Forgot your password?
typodupeerror
Hardware

Comparative CPU Benchmarks From 1995 to 2004 320

Posted by timothy
from the difference-engine dept.
Lux writes "The guys over at Tom's Hardware Guide have been busy recently! They've compared over a hundred different architectures dating all the way back to the Pentium 1 in one huge benchmarking effort. Looking to upgrade an older system? Unlike most benchmarks, which compare modern systems to other modern systems, these charts can help you figure out if the cost of upgrading is worth the speedup or if you should hold off for a bit longer."
This discussion has been archived. No new comments can be posted.

Comparative CPU Benchmarks From 1995 to 2004

Comments Filter:
  • Upgrade (Score:5, Funny)

    by Anonymous Coward on Wednesday December 29, 2004 @01:45PM (#11210588)
    I currently have a 486 [wikipedia.org] with an (upgraded) 900MB hard drive, cdrom drive, and a whopping 32MB of ram. And windows 3.1 + dos. What are my upgrade options?
    • Fixed link (Score:2, Funny)

      by Anonymous Coward
      Whoops, pasted the wrong wikipedia link. The correct one is Intel 486 [wikipedia.org]
    • Re:Upgrade (Score:2, Insightful)

      by stratjakt (596332)
      A 487 math co-processor will turn it into (almost) an equivelant pentium.

      Actually a 486+487 still has enough juice for a homebrewed linux firewall/router, and you can get boards with chips for a buck in the throwaway bin at my local computer shop.
      • The 486 has had a built-in math coprocessor ever since it debuted. After the 486 was around for a while, they made a stripped down version without a coprocessor called the 486SX. The plain 486's were called 486DX.

        You could get a coprocessor for the 486SX, but not the DX. From what I've heard, the original 486SX's were actually re-badged 486DX's whose math coprocessor unit was either not functional or just disabled. When you bought the 487SX "co-processor" you were actually buying a fully functional 486DX t

        • When you bought the 487SX "co-processor" you were actually buying a fully functional 486DX that disabled the other CPU on the board.

          Correct.

          And for the 386, the 386SX was like a 386DX but it was crippled by a 16-bit bus instead of a 32-bit bus. Which made them the same speed as a fast 286.

          Neither 386SX or 386DX had math-coprocessors. You had to add them on later.

        • My wife... (Score:4, Funny)

          by Belial6 (794905) on Wednesday December 29, 2004 @03:07PM (#11211494)
          My wife responded to my description of the SX, DX situation with "So, you want the 486DX, not the 486 Sucks, right?"
      • A 486+487 is still far away from a Pentium.
        Anyway, how will a math coprocessor help firewall or router software? It will help you to play Quake I, but will make no diference for a program that is not intensive on floating point math.
    • So YOU'RE the one hosting that site. Fastest slashdot effect ever.

      -Matt
    • Re:Upgrade (Score:3, Interesting)

      by Zerbey (15536) *
      Put Linux on it, it'll make a perfectly decent home firewall, dns server, web server and mail server. It'll still have plenty of horsepower left over as well.
      • Re:Upgrade (Score:3, Funny)

        by daniil (775990) *
        I fail to see how installing linux on it is an upgrade.
        • Re:Upgrade (Score:3, Insightful)

          by Dysan2k (126022)
          Well, it is if you're running DOS 6.xx + Win3.x or 95. Namely full 32-bit support, native multitasking, better VM; over all, far better tools to work with under Linux than older offerings. Just comes down to what you are wanting to do.

          I still use an old P5-133 w/ 80MB just to do general doc writing and such. I've got it working plenty fast with Mandrake 9.

      • I tried digging out an old 80MHz 486 and put Linux on it...it was soooooo slooooowwwww. 4200RPM (or was it 3600?) hard drive didn't help.

        The thing is, at the time a 486 was current, so was FVWM. Our expectations just aren't backwards compatible.

        • Are you sure? A 486-80Mhz will run slow with a modern GUI such as KDE or GNOME. However, as a server it will run very fast. Never underestimate the power of a 486 :)
        • I wouldn't advise a GUI on an older machine like that. I used to run a 486DX2 66MHz (I think) with 24MB of RAM and an old 512MB hard drive as a web/ftp server. Web pages were served up pretty quick but ftp was a little slow when a client first connected because Linux had to swap some things out to disk. Other than that, it did just fine.

