I just upgraded my Lupu 5.1.1 hardware to a P4 805 - 2.66Ghz dual core with 2X1meg L2 caches. When I run the benchmarks under the System Information utility, my CPU shows only marginally faster (actually the Ray Tracing shows I'm slower) than the standard 1.5Ghz Celeron reference CPU???
Even if the SI util is only using 1 of my cores, I would think I should still be 30 or 40% faster than the Celeron! Should I believe these speeds, or does anyone know why this is happening?
2.6GHz P4 benchmarks barely faster than 1.5GHz Celeron
2.6GHz P4 benchmarks barely faster than 1.5GHz Celeron
[b]Hasten Slowly![/b]
OK, here's another puzzling one... Just for comparison, I booted Lupu on my AMD system. It is an Athlon X2 7850 2.8Ghz dual core with 3 level cache. I pulled a memory stick and made it 1 Gig RAM just to match my Pentium system more closely.
While all the benchmarks win over the standard 1.5Ghz Celeron, they don't win by much. I'd say about 10% on the average, just judging from the bar graphs. I feel it should be nearly twice as fast just based on clock speed alone - very strange! The system does seem noticeably snappy, however.
I understand that benchmarks may not be a truly representative of the real world, but I'm puzzled over the apparent lack of performance? I'm not sure if the benchmarks themselves are unreliable, or if there is some other bottle neck in the system?
While all the benchmarks win over the standard 1.5Ghz Celeron, they don't win by much. I'd say about 10% on the average, just judging from the bar graphs. I feel it should be nearly twice as fast just based on clock speed alone - very strange! The system does seem noticeably snappy, however.
I understand that benchmarks may not be a truly representative of the real world, but I'm puzzled over the apparent lack of performance? I'm not sure if the benchmarks themselves are unreliable, or if there is some other bottle neck in the system?
[b]Hasten Slowly![/b]
Re: 2.6GHz P4 benchmarks barely faster than 1.5GHz Celeron
A 1.6 Pentium M is the equivalent of a P4 3.2.gary wrote:I just upgraded my Lupu 5.1.1 hardware to a P4 805 - 2.66Ghz dual core with 2X1meg L2 caches. When I run the benchmarks under the System Information utility, my CPU shows only marginally faster (actually the Ray Tracing shows I'm slower) than the standard 1.5Ghz Celeron reference CPU???
Even if the SI util is only using 1 of my cores, I would think I should still be 30 or 40% faster than the Celeron! Should I believe these speeds, or does anyone know why this is happening?
Puppys' benchmarks are in accordance with others.
Inspiron 700m, Pent.M 1.6Ghz, 1Gb ram.
Msi Wind U100, N270 1.6>2.0Ghz, 1.5Gb ram.
Eeepc 8g 701, 900Mhz, 1Gb ram.
Full installs
A Celeron M is twice as fast as a Pentium given the same clock speed??? I hope you have some independent benchmarks to support that, especially since Celeron M was enhanced with larger cache so that laptops could compete with desktop P4s. At the very MOST, they would be on par given the same cache size and clock, otherwise Intel has been selling the more powerful CPU (Celeron) cheaper all these years.
However, all that aside, we seem to be going backward in time; i.e., you're suggesting that newer and faster CPUs really means slower, or only marginally faster? In other words, a 1.5Ghz Celeron M (which BTW, is likely 32 bit in the test) is faster than a 2.66Ghz P4 dual core, and only marginally slower than a 2.8Ghz Athlon dual core; both of which are 64 bit CPUs? Even if the benchmark is only employing 1 core, in both cases it should beat the Celeron hands down! PassMark V 7.0 confirms all of this.
Seriously, I have to believe that something else is going on here. Any other suggestions?
However, all that aside, we seem to be going backward in time; i.e., you're suggesting that newer and faster CPUs really means slower, or only marginally faster? In other words, a 1.5Ghz Celeron M (which BTW, is likely 32 bit in the test) is faster than a 2.66Ghz P4 dual core, and only marginally slower than a 2.8Ghz Athlon dual core; both of which are 64 bit CPUs? Even if the benchmark is only employing 1 core, in both cases it should beat the Celeron hands down! PassMark V 7.0 confirms all of this.
Seriously, I have to believe that something else is going on here. Any other suggestions?
[b]Hasten Slowly![/b]
-
jduffy22335
- Posts: 35
- Joined: Thu 16 Apr 2009, 23:29
- Location: Great Mills, MD USA
Re: 2.6GHz P4 benchmarks barely faster than 1.5GHz Celeron
FWIW: My Intel quad core (q6600) 2.4G is reporting to be twice the difference of the reference Celeron in benchmark overall. Either 1/2 when lower is better or about twice when higher is better.gary wrote:I just upgraded my Lupu 5.1.1 hardware to a P4 805 - 2.66Ghz dual core with 2X1meg L2 caches. When I run the benchmarks under the System Information utility, my CPU shows only marginally faster (actually the Ray Tracing shows I'm slower) than the standard 1.5Ghz Celeron reference CPU???
