| View previous topic :: View next topic |
| Author |
Message |
Cossins
Veteran
Joined: 21 Mar 2003
Posts: 1136
Location: Copenhagen, Denmark
|
 Posted: Mon Nov 03, 2003 1:05 pm Post subject: |
 |
|
| Evangelion wrote: | | NicholasDWolfwood wrote: | | Mhz does not equal Performance |
Yes it does. Or are you saying that 2GHz Opteron (for example) is not faster than 1.8GHz Opteron? Sure, comparison fails if you compare different CPU-families, but if you are looking at the same CPU-family, you can compare their performance by looking at their MHz-rating. |
Yes, and as NicholasDWolfwood wrote, MHz does not equal performance. There is a difference between equality and comparability...
- Simon |
|
| Back to top |
|
 |
Evangelion
Veteran
Joined: 31 May 2002
Posts: 1087
Location: Helsinki, Finland
|
| Posted: Mon Nov 03, 2003 1:30 pm Post subject: |
|
|
| Cossins wrote: | | Evangelion wrote: | | NicholasDWolfwood wrote: | | Mhz does not equal Performance |
Yes it does. Or are you saying that 2GHz Opteron (for example) is not faster than 1.8GHz Opteron? Sure, comparison fails if you compare different CPU-families, but if you are looking at the same CPU-family, you can compare their performance by looking at their MHz-rating. |
Yes, and as NicholasDWolfwood wrote, MHz does not equal performance. |
Inside one CPU-family it does. 2GHz Opteron is faster than 1.8GHz Opteron. And 1.8Ghz Opteron is fastre than 1.6Ghz Opteron. That is 100% fact. Of course, one can't do the same comaprison between different CPu-families (say Opteron and P4)
_________________
My tech-blog | My other blog |
|
| Back to top |
|
|
Cossins
Veteran
Joined: 21 Mar 2003
Posts: 1136
Location: Copenhagen, Denmark
|
| Posted: Mon Nov 03, 2003 1:54 pm Post subject: |
|
|
| Evangelion wrote: | | Cossins wrote: | | Evangelion wrote: | | NicholasDWolfwood wrote: | | Mhz does not equal Performance |
Yes it does. Or are you saying that 2GHz Opteron (for example) is not faster than 1.8GHz Opteron? Sure, comparison fails if you compare different CPU-families, but if you are looking at the same CPU-family, you can compare their performance by looking at their MHz-rating. |
Yes, and as NicholasDWolfwood wrote, MHz does not equal performance. |
Inside one CPU-family it does. 2GHz Opteron is faster than 1.8GHz Opteron. And 1.8Ghz Opteron is fastre than 1.6Ghz Opteron. That is 100% fact. Of course, one can't do the same comaprison between different CPu-families (say Opteron and P4) |
Congratulations, you got the point.
- Simon |
|
| Back to top |
|
|
Evangelion
Veteran
Joined: 31 May 2002
Posts: 1087
Location: Helsinki, Finland
|
| Posted: Mon Nov 03, 2003 2:05 pm Post subject: |
|
|
| Cossins wrote: | Congratulations, you got the point.
- Simon |
Well whoppee  . But you do realize that saying "MHz does not equal performance" is simply wrong?
_________________
My tech-blog | My other blog |
|
| Back to top |
|
|
krunk
Guru
Joined: 27 Jul 2003
Posts: 316
|
| Posted: Mon Nov 03, 2003 2:28 pm Post subject: |
|
|
Comparingmhz doesn't demonstrate anything. For example:
"How fast is a 1.8 ghz athlon XP?"
"Oh, about 200 mhz faster than a 1.6 athlon XP."
That tells me nothing.
Nicholas's definition is much more informative:
Performance = Speed (mhz) * Instructions per clock
This a is MUCH more informative description of "speed", which is platform independant. I imagine the 512 L2 cache in the Barton cores give a nice little boost in performance, it's still part of the "XP" line of processors....which means a Barton with the same number of mhz as a Palamino core "XP" would be "faster".  |
|
| Back to top |
|
|
Evangelion
Veteran
Joined: 31 May 2002
Posts: 1087
Location: Helsinki, Finland
|
| Posted: Mon Nov 03, 2003 2:34 pm Post subject: |
|
|
| krunk wrote: | Comparingmhz doesn't demonstrate anything. For example:
"How fast is a 1.8 ghz athlon XP?"
