can you compare a new cpu to a 20 year old one?

[tommo123]

another thread had me thinking on this. my 1st pc had a 166mhz pentium (not sure if p2 or not). how much faster in % terms would a 3.4ghz IB cpu be? (3570k).

i mean if you make it 3400mhz * cores / 166 then it's 82 times faster but that doesn't include any overhead (multiple cores don't scale perfectly no?), or architecture improvements.

 

[Eeqmcsq]

Correct. Clock speed calculations don't factor in architecture improvements. Whenever I want to compare a modern CPU to an older one, I look them up on passmark's list of CPU benchmarks. It gives you a nice rough estimate on how much faster (including multiple cores/threads) the modern CPU is compared to the older one. It's also gives you an idea on where the low powered CPUs (like the AMD E-350) rank in performance compared to desktop processors.

 

[Idontcare]

another thread had me thinking on this. my 1st pc had a 166mhz pentium (not sure if p2 or not). how much faster in % terms would a 3.4ghz IB cpu be? (3570k).

i mean if you make it 3400mhz * cores / 166 then it's 82 times faster but that doesn't include any overhead (multiple cores don't scale perfectly no?), or architecture improvements.

Not quite a 20yr difference, but I compared a PII 450MHz to an FX-8350 in an application (the exact same application) I used spanning the timeframe of the two.

The thread is here, and the specific post I am thinking of is here.

In a normalized setting, the 8-core FX-8350 at 4.5GHz is 141x faster than the single-core PII 450MHz.

 

[Enigmoid]

Not quite a 20yr difference, but I compared a PII 450MHz to an FX-8350 in an application (the exact same application) I used spanning the timeframe of the two.

The thread is here, and the specific post I am thinking of is here.

In a normalized setting, the 8-core FX-8350 at 4.5GHz is 141x faster than the single-core PII 450MHz.

Interesting.

Whats the power use of the pentium 2? (would like to see the raw efficiency increase).

 

[Idontcare]

Whats the power use of the pentium 2? (would like to see the raw efficiency increase).

I have no idea on that, did not measure power when I had the PII

 

[SpeedTester]

Interesting.

Whats the power use of the pentium 2? (would like to see the raw efficiency increase).

I believe it was 25-35 watts. I still have some slot 1 and slot a processors stored away somewhere.

 

[Yuriman]

That doesn't take into account specific instruction sets though, does it? For instance you can use AVX in many tasks on a modern chip, which should result in orders of magnitude more speed.

 

[SPBHM]

I tested a few days ago my Pentium MMX at 237, it run super pi 1M in 10 minutes and 1 second, and "cpumark 99" had a score of 9.45, a newer CPU will probably finish super pi 1M in 10s or less, and CPU mark 99 over 500.

but obviously this completely ignores new instructions and multiple cores..

 

[jhu]

Not 20 years, but I still have a P!!! 450 MHz laptop (from 1999). Compiling Povray 3.6 (which is single threaded) for each processor type and running the benchmark, here's what the results are:

Core i5 3317U (1.7 GHz):      199.26 pps ; 117.21 pps/GHz
Core i5 2400S (2.5 GHz):      235.18 pps ;  94.07 pps/GHz
Celeron 220 (1.2 GHz):         81.15 pps ;  67.62 pps/GHz
Athlon II x4 (2.8 GHz):       179.82 pps ;  64.22 pps/GHz
Pentium !!! (450 MHz):         12.43 pps ;  27.62 pps/GHz

Ivy Bridge core itself is about 4x faster per clock than P!!!.

I haven't run Povray 3.7 (which is multithreaded, haven't had time to test) on the P!!!, but multiple instances of Povray 3.6 can be run to render different areas of an image, which can then be stitched together. Speed of rendering scales linearly with each additional core. With this in mind here's what we get using this method:

Core i5 2400S (2.5 GHz):      940.72 pps ;  376.28 pps/GHz
Athlon II x4 (2.8 GHz):       719.28 pps ;  256.88 pps/GHz
Core i5 3317U (1.7 GHz)*:     398.52 pps ;  234.42 pps/GHz
Celeron 220 (1.2 GHz):         81.15 pps ;   67.62 pps/GHz
Pentium !!! (450 MHz):         12.43 pps ;   27.62 pps/GHz

* Hyperthreading not taken into account
** Povray (3.6 and 3.7) doesn't utilize vector instructions to their full potential, which would require a major code rewrite I think.