Even if the SI util is only using 1 of my cores, I would think I should still be 30 or 40% faster than the Celeron! Should I believe these speeds, or does anyone know why this is happening?
M725 1.6gb was equal or better than P4 2.4gb; not 3.2 as I
stated previously.
http://hostingator.com/pentium_m.html
stated previously.
http://hostingator.com/pentium_m.html
Inspiron 700m, Pent.M 1.6Ghz, 1Gb ram.
Msi Wind U100, N270 1.6>2.0Ghz, 1.5Gb ram.
Eeepc 8g 701, 900Mhz, 1Gb ram.
Full installs
I remember there were the seme confussion in the Pentium III era when the Pentium 4 was released.
The Pentium 4 CPUs were based on the NetBurst micro-architecture, which differed significantly from P6 micro-architecture used in Pentium II and Pentium III microprocessors. CPU performance in that moment were proportional to its frequency and its efficiency, many micro-architectures balanced this two component to achieve better performance.
The NetBurst microarchitecture used a different approach, it attempted to improve performance primarily by increasing CPU frequency, often at at the expense of efficiency. This was a deccision based in the marketing startegy of that era, when higher frecuencies means higher performance for desktop users, using this new architecure the Pentium 4 models could break the 2 GHz barrier in few time. One of key elements in this approach was Hyper-Pipelined Technology, that was significantly longer pipelines than in previous generation of Pentium processors. While longer pipelines are less efficient than shorter ones, they allow CPU core to reach higher frequencies.
Because the first generation of Pentium 4 processors (Willamette core), proved to be performing not significantly faster, and sometimes slower than the fastest Pentium III microprocessors, Intel added more efficiency improvements to subsequent Pentium 4 cores adding: larger size of level 2 cache, faster FSB frequency, SSE3 instruction set, and Hyper-Threading technology.
Latest Pentium 4 systems achieved better performance with these new components, but at that time the AMD Athlon XP models gained marketing terrain with a better micro-architecture. It was so good that at lower frecuencies it could achive better performance than some Pentium 4 models, for this reason AMD implemented the PR rating system.
Today there are more components, and in the same way more variables that could play an important role in the general speed of a system. Programs must be created to use this new technology, plain not optimized code will execute at the same way and you will gain marginal speed only at higher frecuencies.
Equally benchmark programs must be adjusted to measure new technolgy capability. The question is what you want to measure and how to do it?.
clarf
The Pentium 4 CPUs were based on the NetBurst micro-architecture, which differed significantly from P6 micro-architecture used in Pentium II and Pentium III microprocessors. CPU performance in that moment were proportional to its frequency and its efficiency, many micro-architectures balanced this two component to achieve better performance.
The NetBurst microarchitecture used a different approach, it attempted to improve performance primarily by increasing CPU frequency, often at at the expense of efficiency. This was a deccision based in the marketing startegy of that era, when higher frecuencies means higher performance for desktop users, using this new architecure the Pentium 4 models could break the 2 GHz barrier in few time. One of key elements in this approach was Hyper-Pipelined Technology, that was significantly longer pipelines than in previous generation of Pentium processors. While longer pipelines are less efficient than shorter ones, they allow CPU core to reach higher frequencies.
Because the first generation of Pentium 4 processors (Willamette core), proved to be performing not significantly faster, and sometimes slower than the fastest Pentium III microprocessors, Intel added more efficiency improvements to subsequent Pentium 4 cores adding: larger size of level 2 cache, faster FSB frequency, SSE3 instruction set, and Hyper-Threading technology.
Latest Pentium 4 systems achieved better performance with these new components, but at that time the AMD Athlon XP models gained marketing terrain with a better micro-architecture. It was so good that at lower frecuencies it could achive better performance than some Pentium 4 models, for this reason AMD implemented the PR rating system.
Today there are more components, and in the same way more variables that could play an important role in the general speed of a system. Programs must be created to use this new technology, plain not optimized code will execute at the same way and you will gain marginal speed only at higher frecuencies.
Equally benchmark programs must be adjusted to measure new technolgy capability. The question is what you want to measure and how to do it?.
clarf
I seem to recall an article I read on intel processors as to speed in tests.
At the time, intel had a chip with 256K internal cache that was blowing their new cutting edge chip out of the water.
So to make the cutting edge chip more attractive to builders, intel took the old chip and reduced the internal cache to 64k.
Also, remember that test software that is not written to support the second processor core will treat the processor as a single core.