"Oh, about 200 mhz faster than a 1.6 athlon XP."
That tells me nothing. |
It tells you that it's faster than 1.6GHz Athlon XP
| Quote: | Nicholas's definition is much more informative:
Performance = Speed (mhz) * Instructions per clock
This a is MUCH more informative description of "speed", which is platform independant. |
Really? Does it tell me how fast it will compile KDE, what kind of FPS I will get in Quake by using that CPU etc. etc.? Didn't you know that "MIPS" is an acronym for "Meaningless Indicator of Processor Speed"  . Now, some people try to claim that it stands for "Millions of instructions Per Second", but don't listen to them.
Mhz as an indicator of speed will become irrelevant when we come up with clockless CPU's
_________________
My tech-blog | My other blog |
|
| Back to top |
|
|
Cossins
Veteran
Joined: 21 Mar 2003
Posts: 1136
Location: Copenhagen, Denmark
|
| Posted: Mon Nov 03, 2003 3:58 pm Post subject: |
|
|
Evangelion:
As you said yourself, clock speed = performance when we compare processors from the same family. The thing is: we don't.
Since other processors than Athlons and/or Pentiums exist, clock speed != performance.
Now, isn't it obvious that a P4 2 GHz is faster than a P4 1.8 GHz? Yes. But it isn't quite obvious that an Athlon XP Barton 2500+ (running at 1.8 GHz, IIRC) is much faster than your average P4 2 GHz.
Now, what was the point of starting this discussion?
- Simon |
|
| Back to top |
|
|
professorn
Apprentice
Joined: 18 Sep 2003
Posts: 235
Location: Stockholm, Sweden
|
| Posted: Mon Nov 03, 2003 5:58 pm Post subject: |
|
|
But its still possible to change the MIPS in one famely  |
|
| Back to top |
|
|
Evangelion
Veteran
Joined: 31 May 2002
Posts: 1087
Location: Helsinki, Finland
|
| Posted: Mon Nov 03, 2003 7:03 pm Post subject: |
|
|
| Cossins wrote: | Evangelion:
As you said yourself, clock speed = performance when we compare processors from the same family. The thing is: we don't. |
Well, I wasn't talking just about P4 vs. XP comparison, but in general regarding the claim that Mhz does not signify performance.
| Quote: | | Since other processors than Athlons and/or Pentiums exist, clock speed != performance. |
When comparing different CPU's, yes. When comparing same CPU's, no. So claim that "Mhz has nothing to do with performance" is not entirely correct.
_________________
My tech-blog | My other blog |
|
| Back to top |
|
|
Malakin
Veteran
Joined: 14 Apr 2002
Posts: 1692
Location: Victoria BC Canada
|
| Posted: Mon Nov 03, 2003 7:09 pm Post subject: |
|
|
| Quote: | | Prescotts run at an average of 103 degrees CELCIUS. Don't get it. |
You say degrees and then you say watts, it's watts of course but the 3.2Ghz P4 is already maxing around that so it's nothing new.
| Quote: | | The reason AMD has the reputation of being "hotter" is because of inadequate cooling. |
I'd have to dissagree with this. AMD has a reputation for being hot because back in the days of 1Ghz Athlons and 1Ghz Pentium 3's the Athlons _were_ much hotter. Pentium 3 1Ghz was 26-33 watts while an Athlon 1Ghz was putting out 49 watts. (both typical power consumption, not max)
| Quote: | | Now, I won't lie, Athlons are on average hotter than *lower end" P4's |
Well low end these days would be a P4 Northwood 1.8Ghz at 50 watts (typical) and an Athlon XP Thoroughbred 1800 at 46 watts (typical) so although close the P4 is still producing more heat.
Or maybe you meant Duron/celeron.
