Another perspective on CPU value

As you might have already seen, the value section of our grand Athlon II vs. Core i3 showdown article includes performance-per-dollar graphs based on whole system prices. We began including these numbers some time ago at the behest of a readers who rightfully feel that, if they need to buy a whole PC to begin with, then processor pricing is only part of the equation. For example, if processor A costs $200 and only performs 50% better than processor B, which is priced at $100, then the prospect of paying double for 50% extra performance may not seem that worthwhile. In a $900 system, however, paying an extra $100 for 50% of extra performance suddenly becomes far more enticing.

We changed our full-system formula slightly this time, factoring in all the components you’d need to build a workable PC, including a case, power supply, and optical drive. Picking the right components proved to be somewhat of a challenge. In the end, we settled on parts we thought representative of a system your average enthusiast might buy: a new-ish motherboard with USB 3.0 connectivity, 4GB of RAM (or 6GB for the Core i7-900 config), a Radeon HD 5770 graphics card, a 640GB hard drive, a DVD burner, and a good quality case-and-power-supply bundle. Those parts pretty much correspond to the middle-of-the-road Utility Player build from our system guides.

What happens if you’re not shopping for a middle-of-the-road system, though? What if you’re on a tight budget and really have to cut corners? How do the performance-per-dollar CPU rankings end up looking then? We were curious to find out, so I fired up Excel, entered the low-end components below into our giant spreadsheet of doom, and watched the rankings rearrange themselves.

Platform Total price Motheboard Memory Common components
AMD 785G $410.90 Asus M4A785-M

($79.99)

2GB Kingston DDR3-1333

($56.99)

HIS Radeon HD 5670 512MB graphics card ($89.99), Western Digital Caviar Black 640GB hard drive ($74.99), Samsung SH-S223L DVD burner ($28.99), Antec Three Hundred with 430W PSU ($79.95)
Intel G43 $410.90 Asus P5G43T-M Pro

($79.99)

Intel H55 $420.90 Gigabyte GA-H55M-S2H

($89.99)

Intel X58 $545.90 Gigabyte GA-EX58-UD3R

($188.99)

3GB Crucial DDR3-1600

($82.99)

For our low-end component choices, I went with very affordable, micro-ATX motherboards with integrated graphics on all but the X58 platform. I stuck to 2-3GB of RAM, and I replaced our graphics card with a $90 Radeon HD 5670, since having a decent discrete GPU has uses beyond extreme gaming. Also, I opted for a cheaper case-and-PSU combo priced just under $80. All of this brings costs down quite significantly: minus the CPU, our AMD platform now costs around $411, down from almost $609 in our original set of numbers. Even the X58 platform has tumbled from $759 to $546.

And here are the results.

The points in our scatter plots follow almost the same pattern as in our big roundup, but all of a sudden, budget-friendly processors like quad-core Athlon IIs score much better in our performance-per-dollar rankings—no great surprise, perhaps. Interestingly, the Core i5-750 and Phenom II X4 965 remain neck-and-neck in the top two spots, albeit with their positions switched. It’s tough to beat Intel’s cheapest quad-core Nehalem, especially once you start taking power into the equation, as we’re about to do.

Again, very little is new under the sun when it comes to our scatter plot. Looking over to our power-efficiency-per-dollar bar graph, we see the Core i3-530 and Core i5-750 have switched places, with the former now claiming the gold medal. The Athlon IIs have climbed up the rankings a little, but they still pale in comparison even to the old Core 2 Quad Q9400. AMD has higher load power consumption pretty much across the board, and no amount of fiddling with value numbers is going to change that.

I think we can draw two conclusions from this little exercise. One, quad-core Athlon IIs are very clearly the best deals if you’re building a low-end rig and don’t care too much about power efficiency. Two, even in a low-end rig, Intel’s Core i5-750 still delivers by far the best mix of performance per dollar and power efficiency per dollar, at least among the processors in our test suite. If you can afford to plunk down $200 on a CPU, you can’t go wrong with it.