          On the other hand, my current Pentium 166 with 32MB of RAM is an absolute dog when I load up X and Blackbox. I think it's a combination of swap thrashing and slow video (pr
    • When you see better computers than the one you're using at Goodwill/Salvation Army/St. Vincents .. it's time to upgrade.
    • Cool it with liquid nitrogen and overclock the hell out of it!
      • It's probably an urban legend, but i've heard of a guy who overclocked a 386 to run at 300 MHz. For a few seconds before the meltdown.
  • Heat Output (Score:5, Funny)

    by eln (21727) on Wednesday December 29, 2004 @01:45PM (#11210589) Homepage
    It's easy...if the room is getting a little too chilly for my liking, I upgrade to a faster processor. Problem solved.
    • My Athlon 2600+ is good at being a space heater when I run emerge -uD world! But the Thermaltake fan I have on there doubles as a jet engine. 5000 RPM. Does a great job of moving the heat out to the room though.
  • by dreamchaser (49529) on Wednesday December 29, 2004 @01:45PM (#11210593) Homepage Journal
    A /.'ing that I can actually cheer for!
  • by stratjakt (596332) on Wednesday December 29, 2004 @01:46PM (#11210596) Journal
    Artificial benchmarks tend to exaggerate minor differences in speed that aren't noticable or relevant in human time.

    The best analysis of whether you should upgrade is a subjective one. Sit down at the computer. Does it do what you want or not?

    Benchmarks tell me my Radeon 9800 is horribly out of date and imply its too weak to play any modern games. But I know from experience, that's bullshit.
    • *sits at his Macintosh from 1984*

      I can honestly say I don't want to upgrade.
    • Totally.

      I didn't see the Doom3 benchmark anywhere!

    • Agreed. My CPU just turned 5 years old.

      That's right, I'm a slashdotter with an AMD Athlon 550mhz, 384MB ram, and 80gb hard drive (the latter 2 were obviously some small upgrades).

      This does everything I need, why should I spend money on it? It plays music, is fine for my homework, IM, web browsing, and plays *cough* videos *cough* just fine. In the meantime, it's also my server running SSH, Samba, Apache, ProFTPD, etc...

      I'm now at the point where I refuse to upgrade until it melts down on me. It's been t

  • Well (Score:5, Funny)

    by 0racle (667029) on Wednesday December 29, 2004 @01:46PM (#11210600)
    I went from a 486 to a Sempron 2500+. Unfortunatly the artical doesn't go back far enough so I can't tell if it was worth it.
  • No Apples? (Score:2, Insightful)

    by WormholeFiend (674934)
    They must've dropped them on the road on the way to the benchmarking lab. :(
  • Upgrade (Score:3, Interesting)

    by FiReaNGeL (312636) <fireang3l@hotma i l .com> on Wednesday December 29, 2004 @01:50PM (#11210652) Homepage
    Back when I upgraded my 386 16 Mhz, I told myself that I'd upgrade every 10x in performance gain. I upgraded to a Pentium 90 Mhz, then to an Athlon 900 Mhz. It seems that with the recent troubles of AMD/Intel at breaking the 4 Ghz barrier that I won't keep my 'promise' anytime soon, sadly.

    How will they keep their market alive if they can't upgrade the performance? Its not like CPU chips are burning easily anyhow... so why get a replacement if the performance gain is not worth it? (Especially for web browsing / text editing only folks who upgrade based on marketing ONLY... yes! 3 Ghz more will make your internet go faster! Heh)
    • Re:Upgrade (Score:3, Informative)

      Because megahertz is not totally where it's at, contrary to Intel's marketing hype. I'm currently running an Athlon XP-M 2500+ overclocked to 2.4ghz, and it performs faster than a 3.2ghz Pentium 4. An Athlon 64 3200+ runs at 2.2ghz iirc, and outperforms a 3.2ghz p4 by an even larger margin.

      With multi-core chips and on-die memory controllers, the benefits of performance will be felt, even if the clock speed is constrained to 4ghz for now.
      • Re:Upgrade (Score:3, Interesting)

        Are games currently multi-threaded enough for multi-core to have any effect? Do games developers even want that debugging headache?
        • No, but games aren't currently *that* CPU dependant. If you have enough memory bandwidth and video card power, you can run almost any game now.