Duron 1.3 Ghz is 55W (typical) while a Celeron 1.8Ghz is 66W (typical), both around the same in performance. Duron/Celerons are some of the worst chips when it comes to heat/performance since removing some of the cache barely effects power usage but destroys their performance.
| Quote: | | but if you are looking at the same CPU-family, you can compare their performance by looking at their MHz-rating. |
Let's not forget cache sizes and FSB speeds :) |
|
| Back to top |
|
|
mmealman
Guru
Joined: 02 Nov 2002
Posts: 348
Location: Florida
|
| Posted: Mon Nov 03, 2003 7:48 pm Post subject: |
|
|
I have a much older 1.4Ghz Thunderbird AMD that I would personally put in the "should've bought a pentium" range of experiences. There was nothing at all wrong with the chip itself, it's just that the motherboards at the time where all using some VIA KT133 chip that was basically a POS.
The Thunderbirds ran hot and the bad motherboard chipping caused headaches. I think most of the "AMD is bad because..." stories you can link back to these much older designs.
I soon after replaced the above with an XP 1800 on a nice ASUS board and I couldn't have been happier with it over the years. It's quiet, doesn't run hot and has been extremely stable and reliable.
I'm replacing the above with an XP 2800, because I think the XP design is just solid and honestly the chips give you the best bang for the buck. There's no question that the newer 3.2 Pentiums are faster, but you pay for that speed in both price and noise levels. |
|
| Back to top |
|
|
krunk
Guru
Joined: 27 Jul 2003
Posts: 316
|
| Posted: Mon Nov 03, 2003 8:40 pm Post subject: |
|
|
| Quote: |
It tells you that it's faster than 1.6GHz Athlon XP
|
Hmmm, don't think you finished reading my post:
| Quote: |
I imagine the 512 L2 cache in the Barton cores give a nice little boost in performance, it's still part of the "XP" line of processors....which means a Barton with the same number of mhz as a Palamino core "XP" would be "faster".
|
Also as Malakin pointed out, cpu's in the same family have different BUS speeds with the same Mhz rating, which significantly affects performance as well.
and
| Quote: |
Really? Does it tell me how fast it will compile KDE, what kind of FPS I will get in Quake by using that CPU etc. etc.? Didn't you know that "MIPS" is an acronym for "Meaningless Indicator of Processor Speed" . Now, some people try to claim that it stands for "Millions of instructions Per Second", but don't listen to them.
|
Actually, yes it does......much more than a simple Mhz comparison will.
mmealman:
yeah those 133's chipsets were crap. I had a Athlon 1.4 and it put off heat like a space heater. I put mine in a KG7-RAID though, and had a very nice experience with it. It was the most stable board I've ever used until this ASUS A7N8X I have right now which is on par......VIA is notorius for buggy chipsets. I don't care how many benchmarks VIA wins, I'll wait till NVIDIA or someone else comes up with a solution before I buy VIA again....they're kt266 and kt266A was crap too. |
|
| Back to top |
|
|
TheJammnman
n00b
Joined: 29 Oct 2003
Posts: 29
Location: Somewhere in my head.
|
| Posted: Tue Nov 04, 2003 7:33 pm Post subject: mother boards |
|
|
I have a Kt3ultra2 motherboard by MSI with a VIA chipset and have no problems with it. I also have a 1.73ghz AMD athlon XP, my computer does not get very hot. The hottest part of my computer is the power supply. Also the more processer/system use the hotter the power supply runs.
_________________
There are only two you need in life.
WD-40 and Duct tape.
If doesn't move and should, use WD-40.
If it does move and shouldn't, use Duct tape. |
|
| Back to top |
|
|
kwiqsilver
Guru
Joined: 19 Mar 2003
Posts: 360
|
| Posted: Tue Nov 04, 2003 11:55 pm Post subject: |
|
|
| Evangelion wrote: | | Cossins wrote: | | Evangelion wrote: | | NicholasDWolfwood wrote: | | Mhz does not equal Performance |
Yes it does. Or are you saying that 2GHz Opteron (for example) is not faster than 1.8GHz Opteron? Sure, comparison fails if you compare different CPU-families, but if you are looking at the same CPU-family, you can compare their performance by looking at their MHz-rating. |
Yes, and as NicholasDWolfwood wrote, MHz does not equal performance. |
Inside one CPU-family it does. 2GHz Opteron is faster than 1.8GHz Opteron. And 1.8Ghz Opteron is fastre than 1.6Ghz Opteron. That is 100% fact. Of course, one can't do the same comaprison between different CPu-families (say Opteron and P4) |
MHz does not equal performance, even in the same family. Run a barton chip at 2GHz and benchmark it. Then drop the multiplier (but not the bus) by 50% and benchmark it again. The drop in performance will be less than 50%. A chip with a higher clock will be faster than the same chip at a lower clock, but not in proportion to the clock difference. If performance equaled clock speed, doubling the clock would double the performance.