Comments closed
    • wira020
    • 10 years ago

    I use a cheap processor in fairly expensive systems.. and cheap motherboards too… why?.. because when a new processor come out, i wont have to think too much to upgrade it… i get the best gpu my money can afford so i wont have to upgrade it when a more demanding games come out.. top of the line gpu should be able to survive a few generations of gpu…

    But that’s just my preferences… and that is why i suggest you could just factor out the common components and leave that to the readers to decide… if not, at least you guys could also include another graph with common components factored out… different people, different preferences…

    • Bensam123
    • 10 years ago

    I would actually stick to just the memory, processor, and motherboard.

    Adding in all the other components only works to clutter results. In processor reviews all you’re supposed to be reviewing is the processor and the price proposition should be based off of it.

    Just the motherboard, processor, and memory gives leeway for systems that do change based on a persons needs, but still represents the area of the market the specific processor is in and how it should be priced out compared to it’s competition.

    The only real difference between processors most of the time is those two things. There really isn’t any reason to include anything else unless you guys are working on a system building guide.

    • oldDummy
    • 10 years ago

    To make a long story short;
    Upon finding out that one of the motherboards I was experimenting with maxed out at 65W and the other at 95W changing the CPU from a QX9650 to an E8600 was required.
    .
    §[<http://www.sudhian.com/index.php?/forums/viewthread/106914/<]§ while not the intent going from the 3G quad to a 3.3G C2D shows exactly how much dual cores suck compared to Quads if your doing any multitasking. Running Prime95 and Vantage P concurrently caused the C2D to stutter and lose frames, enough to notice it. When I started with the quad that wasn’t the case at all. My point is that price and sometimes cores matter.

    • mdk77777
    • 10 years ago

    While I agree with the general results, It would be interesting to see a metric not based on Cinebench.
    It is based on an intel compiler, that has been shown to contain code which intentionally cripple AMD CPU. Would a compiler neutral test suite show a huge difference? Perhaps not, but it would be interesting to see the results. My point is that the level of SSE optimization used by Cinebench might yield only a small advantage to INTEL, but that error(if you like) is compounded when you start to layer value, power consumption, and value per power consumption on a flawed test.

    • dale77
    • 10 years ago

    How about flipping the x axis so that 0 price is on the right.

    This would have the best value processors in the top right where they belong.

    Have the “better” metric ascending on both scales.

    • opinionated
    • 10 years ago

    Given that on Newegg the Core i7 860 is about $90 more than the Core i5 750, and I believe I read somewhere that the i7 860 performs about the same as the i7 920, could we assume the Core i7 860 would be plotted around the coordinates ($700, 460%) on the Overall Performance per Dollar (system price) plot?

    Although, looking at the position of the i7 870, it seems like the i7 860 should be plotted a little higher than that.

      • alex666
      • 10 years ago

      The 860 is almost $300US cheaper than the 870 even though the 860 is rated at 2.80GHz while the 879 is 2.93. Why .13GHz is worth $300 is beyond me, but the 860 should have a much better price-performance ratio. BTW, I’ve got an 860 modestly OCed to 3.60 and I love it. I thought the 870 was way overpriced.

    • crochat
    • 10 years ago

    I don’t agree with adding discrete graphics to a budget system when good integrated solutions are available for none gaming use. Also discrete graphics has a big impact on efficiency numbers.

    • crochat
    • 10 years ago

    I think efficiency results are strongly affected by discrete graphics, and in my opinion discrete graphics doesn’t make sense for a budget computer.

      • indeego
      • 10 years ago

      Intel’s massive market share disagreesg{<.<}g

        • travbrad
        • 10 years ago

        Actually, Intel’s massive market share AGREES, since all of their graphics are integrated, not discrete.

        There’s a million different configurations one could put together though. You can’t expect TR to do that many combinations.