          Best example is Sims 2. If you have a T&L Video card, you can run the game with a 900mhz processor. If you don't, you need a 2.4ghz processor.
        • Most games are multi-threaded just a little bit, but not enough for multicore to help rather than hurt. They'll certainly consider multicore in next generation titles, but most titles today aren't anything like cpu bound (they're all GPU bound).
      • Because megahertz is not totally where it's at, contrary to Intel's marketing hype.

        Somewhat true, but historically most cpu improvement was due to clock speed increases, which have slowed way down.

        CPUs just aren't getting faster like they used to. We've hit something of a wall. It would be interesting to seem Tom's benchmark charts plotted against release dates.

        The move to multiple cores is a white flag waving. Two cpus aren't in any was as good as one that's twice as fast.

    • Another one of those people who think that clock speed equates to performance. :sigh:

      I'd take a 1 Ghz proc that executes 5 ops/clock over a 3 Ghz proc that executes 1.5 ops/clock, if I were given a choice. Price is also a major determinating factor for me.

    • Re:Upgrade (Score:3, Insightful)

      by jejones (115979)
      Clock speed != performance, no matter what the Blue Man Group might want you to think.
    • If they can't pump up the MHz they'll get you with everything else:

      "Anti-Virus Projection" - AMD & NX Flag (Earlier today)
      "Multi-Core chips"
      "Dual processors"

      etc, etc. The industry will continually find another reason for you to upgrade. And I'll keep doing it, because I like to be on the Bleeding Edge (TM).
      • Re:Upgrade (Score:3, Insightful)

        by wfberg (24378)
        Dual processor is very, very useful. Even for lowly non-server use.

        You see, I have a dual cpu system, and for the longest time I thought XP must be the most stable windows OS evar!

        Turns out the OS never really crashes because there's always a cpu left to bring up the ctrl-alt-del screen with, so you can kill all the OS processes on the other CPU that DID crash..
    • As others have pointed out, MHz isn't the only factor in speed. I don't understand the obsession with speed in the first place - if the box works, use it. If it's too slow to do what you need, upgrade it.

      I went from a 486-66 to a P133 to a P2-266 (doubling clock each time), then a 1.33GHz. I don't use any "rule" based on performance gain - I just upgrade when I have the money, and I have a need. That 1.33GHz is going on 4 years old and does everything I ask of it (thankfully I spec'd it out well enough
      • The reason that we're all obsessed with Mhz is that it's the only performance guage offered.

        If for example, processor makers started guaging their performance in MIPS (or BIPS) or something, it would make more sense I guess.

        (someone will correct me if I'm wrong anyway lol)
        • Re:Upgrade (Score:3, Informative)

          by GlassHeart (579618)
          If for example, processor makers started guaging their performance in MIPS (or BIPS) or something, it would make more sense I guess.

          It would make about as little sense. On many CPUs (particularly the CISC CPUs), instructions take wildly different amounts of time to complete. A NOP might complete in one clock cycle, while an obscure legacy instruction might take twenty. Running only NOPs, the CPU would be 1 BIPS if driven at 1 GHz. Running the other instruction, it'd be 50 MIPS. Somewhere in the middle wou

      • I see you got a dell as well. Congradulations on buying the cheap side of motherboards.

        Actually my Dell never had a probelm. Of course i didn't spec out enoguh RAM, and can't seem to part with the $200 for RDRAM needed to run Doom 3. That and the $75 for the video card upgrade. Ah well, I heard it sucks anyway.

        the trick with MHZ is that Intel reinforced that more mHZ is more speed. Look how many people laughd at Apple for years when the G4 running at 400mhz was out performing 1 ghz intel chips.

        Simpl
    • When we got our first Athlon 1 GHz a few years ago, it was about 3 times faster at floating point operations than the P-3 500Mhz that had been our top number cruncher.
  • turns out they left the pentium 1 in the server, so it's dead now... they'll figure out the cost of upgrading it later.
  • 100 architectures?! (Score:5, Interesting)

    by Jhan (542783) on Wednesday December 29, 2004 @01:53PM (#11210685) Homepage
    Actually, they only benchmark one architecture, x86. A real shame, I would love to see a thorough comparison of *multiple* processor architectures over a long period of time.
    • by Kaellenn (540133)
      While it might be an interesting exercise, I think it's well beyond the scope of this article. The focus is clearly on personal computers, and the (by-far) dominant architecture we've had on our desks over the past 10 years.