You might as well say "transistor count equals performance". It's just as wrong, and in the same way.
The only thing that equals performance is performance (i.e. benchmark results).
What you might be trying to say is "performance scales with clock speed" (when the clock goes up the performance goes up). That is true.
Personally, I prefer the Athlon because for $X you can get higher frame rates, lower compile times, etc. than an intel solution.
The computer I'm putting together now has a Celeron (P4 based 2.2GHz), because my Athlon (aka the heater) runs at over 150F, which is just too much for a 24/7 box.
The celeron has been running emerge for two days straight now, and its heatsink is still cool to the touch. It's got enough power to be a good workstation and run some servers for my lan, but I wouldn't want to use it to play any game more intensive than spades or xbill. |
|
| Back to top |
|
|
Malakin
Veteran
Joined: 14 Apr 2002
Posts: 1692
Location: Victoria BC Canada
|
| Posted: Wed Nov 05, 2003 2:16 am Post subject: |
|
|
| Quote: | | The computer I'm putting together now has a Celeron (P4 based 2.2GHz), because my Athlon (aka the heater) runs at over 150F, which is just too much for a 24/7 box. |
(150F = 65.5C) This _is_ too high. Must be a cheap oem cooler.
Celeron 2.2Ghz puts out 57W, If you wanted to stick with Intel you could get a Pentium 4 1.8Ghz which would be faster and puts out only 49.6W although it costs quite a bit more. An Athlon XP1800 would be cheaper then the celeron and only puts out only 46W. |
|
| Back to top |
|
|
krunk
Guru
Joined: 27 Jul 2003
Posts: 316
|
| Posted: Wed Nov 05, 2003 2:39 am Post subject: |
|
|
| I'd almost say 65.5 C is a misread.....I haven't seen many athlons that didn't lock when they hit 60 C (actually at a 60 C read, the die temp is much higher since the old T-birds relied on a diode under the socket). |
|
| Back to top |
|
|
kwiqsilver
Guru
Joined: 19 Mar 2003
Posts: 360
|
| Posted: Wed Nov 05, 2003 3:12 am Post subject: |
|
|
I've got a Zalmann flower cooler on there with a huge fan. The case has 6 80mm fans. The chip is an AthlonXP 2000+.
Sometimes I'd get home from work and the system would be locked. Upon restart, the boot screen would list it at upper 140s or lower 150s. I set the shutdown temp to 160F or so.
According to this graph at tom's hardware, the Celeron runs cooler than the Athlon:
Athlon: 85-90C
Celeron: 68-70C
The celeron also wins my index finger test, which I give the most weight to. |
|
| Back to top |
|
|
Evangelion
Veteran
Joined: 31 May 2002
Posts: 1087
Location: Helsinki, Finland
|
| Posted: Wed Nov 05, 2003 9:13 am Post subject: |
|
|
| krunk wrote: | | Also as Malakin pointed out, cpu's in the same family have different BUS speeds with the same Mhz rating, which significantly affects performance as well. |
By "family" I means similar CPU's. If the CPU has different amount of cache and/or different FSB-speeds, it's no longer similar.
| Quote: | | Actually, yes it does......much more than a simple Mhz comparison will. |
MIPS doesn't really tell anything about performance. Remember when Apple told how the Cube is a "supercomputer on your desktop"? they based that claim on it's MIPS-rating. In reality, Macs of that time were getting bitch-slapped by PC's.