          • Voldenuit
          • 10 years ago

          l[

            • JustAnEngineer
            • 10 years ago

            I am a proponent of the “trickle-down” theory of computer upgrades. “Mom & Dad PCs” are mostly the parts of my “Gaming PCs” from three or four builds ago.

            The current iterations are running AthlonXPs at 2+ GHz with 2 GiB PC3200 DDR memory and Radeon 9800Pro/XT AGP graphics under Windows XP. If one of those fails, it’s a terrific excuse for me to upgrade again.

    • cheapbastard
    • 10 years ago

    For a cheapbastard this is how you put a reasonable system together
    now this is still an expensive system (I’m still using “premium” components)

    BIOSTAR A785G3 AM3 $59.99
    2GB Kingston DDR3 $56.99
    Seagate 500GB $54.99
    samsung cd/dvd $19.99
    case+psu $46.99
    total $238.95

    Now choice of CPU makes much more sense

    I think putting a cheap processor in an expensive system makes no sense
    and I think TR may have forgotten how to put a cheap system together

      • Voldenuit
      • 10 years ago

      cheapbastard you are my hero.

      • FuturePastNow
      • 10 years ago

      Now that right there is a mom and dad computer.

        • hermanshermit
        • 10 years ago

        Nope, Mom and Dad computers are retired early core 2 duos from the company for $50 a go, $80 with 17″ LCD.

        Really if you don’t game or encode video, anything post Pentium 4 (actually mostly for the power and heat) isn’t going to make much of a difference. That’s why netbooks are such a success.

        I seriously don’t think it’s worth bringing such low end hardware new into the world when there are already many such systems around with basically the same performance.

      • opinionated
      • 10 years ago

      What case and power supply are you using here? I’m a bit concerned about the quality of such a cheap power supply.

        • cheapbastard
        • 10 years ago

        it’s the HEC 6K60BSX585 @ ne

        • MadManOriginal
        • 10 years ago

        I would be a little concerned about the PSU as well but adding a solid PSU like a ~$40 Seasonic and a cheap case for $30-40 (if you can’t find a deal) would add about $30-40 to the total. A significant percentage perhaps but not a huge absolute amount.

          • Rakhmaninov3
          • 10 years ago

          Yeah you really don’t want to compromise on PSU or RAM quality–they can cause a whole boatload of weird problems with a system. Pay a smidge more and get a reliable brand.

    • mongoosesRawesome
    • 10 years ago

    This is great! Thanks for the update.

    • blitzy
    • 10 years ago

    Would there be any chance of getting your excel sheet so we could put in our own price figures? For those of us not in the US the prices can differ quite a bit.

      • MadManOriginal
      • 10 years ago

      It’s really not that hard and you don’t need any raw data just work off the CPU-only chart on page 17 of the original article. See post #22.

        • blitzy
        • 10 years ago

        thats true i never thought of that 🙂

        although it would be good to have the sheet so we dont all have to make our own one

        • blitzy
        • 10 years ago

        I tried making a quick scatter chart with prices in NZD, but I had the following problems:

        Its actually tricky to be accurate taking the numbers from the original charts Y axis, because they go up in 50’s and its difficult to guess the exact performance number. E.g. I might look at it and guess 458, when really its 415

        The other problem was I couldn’t find how to make a scatter chart with the axis the same way Cyril has in his charts, in either Excel or Google docs both ways either I couldn’t figure out how or it doesn’t do it.

        §[<https://spreadsheets.google.com/ccc?key=0Ag5Q802KQ3vNdHJfaV8wYXlPUXd2VlRGaFJxRTY1V1E&hl=en<]§ Heres my document if u wanna see the data. It did get me thinking though, if I got a Q9400 and put it into my existing mobo that would be another reasonable option thats not a total pc upgrade.