      While the slashdot crowd might find such a benchmark informative, the general Tom's HWG user probably would not.
    • by Rebar (110559) on Wednesday December 29, 2004 @02:43PM (#11211277)
      Okay, since Tom's is completely dead at the moment, I'll share the results of my own useless benchmark. I ran the very same single-threaded stupid command on several different machines and recorded user time. Users enter stupid commands anyway, so I figure this is as good as any other test.

      I did say this is useless, right? Good. Note that most of these machines are multi-cpu machines, and it looks like I only did this on Power4, PPC, Intel and AMD, and Alpha systems.

      My stupid command is:
      dd if=/dev/zero bs=32768 count=3200 | time gzip > /dev/null

      Here are the machines and the USER-time result in seconds:

      MHz Secs CPU Arch
      3185 1.04 Intel Xeon CPU 3.20GHz
      3057 1.08 Intel Xeon CPU 3.06GHz
      2795 1.22 Intel Xeon MP CPU 2.80GHz
      2786 1.22 Intel Xeon CPU 2.80GHz
      2395 1.39 Intel Xeon CPU 2.40GHz
      1800 2.00 AMD Athlon 64 Processor
      1533 2.44 AMD Athlon MP 1800+
      1300 3.18 IBM Power4
      1108 3.69 IBM P690 Power4
      1108 3.71 IBM P690 Power4
      1000 4.36 EV6.8CB
      1150 4.4 EV7 21364
      1000 4.79 AMD Duron OC 133FSB
      1000 5.1 EV6.8CB 21264C
      1000 5.37 PIII Xeon Coppermine core
      1000 5.5 PowerPC RS64-IV
      866 5.76 PIII Coppermine core
      700 6.1 EV6.7 21264A
      500 12.36 PIII Katmai core
      600 14.9 EV5
      400 14.99 AMD K6
      350 17.23 Pentium 2
      532 19 EV5.6 21164A
      300 27.14 Pentium MMX
      300 34 EV5

      Due to the Lameness Filter, I can't make the above data any prettier, but I'd bet you can figure it out.

      Of course with differences in OS, compilers, memory speeds, etc. you can't really draw any conclusions from this, EXCEPT this is how fast this particular command runs on these exact systems, AND you can run it on yours to compare how fast a stupid command will finish, which is good to know.

      --rebar

      • Did you not test IBM's 970 (better known as Apple's G5), or am I just blind? I'd like to know where it stacks up....
        • No, I only tested systems available to me for either evaluation or that were production somewhere; the G5 had not made it to the big-iron world at the time (has it yet?), but feel free to run that simple command on your G5 and compare. These numbers are anywhere from several years old to 1 year old, and I don't have access to all the latest-and-greatest hardware.

          Oh, and the AMD64 is running in 32 bit mode; IIRC it was about 20% faster when running Gentoo in 64 bit mode.


      • Hate to break it to you, but gzip is actually a useful benchmark. I spend more time waiting on file compression commands on my old computer than anything else, because I tend to like to tar and compress infrequently used directories. I also re-do gzipped downloads in bzip to save space.
        • I agree, but I figured someone would point out that compressing a stream of zeros is not quite like compressing real data - but at least the stream of zeros is available on (nearly) all platforms, and it is very nearly CPU bound and so a pretty good relative test, for what that is worth.

          I hate waiting on gzip so much that I wrote a multi-threaded program to make gzip files faster on machines with >1 CPU; see http://lemley.net/smp.html if you have more than one CPU and hate to wait on compression program
      • 1.64user athlon64 2000mhz , with some typical desktop apps open(but idle) simultaniously.
  • Recently? (Score:3, Informative)

    by jargoone (166102) * on Wednesday December 29, 2004 @01:53PM (#11210691)
    They haven't been busy recently. They just updated the guide they did quite some time ago. Not very much new to see here...
  • by chaoskitty (11449) <johnNO@SPAMsixgirls.org> on Wednesday December 29, 2004 @02:01PM (#11210771) Homepage
    "Comparative CPU Benchmarks From 1995 to 2004"

    I only see x86 CPUs. What about the PowerPCs, SPARCs, MIPS, Alphas, ARMs, and so on?

