Left In My Pocket,The BEST Benchmark I would tend to agree, but how to determine how many ips a cpu can perform when that number VARIES based on the app being ran, its parellelism, its optimisations, its function, load ballance, tons of factors prevent a single number from even being devisied.
And for the record quake 3 is not optimised for the p4. The simple playbavck demos love the rdrams high bandwidth, when the graphics intensive demos are ran(nv15) the cpus are much closer and in some cases the amd chip wins.
The Cash Left In My Pocket,The BEST Benchmark but once you have twice the clock speed thats twice instuctions per SECOND, still intel have that and even i accept that for an important guide line of their strategy.
then again AMD is still using .18, not forlong thogh.
anyway AMD kick Intel in the nuts!
<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>
then again AMD is still using .18, not forlong thogh.
anyway AMD kick Intel in the nuts!
<font color=green>
*******
*K.I.S.S*
*(k)eep (I)t (S)imple (S)tupid*
*******
</font color=green>
eden
Champion
- Oct 3, 2001
- 26,541
- 0
- 50,780
Well ok, maybe you are right, so I'll take your word on it.
But Mat is also right here, if you took the NV15 demo, even with optimizations pushed, and the large bandwidth, it shows how much a P4 regardless of clock speed, can be put on strain in such situation. Ya know, I've always and always knew, that if they added back the FPU, the SSE2 opt. would definitly push a P4 high per clock than AthlonXP's (not IPC but when looking at similar clocked CPUs such as P4 vs AthlonXP in benchs) and THEN I may not have been discussing this. I guess I am just not satisfied with optimizations that don't push the CPU to beat the other similar clocked in the home PC environment.
One powerful example is also 3d Studio Max. WHAT IF, it was opt. for P4, which it is, and we took a scene to render which takes usually 3 minutes for Athlon, 6 for P4, once optimized, puts the P4 at 4 min? I agree it's a good opt, but like you said and which I now agree and understand, it's limited by its IPC in 3d Studio Max. So say they added an FPU, moved SSE2 to ALU 32-bit, THEN it possibly could go down to 2 minutes even, whether it's 2.4GHZ or 1.66GHZ! (Assuming that without the opt enabled, the new components would down the time from 6 min to 4)
Anyways I now understand a bit more, I guess we both see opt. different ways, and although we agree they are beneficial, they don't always help to a much needed extent, and there you are limited by your CPU's integrity. That is what I never liked of the P4 anyway, lack of FPU is not always made up by insane clock speeds.
--
Thunderbirds in wintertime, Northwoods in summertime! :lol:
But Mat is also right here, if you took the NV15 demo, even with optimizations pushed, and the large bandwidth, it shows how much a P4 regardless of clock speed, can be put on strain in such situation. Ya know, I've always and always knew, that if they added back the FPU, the SSE2 opt. would definitly push a P4 high per clock than AthlonXP's (not IPC but when looking at similar clocked CPUs such as P4 vs AthlonXP in benchs) and THEN I may not have been discussing this. I guess I am just not satisfied with optimizations that don't push the CPU to beat the other similar clocked in the home PC environment.
One powerful example is also 3d Studio Max. WHAT IF, it was opt. for P4, which it is, and we took a scene to render which takes usually 3 minutes for Athlon, 6 for P4, once optimized, puts the P4 at 4 min? I agree it's a good opt, but like you said and which I now agree and understand, it's limited by its IPC in 3d Studio Max. So say they added an FPU, moved SSE2 to ALU 32-bit, THEN it possibly could go down to 2 minutes even, whether it's 2.4GHZ or 1.66GHZ! (Assuming that without the opt enabled, the new components would down the time from 6 min to 4)
Anyways I now understand a bit more, I guess we both see opt. different ways, and although we agree they are beneficial, they don't always help to a much needed extent, and there you are limited by your CPU's integrity. That is what I never liked of the P4 anyway, lack of FPU is not always made up by insane clock speeds.
--
Thunderbirds in wintertime, Northwoods in summertime! :lol:
TRENDING THREADS
-
-
Question Converting a laptop hard drive into an external hard drive
- Started by hack3rcon
- Replies: 6
-
-
-
Discussion What's your favourite video game you've been playing?- Started by amdfangirl
- Replies: 4K
Facebook
Twitter
Instagram