_________________
My tech-blog | My other blog |
|
| Back to top |
|
|
Evangelion
Veteran
Joined: 31 May 2002
Posts: 1087
Location: Helsinki, Finland
|
| Posted: Wed Nov 05, 2003 9:17 am Post subject: |
|
|
| kwiqsilver wrote: | | MHz does not equal performance, even in the same family. Run a barton chip at 2GHz and benchmark it. Then drop the multiplier (but not the bus) by 50% and benchmark it again. The drop in performance will be less than 50%. |
Congratulations. You just proved the fact that performance and Myhz don't scale identically. But I thought everyone knew that?
| Quote: | | A chip with a higher clock will be faster than the same chip at a lower clock, but not in proportion to the clock difference. |
And that proves.... what? That performance and MHz don't scale identically? But that's common knowledge already.
| Quote: | | If performance equaled clock speed, doubling the clock would double the performance. |
When I say that "MHz equals performance" I mean that fact that if you take two otherwise identical CPU, but one has higher MHz than the other, the one with higher MHz will be faster. When they come up with clockless CPU's, come talk to me.-
| Quote: | | What you might be trying to say is "performance scales with clock speed" (when the clock goes up the performance goes up). That is true. |
Well, yes. That is what I'm saying.
_________________
My tech-blog | My other blog |
|
| Back to top |
|
|
Cossins
Veteran
Joined: 21 Mar 2003
Posts: 1136
Location: Copenhagen, Denmark
|
| Posted: Wed Nov 05, 2003 9:49 am Post subject: |
|
|
| Evangelion wrote: | | Well, yes. That is what I'm saying. |
No, it isn't... You say that MHz equals performance, as in =. That means they are more or less linearily proportional. Also, you said earlier that a 1800 MHz processor is 200 MHz "faster" than a 1600 MHz processor, which is thus untrue. No one besides you is comparing CPUs only within the same family, as that is quite a narrow perspective. The point is that the fastest CPU is not necessarily the one with the highest clock speed.
And Evangelion, I don't think you lead this discussion anywhere with that being stubborn, acting like a spoiled kid...
- Simon |
|
| Back to top |
|
|
krunk
Guru
Joined: 27 Jul 2003
Posts: 316
|
| Posted: Wed Nov 05, 2003 2:17 pm Post subject: |
|
|
| Quote: | | When I say that "MHz equals performance" I mean that fact that if you take two otherwise identical CPU, but one has higher MHz than the other, the one with higher MHz will be faster. When they come up with clockless CPU's, come talk to me.- |
Even that narrow of a definition doesn't follow. For example, I could just as easily say: Take two otherwise identical CPU, but one has more L2 cache than the other, the one with more L2 cache will be fster.....Therefore, L2 Cache equals performance!!!
In fact, that same statement holds true if you insert: Mhz, L2 cache, L1 cache, onboard memory controller, Front Side Bus (which doesn't necessarily change the Mhz), instructions per second, FP, integer calculations per second, ad nauseum. |
|
| Back to top |
|
|
kwiqsilver
Guru
Joined: 19 Mar 2003
Posts: 360
|
| Posted: Wed Nov 05, 2003 9:16 pm Post subject: |
|
|
| Evangelion wrote: | | kwiqsilver wrote: | | MHz does not equal performance, even in the same family. Run a barton chip at 2GHz and benchmark it. Then drop the multiplier (but not the bus) by 50% and benchmark it again. The drop in performance will be less than 50%. |
Congratulations. You just proved the fact that performance and Myhz don't scale identically. But I thought everyone knew that?
When I say that "MHz equals performance" I mean that fact that if you take two otherwise identical CPU, but one has higher MHz than the other, the one with higher MHz will be faster. When they come up with clockless CPU's, come talk to me.- |
That's not an equals relationship though, that's a scaling relationship. They're not synonymous.
| Quote: |
| Quote: | | What you might be trying to say is "performance scales with clock speed" (when the clock goes up the performance goes up). That is true. |
Well, yes. That is what I'm saying. |
Then I agree with you. And I don't think anybody here would disagree with that. |
|
| Back to top |
|
|
Malakin
Veteran
Joined: 14 Apr 2002
Posts: 1692
Location: Victoria BC Canada
|
| Posted: Wed Nov 05, 2003 11:32 pm Post subject: |
|
|
| Quote: | According to this graph at tom's hardware, the Celeron runs cooler than the Athlon:
Athlon: 85-90C
Celeron: 68-70C |
The article doesn't say what the temperatures are for or where they were obtained from. These appear to mostly (all?) be maximum operating temperatures, not what temperature the CPU runs at. Anything over 60C is too hot for current CPU's IMO whether they're crashing or not.