          • MadManOriginal
          • 10 years ago

          Yes well it wouldn’t be exact, I wasn’t suggesting making an actual scatter plot either 😀 What really matters if you’re willing to agree with the overall price/performance metric that TR uses is the relative ranking. So the first scatter plot on this page: §[<http://www.techreport.com/articles.x/18448/17<]§ that is for CPU only is the place to start, you take the relative rankings and add the rest of the system to determine system price/performance value. The total system price is something you have to determine based on your local prices, deals, and what you might be able to reuse from an old system but at least with relative performance rankings you can estimate whether one system is clearly a better value. It's all personal too depending on absolute price and budget and how much 'more performance' is worth. Making the raw data for the chart available would be good if you have to get it exact, I guess my point was that you can get 'pretty close' with what's available and pricing is always going to vary a bit locally plus there are x-factors like overclocking or unlocking cores.

          • lammers
          • 10 years ago

          I edited your chart and sent it back to you to view let me know if you have any questions.

    • curls
    • 10 years ago

    yay! i5-750 #1! love my pc

    • Dposcorp
    • 10 years ago

    Excellent Job Cyril, although one thing it is hard to factor is prices going up and fluctuating even over the last 6 months on existing products.

    Do you guys know what is going on with the i7 920 pricing?
    Also, some of the 3x2GB kits of ram are wacky, but I know prices went up for all ram.

    I built a 920 system 2-3 months ago, between thanksgiving and Xmas, and it seems like pricing went up since then.

    I picked up a i7-920 from Microcenter for $199, a MSI X58-M Micro ATX board for $150 after rebate, and 3 x 2GB sticks Of G.skill ram for around $100, but that was in about August or September.

    Now I see all that stuff at Newegg for $630 plus shipping.

    I was also seeing a lot of $425-$475 used I7-920/Asus X58/6GB ram combo in the forums at that time as well, but not anymore.

    It is just interesting that I can use my combo for 3 months and sell it and make a good amount of money.

      • Bauxite
      • 10 years ago

      Interesting, your board combo and timing are almost exactly the same as mine, august-ish.

      I actually felt dumb for not waiting, because I assumed surely with w7 and a cpu refresh around the corner combined with vendors everywhere hurting for sales surely everything was going to drop.

      Boy was I wrong…I ended up buying at nearly the best possible time in the last year+

        • Dposcorp
        • 10 years ago

        Yeah, sounds like we both had good timing.
        Also, mine ended up being a D0 and is currently running @3.5Ghz at 45 idle, 60 load.

    • axeman
    • 10 years ago

    No surprise here – when it comes to Intel, cost versus performance is always a case of rapidly diminishing returns. Of course AMDs situation is different, since they need to keep rock bottom prices to remain a viable option.

    • axeman
    • 10 years ago

    This is a mistake:

    l[

    • MadManOriginal
    • 10 years ago

    Thanks for the second look. I do think people protested too much about the specific components and pricing out of laziness. All one has to do is look at the overall performance/dollar graph and shift their favored system build total price left or right along the X-axis since the Y-axis relative performance will remain the same. A little simple logic and understanding of what underlies the graph rather than just take the scatter plot as an unchangable whole allows one to make a judgment for their particular needs.

    • dpaus
    • 10 years ago

    We standardized on the Phenom II platform(s) a little over a year ago for client systems. At the time, the nominal competition was the Core i7 line, as the i5s were just a rumour. We’ve been incrementally updating the base platform over the intervening months, moving from AM2+ to AM3 motherboards, currently standardized on the 965, and now preparing to start deploying Thuban 6-core CPUs (yes, including upgrading some of the original systems).

    One of the factors in our choice was the flexibility that AMD has shown; while the run of the LGA775 was impressive, we were even more impressed with the run we got out our our s939/s940 systems (except for the FX-74 “4x4s” – let’s just not talk about those, OK?), and it looks like we’re going to see a repeat of that with the AM3 platforms; all the way from quads to 12-cores with 5-minute, in-the-field upgrades.