At the time, intel had a chip with 256K internal cache that was blowing their new cutting edge chip out of the water.
So to make the cutting edge chip more attractive to builders, intel took the old chip and reduced the internal cache to 64k.
Also, remember that test software that is not written to support the second processor core will treat the processor as a single core.
-
Ralph124C41
- Posts: 4
- Joined: Fri 15 Aug 2008, 03:56
I think the whole thing is a scam. I haven't seen any significant speed improvement for about five years now. They ran out of steam on the frequency scam, and the current scam is to see how many processors they can cram on a chip. The software is all a scam too. XP was supposed to be faster than WIN98, and Vista was supposed to be faster than XP. Windows 7 is supposed to be even faster. It's all a crock. I think the speed is really determined by the hard drive performance, and that hasn't improved significantly either for a decade. If you don't believe me, do a benchmark on a Conner CFA-540 hard drive, circa 1994. Compare its seek time to you new fancy-shmancy Terror-Byte hard drive. And just to make sure nothing is fast enough, they invented the Virus scam. Now, the speed of the computer is determined by how much Norton Antivirus can bog down the entire system. And then there's "Cloud" computing... That gives the scam artists the ability to pin the blame on the Ether...
Puppy Linux forever!
Puppy Linux forever!
Haha, although a bit cynical, Ralph has some points. In fact, I was just talking about this to a friend the other day. My first store bought computer (I built 2 homebrews before that) was a Cromemco Z1 4mhz 8bit system with 64k. It ran CDOS, a modified version of CPM. The entire OS was about 7 or 8k. From a workstation point of view, it booted and ran programs about the same speed as any of my modern systems today even though today's CPUs are 8X the bus width, clocked a 1000X faster, and there is 1000X more RAM. The reason is the software generally expands to consume all available resources. So you get a pretty GUI interface, fancier programs, and so on, but at least from an efficiency standpoint, run about the same as my 1976 system.
[b]Hasten Slowly![/b]
One last thing... Running the Raytracing benchmark indicates 58 seconds on the dual core, whereas the reference Celeron is about 41 seconds (i.e., the dual core is slower). I actually measured this with a stopwatch just to make sure nothing was skewed in software. I also checked with BIOS to make sure the CPU clock was accurate and the bus multiplier was correct. I then launched Htop so I could watch CPU activity concurrently with the tests. On all 6 benchmarks (including Raytracing), BOTH cores immediately go to 100% utilization for the duration of the test. So for anyone wondering whether these aged tests were SMP, the tests DO employ both cores (jduffy, it would be interesting to see if all 4 cores are utilized on your system?).
In summary, the reference 1.5Ghz single core Celeron DOES appear to beat a 2.66Ghz dual core P4 CPU, at least on the Raytracing benchmark. Of course, this may or may not have anything to do with the real world (i.e., real application software). On the surface the discrepancy seems ludicrous, but I suppose issues like cache can become stumbling blocks if the software doesn't take advantage of the methodology. For instance, if the CPU experiences excessive cache misses, then obviously things will proceed much slower. Perhaps even the dual core architecture itself may be partially in the way as multiple cores still share the same bus. It would be interesting if there was a way to shut down one core and try the tests. Does anyone know if this is possible?
In summary, the reference 1.5Ghz single core Celeron DOES appear to beat a 2.66Ghz dual core P4 CPU, at least on the Raytracing benchmark. Of course, this may or may not have anything to do with the real world (i.e., real application software). On the surface the discrepancy seems ludicrous, but I suppose issues like cache can become stumbling blocks if the software doesn't take advantage of the methodology. For instance, if the CPU experiences excessive cache misses, then obviously things will proceed much slower. Perhaps even the dual core architecture itself may be partially in the way as multiple cores still share the same bus. It would be interesting if there was a way to shut down one core and try the tests. Does anyone know if this is possible?
[b]Hasten Slowly![/b]
Fancier with stuff we don't need I might add. It's good to see discussion about real time performance rather than just figures.gary wrote: "The reason is the software generally expands to consume all available resources. So you get a pretty GUI interface, fancier programs, and so on, but at least from an efficiency standpoint, run about the same as my 1976 system.
My fastest PC in actual use was an IBM with a P3 850 and SD133 MHZ Ram running win 98 se. It was supposedly designed to run (and supplied with) ME. Was anything designed to run ME? If that processor hadn't failed I'd be running Puppy on it now.
Recently the mainboard in the computer I'm typing on right now failed. I was fortunate enough to come by a similar chipset mainboard to replace it with so as not to have XP issues with a different chipset. (I hang my head in shame for having bloatware between my between my keyboard and ethernet ports)
Along with the mainboard came a P4 2.8 Ghz to replace the 1.8 GHZ. It does a few things quicker, but not much. The most noticeable improvement is due to USB 2 ports!