So a Core i5 2400S is at least 75x faster than P!!! 450. A Core i7-3970X should be at least 158x faster than the P!!! 450 (more if hyperthreading is taken into account).

 

[Idontcare]

Ivy Bridge core itself is about 4x faster per clock than P!!!.

That is impressive! Increasing IPC by a factor of 4x alone is crazy. I guess we have some idea of where those hundreds of millions of transistors are going.

P!!! had a silly small number of xtors compared to an IB core.

 

[jaqie]

and the pentium 3 was and still is very fast for its clock speed. I have a dual tualatin p3 1.4GHz system here that is quite able to be a good desktop machine. I also have several coppermine P3 1ghz systems and for basic tasks only they are also quite capable, and run win7 acceptably. That is quite a feat for as old as they are.
I actually just got an IBM x41 tablet which has a pentium M 1.5GHz which is a tualatin modified for higher FSB and DDR2 use and I am using it as my main laptop. Besides being single core and having a terribad video card it is quite snappy, and running win7.

Don't forget p3 used SDR pc100 or pc133 ram, that was the main bottleneck to it performing better!

The pentium 3 was one hell of a CPU, nomatter how you slice it.

 

[bononos]

.....
The pentium 3 was one hell of a CPU, nomatter how you slice it.

As I understand it, the P3 evolved into the Pentium M under the Israeli design team who also put out the Core architecture cpus. So current Intel cpus do owe quite abit to the P3.

I wonder what happened to the P4 team, were they censured, disgraced and put into cold storage?

 

[jaqie]

You are correct for the P3 -> Pm -> C2. They did also take some of the P4 (netburst) and put it into the i series, though, so that got the best of both (though IMO hyperthreading is not good, in many people's it is, so I give credit here)

My coppermine P3 systems give like 256MB/s bandwidth from memory, the dual tualatin (and a few single ones) got 512MB/s or so (from two memory controllers akin to dual channel today), and the P mobile gets 2GB/s or so... performance of the p3-ish cpu has been shown to scale with that, in my personal experience as well.

 

[Idontcare]

I wonder what happened to the P4 team, were they censured, disgraced and put into cold storage?

The "team" did what they were asked to do.

Had they failed in doing what they were asked to do then you could start wondering about the repercussions of those failings.

They succeeded, but I don't think the end-result was quite what the champions of the program had envisioned. So the question really ought to be "what happened to the risk-takers who decided to take risk in launching the netburst initiative?"

And if you think about American business mentality the one thing that is repeatedly driven home is "we need more people who are willing to take risk". When everyone becomes a Wally (dilbert reference) and just worries about covering their own neck to save their own job, that is when the company dies from lack of innovation.

So one would hope Intel rewarded the risk takers, yes the final product was a bit of dog (prescott at least), but the mindset of being willing to do something different is something that needs to prevail. Albeit redirected towards taking risk in a different vector (power reduction).

AMD slaughtered its risk-takers, lambs to the butcher style. Don't expect too many Dilberts or Tinas to be willing to stick their neck out on projects going forward. Those who remain at AMD are still there because they knew well enough to keep their mouth shut. And that is exactly what they have going forward, a scared-shirtless workforce who doesn't want to say boo because they don't want to be on the list for the next round of layoffs.

 

[sm625]

I have a windows 95 machine with a 66MHz cyrix. It is almost exactly 20 years old. If you take a brand new intel chip with DDR3 motherboard and you attach that ancient hard drive and install windows 95 on it, the computer still wont feel very much faster.

 

[jaqie]

actually, it will, if you do a few tricks to help it. How do I know? see my sig. I play with things like that as a matter of course.

 

[bononos]

I'm sure they were many successes for the engineering team but as a whole, it looked like Intel was swimming against the tide throughout the lifetime of the P4. There were plenty of booboos to go around even before Prescott like the slow speed of the debut chip and the the rambus mistake.

It could have been a rushed product because the P3 was getting slaughtered by the Thunderbirds and Intel needed something new out there fast.

 

[lagokc]

The "team" did what they were asked to do.

Had they failed in doing what they were asked to do then you could start wondering about the repercussions of those failings.