The link I provided before is much more reliable as it states what everything means and the sources for the information so you can go through intel/amd documents and actually see the sources for yourself if you were so inclined.
http://users.erols.com/chare/elec.htm
Here's a good graph that shows CPU performance, you can compare just about anything here:
http://www6.tomshardware.com/cpu/20030217/cpu_charts-24.html
These two cpu's are currently about the same price, about the same heat output but one is over 50% faster:
173FPS Athlon XP2100 (56W)
113FPS Celeron 2.0Ghz (53W)
| Quote: | | I've got a Zalmann flower cooler on there with a huge fan. |
Which model of cooler? How big is the fan and what rpm is it running at? Is your XP2000 a Palomino or a Tbred-A or a Tbred-B? Is it attached with thermal tape, regular heatsink compound or silver heatsink compound? If it's compound is it a really thin layer and nothing more then what's needed for both surfaces to contact each other?
If it really is hitting 65C then something should be improved with the cooler setup. A stock AMD cooler will run this chip around ~48C. |
|
| Back to top |
|
|
kwiqsilver
Guru
Joined: 19 Mar 2003
Posts: 360
|
| Posted: Thu Nov 06, 2003 12:42 am Post subject: |
|
|
| Malakin wrote: |
The article doesn't say what the temperatures are for or where they were obtained from. These appear to mostly (all?) be maximum operating temperatures, not what temperature the CPU runs at. Anything over 60C is too hot for current CPU's IMO whether they're crashing or not.
|
I think I need to re-seat the heatsink, or maybe I just have too much heat production in there. When I first got the chip it wasn't that hot, even with the amd stock heatsink.
| Quote: |
The link I provided before is much more reliable as it states what everything means and the sources for the information so you can go through intel/amd documents and actually see the sources for yourself if you were so inclined.
http://users.erols.com/chare/elec.htm
Here's a good graph that shows CPU performance, you can compare just about anything here:
http://www6.tomshardware.com/cpu/20030217/cpu_charts-24.html
These two cpu's are currently about the same price, about the same heat output but one is over 50% faster:
173FPS Athlon XP2100 (56W)
113FPS Celeron 2.0Ghz (53W)
|
Oh well...maybe I should have gotten another Athlon. I wanted my new machine to be very cool and quiet. It meets both of those requirements. It's also more than fast enough for me, since the most intensive task I run on it is kino. But I would have preferred AMD pricing.
| Quote: |
| Quote: | | I've got a Zalmann flower cooler on there with a huge fan. |
Which model of cooler? How big is the fan and what rpm is it running at? Is your XP2000 a Palomino or a Tbred-A or a Tbred-B? Is it attached with thermal tape, regular heatsink compound or silver heatsink compound? If it's compound is it a really thin layer and nothing more then what's needed for both surfaces to contact each other?
If it really is hitting 65C then something should be improved with the cooler setup. A stock AMD cooler will run this chip around ~48C. |
I don't remember the specific zalman model. It's the larger of their pure copper flower Athlon models. The fan is 90 or 100 mm I think and running at full speed. It moves a lot of air...straight into the side of the case. I was thinking about drilling some holes near it, or installing a screen for better flow. When I leave the case open, it stays much cooler. The heatsink is riding on a thin layer of arctic silver II.
I think part of my problem might be the outside heat. Recently it's been running much cooler. When I had the worst issues, it was 120F outside, and that room was getting up to 90 during the day (I have the AC set to work less when I'm not home). After a few days of locking up, I decided leave the case open with a 6" fan pointed in on the chip, and it was better.
I should see how hot it's running now. |
|
| Back to top |
|
|
krunk
Guru
Joined: 27 Jul 2003
Posts: 316
|
| Posted: Thu Nov 06, 2003 12:59 am Post subject: |
|
|
Without excellant cooling it's very hard to approach ambient with any heat producer....let along little power stations like CPU's. With mid-range cooling solutions (like the Zalman), it's good to get about 10 to 15 degrees C over ambient (that's ~50-60 F). So if your ambient temps were hitting 90 degrees, than the math says your cpu would be about 140-150 F....right about where you said it is
It's not the cpu................it's the jungle, baby! |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Gentoo Chat