    For instance, the m68060 was the first consumer level processor with branch prediction and branch folding, superscalar dispatch, and real-world throughput of more than one instruction per clock cycle. Except for floating point where it performed only modestly, the m68060 seriously outperformed the Pentium in spite of only having a 32 bit data bus as compared with the Pentium's 64 bit bus. Isn't this significant in illustrating the influences in processor architecture?

    http://www.sixgirls.org/ is an m68060 Amiga running NetBSD 2.0. Still very useful after all this time. Where are all those Pentium 60 machines?
    • The reasons those CPUs didn't get benchmarked (aside from the time and effort it would take) is because they really aren't that relevant to the article. No one used those CPUs for any gaming, and gaming is what drives the CPU market. Not business apps, not scientific apps, and certainly not Mom and Dad word processing or desktop publishing their way to fame and fortune.

      Games drive consumer CPU advances, plain and simple, and the CPUs you mentioned had virtually nothing (comparitively) to offer in terms o
      • Not business apps, not scientific apps, and certainly not Mom and Dad word processing or desktop publishing their way to fame and fortune.

        I'm sure IBM and Motorola would disagree with that. I'd bet Apple (who buys from IBM) would disagree with you. I'd bet Sun disagrees with you.

        In fact, maybe the only business that would agree with you would perhaps be HP with their now defunct Alpha product (Started by Digital and destroyed by Compaq).

        This is the problem with /. - pimply faced school-children that t
    • Where are all those Pentium 60 machines?

      Well the best place for them is as a firewall between the broadband modem and the new PC. People are using separate hardware firewalls, right? No? Oh dear...

  • by Mike626 (70084) <injoke.gmail@com> on Wednesday December 29, 2004 @02:01PM (#11210772) Homepage
    Part 1: http://www.tomshardware.com.nyud.net:8090/cpu/2004 1220/index.html
    Part 2: http://www.tomshardware.com.nyud.net:8090/cpu/2004 1221/index.html
  • 8mb card for PCI? (Score:4, Insightful)

    by jensen404 (717086) on Wednesday December 29, 2004 @02:06PM (#11210820)
    Why did they use an 8mb video card for the older motherboards that don't support AGP?

    Matrox Mystique G170
    Memory: 8 MB SD-G-RAM

    They should use the fastest availible video card if they are testing CPU speed. My 200mhz pentium pro with a 16mb TNT card ran Quake 3.
  • My results (Score:3, Informative)

    by freelunch (258011) on Wednesday December 29, 2004 @02:07PM (#11210834)
    The article is a bit slashdotted but it looks like it doesn't go back all that far.

    Just a teaser, I have been running a collection of benchmarks since the Pentium 90.

    At the time, I was involved in a huge UNIX engineering workstation benchmark. I felt we needed something more constant than the applications to compare performance (the engineering apps constantly change). So I quickly assembled everything I could find that could be easily run. These are mostly 'toy' benchmarks, but the results are still interesting.

    For these int benchmarks, higher is better:

    c4.s c4.64 dhry21 hanoi heapsort nsieve nsieve TOTAL
    Kpos/sec Kpos/sec MIPS mvs/sec high High Low
    MIPS MIPS MIPS
    P 90 92.7 94.2 68.6 51.2 43.55 111.0 33.3 494.6
    md64b 4050.1 4167.8 4914.3 2708.8 3333.7 3333.7 610.4 21782

    Float: Higher is better, except for the fft's.
    flops20 fft tfftdp
    MFLOPS MFLOPS MFLOPS MFLOPS TOTAL time time
    (1) (2) (3) (4)
    P 90 13.3 12.8 18.1 23.8 68.0 3.07 16.81
    amd64 1120.9 1004.3 1480.9 1834.7 5440.8 0.04 0.42

    The P90 was running RedHat. The AMD64 is my new desktop, a 90nm 3000 OC'd to 2430 Mhz. My data also includes systems from DEC, HP, IBM, Sun and SGI. I also ran 10 matrix multiply benchmarks as part of the effort.