They succeeded, but I don't think the end-result was quite what the champions of the program had envisioned. So the question really ought to be "what happened to the risk-takers who decided to take risk in launching the netburst initiative?"

And if you think about American business mentality the one thing that is repeatedly driven home is "we need more people who are willing to take risk". When everyone becomes a Wally (dilbert reference) and just worries about covering their own neck to save their own job, that is when the company dies from lack of innovation.

So one would hope Intel rewarded the risk takers, yes the final product was a bit of dog (prescott at least), but the mindset of being willing to do something different is something that needs to prevail. Albeit redirected towards taking risk in a different vector (power reduction).

AMD slaughtered its risk-takers, lambs to the butcher style. Don't expect too many Dilberts or Tinas to be willing to stick their neck out on projects going forward. Those who remain at AMD are still there because they knew well enough to keep their mouth shut. And that is exactly what they have going forward, a scared-shirtless workforce who doesn't want to say boo because they don't want to be on the list for the next round of layoffs.

Bulldozer's huge changes in architecture were a pretty big risk. Do you have knowledge that the people responsible for those changes were the ones AMD laid off or are you guessing here?

 

[jhu]

That is impressive! Increasing IPC by a factor of 4x alone is crazy. I guess we have some idea of where those hundreds of millions of transistors are going.

P!!! had a silly small number of xtors compared to an IB core.

Some of those gains are no doubt due to the presence of faster memory and more cache. If anyone can suggest useful FP operations that can fit in 16KB + 16KB of code and data (P!!!'s L1 cache), that might help isolate things better.

 

[SPBHM]

As I understand it, the P3 evolved into the Pentium M under the Israeli design team who also put out the Core architecture cpus. So current Intel cpus do owe quite abit to the P3.

Banias (first Pentium M) had many evolutions in comparison to Pentium 3 (Tualatin), it adopted the quad data rate FSB from the P4, more l2, SSE2 and other enhancements, later it evolved to Dothan, Yonah (Core Duo)...

Pentium 3 was also a direct evolution from Pentium II and Pentium Pro... so P6 aged quite well I guess.

 

[tweakboy]

No you cant. It would take that old machine like couple days to a week to Export video into avi from Premiere or Vegas

Now it would take that same project 45 seconds to render. gl

 

[aigomorla]

my cely 300A was the best bang for the buck cpu i ever had..

nothing beats it... absolutely nothing!

u set 1 dip switch and poof! your flying at 450mhz!

 

[tommo123]

Not quite a 20yr difference, but I compared a PII 450MHz to an FX-8350 in an application (the exact same application) I used spanning the timeframe of the two.

The thread is here, and the specific post I am thinking of is here.

In a normalized setting, the 8-core FX-8350 at 4.5GHz is 141x faster than the single-core PII 450MHz.

nice info, ta

now i won't go asking the next question i'm thinking of ... how much faster are PCs today versus my 1st computer (pre pc) - an amiga 500+

i think their software was optimized to hell though

 

[IntelEnthusiast]

Wow comparing my old Intel® Pentium® 75MHz processor vs. my new Intel Core™ i5-3570K isnt even close. I can do so much more with this system so much faster that isnt even close.

 

[Exophase]

That is impressive! Increasing IPC by a factor of 4x alone is crazy. I guess we have some idea of where those hundreds of millions of transistors are going.

P!!! had a silly small number of xtors compared to an IB core.

I'd love to see an in-depth comparison using perf counters. 4x seems too high to represent an average over many tests. I'd expect something more like a 2x, maybe 2.5x average advantage. But that'd be with the same binaries. In this case, even if auto-vectorization only plays a modest improvement, the new CPU still benefits from being able to use scalar SSE over x87. That can help the compiler out a lot.

Could also see some higher than typical improvements for anything very memory bound, especially memory bandwidth bound - those old Katmai P3s had the weaker off-chip L2 caches, a pretty small DTLB, and not a lot of aggregate bandwidth (over a non-integrated memory controller). Around 1.77 MB/s/MHz at peak clocks vs 6-7 for the IB system. And the IB is vastly more likely to utilize it.

Really have no idea what PovRay's code is like.. Am hesitant to make blanket assumptions based on a naive idea of how raytracing works..