    I have never gotten around to publishing the results or the collection of benchmarks.. Maybe it is time.
  • But... (Score:5, Funny)

    by Junior J. Junior III (192702) on Wednesday December 29, 2004 @02:07PM (#11210837) Homepage
    I bought my computer in 1982... how will I know if it's worth upgrading if the data only goes back to 1995?
  • Exciting? (Score:5, Informative)

    by SpinningAround (449335) on Wednesday December 29, 2004 @02:11PM (#11210883)
    As I recall, they claimed that part one of the article took something like 300 hours to put together. Seems like a lot of work to tell me that processors have become a lot faster in the last 10 years.


    Actually I shouldn't give Tom's Hardware a hard time (like everyone else seems to). As articles go, the reviews of high-end ink-jets [tomshardware.com], the 8-channel RAID6 card [tomshardware.com] and the Viewsonic media center [tomsnetworking.com] were quite interesting (and a lot more recent than the CPU round-up too).


    These days though, my favourite reviewer is Dan [dansdata.com] (who posts here now and then). Dan seems to understand that a million graphs showing you the statistically insignificant difference between the latest mobos / graphic cards / processors / ram sinks don't really make a great site.

    • well, i agree that tom isnt very hot regarding reviews (anand is better, for special stuff of course storagereview or beyond3d rule), but i wouldnt say dan is a reviewer.
      He just takes every shit he gets for free and tells something about it, which is either "its cool" or "it sucks", only with more words.
      Every 2nd article is eiter a beggin for money (he actually posts news items "please send me money via paypal") or a blatant product placement ("click here to get those GREAT photon lights for so much cheape
  • but I'm sure that I read a similar article at tom's this year comparing x86 CPUs from ~386 to recent systems.

    IIRC they decompressed zip files and encoded video stuff among other things. It's quite impressive to see a fairly modern CPU performing the same task in minutes which used to take several hours on one of those 'antiques'.
  • If you installed Intel's PII Overdrive for Socket 8 mommyboard you could obtain MMX, but that was the only way.
  • by Renraku (518261)
    I've got a Duron 1200 and am trying desperately to upgrade to an nForce2 / XP 3200+.

    After a bad board and bad processor (from partspc!) blew a RAM chip I had bought for it, I ended up RMAing all three of them, likely with little or no compensation coming my way.

    So here I am, after Christmas, stuck with one functional 512 stick of PC3200 Corsair after have spent $400 or more on my upgrade fest.

    That's why I don't upgrade very often, besides the lack of money. Because I know that there's a good chance what
  • People don't really detect a perceivable performance boost for less than 25% more actual performace. And the perception of that 25% varies wildly from 'fneh' to 'wowieeeee'.

    But I'm talking to a bunch of gamers for whom that last 0.4% performance boost is worth more than a hot cheerleader full of X.

    • But I'm talking to a bunch of gamers for whom that last 0.4% performance boost is worth more than a hot cheerleader full of X.

      These are the same people that get into squabbles over which display adapter is better when A gets 150FPS (frames per second) and B gets 140FPS but with better double-dodecahedron rendering or whatever. It makes no difference to them that most human eyes cannot distinguish between 40FPS and 80FPS, let alone anything above 100. However, if the [abstract and wholly meaningless n
      • I must disagree with you on the "most human eyes can't distinguish the difference between 40 and 80fps." This may be somewhat more true of movies which are made with motion blur that blends the frames together. But you can certainly still tell the difference for the better when a movie is captured at 60fps rather than 30. Its not as big a difference as changing the framerate in a game is due to the motion blur but its certainly more visually pleasing when a video is captured and played at a higher framerate
    • Actually, a worthy performance boost should be an order of magnitude (10x).

      I don't upgrade until there's a computer 10x better than the one I have right now, and available at a good value.
  • The primary end-result of the evolution of this commodity hardware is the fact that expensive software is now just obsolete - plain and simple. Ten years ago, there was a justified price premium associated with state of the art software algorithms. I still see these zealots for the DB companies raising these red-herring issues as to why every organization should still spend hundreds of thousands of dollars on expensive DB software. With the evolution of hardware the way it is, any credible SQL DB Engine co
  • If you take a look at the big chart, why is it that the clock speeds have increased by ~10 fold from start to end, but the front side busses have only increased ~5 fold?

If I'd known computer science was going to be like this, I'd never have given up being a rock 'n' roll star. -- G. Hirst

Working...