    • UltimateImperative
    • 10 years ago

    What would be insanely nice would be to have an interactive thing to look at the value graph, that would allow you to set the system prices, which benchmarks to include/exclude, &c.

    That would enable you, for instance, to look at performance/$ for systems with integrated graphics, without taking gaming benchmarks into account.

    • rhema83
    • 10 years ago

    Great addition to the already superb Athlon II vs Core i3 article, Cyril!

    To sum it up in two lines…

    1) Using “good enough” components (such as a micro-ATX motherboard) improves performance per dollar with negligible penalty

    2) Core i5-750 and H55 are the mainstream king of value

      • just brew it!
      • 10 years ago

      Considering how much is integrated onto the motherboard these days, I agree that mATX is “good enough” for a large percentage of usage scenarios. With gigabit NIC, reasonable audio, 4+ SATA ports, and 6+ USB ports now standard even on inexpensive boards, most people just don’t need all those expansion slots any more.

      5+ years ago, I would not have considered a mATX mobo, unless it was for a cheap secondary/kids’ system. Today, my primary desktops at home and work are both mATX based.

        • Voldenuit
        • 10 years ago

        Yeah, the last mATX mobo I had (an ASUS P5EVM HDMI, iirc) had 6 SATA ports, 1 PEG, 1 PCIEx1 and 2xPCI. Unfortunately, the Atheros NIC (damn you, ASUS!) was buggy enough that I ended up sacrificing one of my PCI slots for an external ethernet card (the other was already obstructed by the GPU cooler).

        • eitje
        • 10 years ago

        just wait until you begin to realize the options available on mITX! 🙂

          • JustAnEngineer
          • 10 years ago

          Newegg’s only got the one (DFI) LGA-1156 mini-ITX motherboard. I’m waiting for the Zotac H55-ITX to finally show up.

            • eitje
            • 10 years ago

            picked up the DFI last week. I’ll be posting thoughts and comments in the forum sometime this or next week, I’m sure. 🙂

            • MadManOriginal
            • 10 years ago

            I’m hoping for some ITX mobos that have robust power sections and can handle the higher current CPUs when overclocked. The DFI is nice but seems a touch limited in oc’ing, if their past is any indication DFI can do solid power sections on small boards.

    • FireGryphon
    • 10 years ago

    l[

      • Rakhmaninov3
      • 10 years ago

      If you’re payin’ twice as much, you’d be hoping to get close to twice the perf (200%) instead of just 150%.

    • codedivine
    • 10 years ago

    I think one hard-to-quantify variable is upgradeability. AM3 will be compatible with bulldozer AFAIK but we know nothing about what will happen to LGA1156. If you are building a budget system, chances are that you are not the kind of person who replaces your motherboard every year.

      • Firestarter
      • 10 years ago

      Chances are that you aren’t going to replace the CPU either.

        • Voldenuit
        • 10 years ago

        No, but you might replace your CPU every 2 years instead of getting a new system.

        Lots of s939 users stuck with their platform for 3-4 years, especially when s939 Opterons became cheap. I had an A64 3000+ that I happily upgraded with an Opteron 165, and it served me for many years.

        LGA775 was an unprecedented stint for intel, who usually goes through sockets like a fashionista discards shoes, but since buying into it early involved getting a Pentium 4, it’s not exactly the poster child for value.

        That said, Llano will require a new socket, so AM3’s days are probably numbered.

    • JustAnEngineer
    • 10 years ago

    Presumably, you selected the M4A785T-M rather than the M4A785-M. The latter uses DDR2 memory rather than DDR3.

    • KGA_ATT
    • 10 years ago

    AMD is taking a hit in this area(power efficiency) with many of their CPU’s.

    As far as their flagship desktop processor the Phenom II X4 965, they have already stepped from 140w Thermal Design to a 125w Thermal Design. Perhaps it can go further? (edit:along with additional power management changes)

    Power efficiency is becoming far more of a concern with the worldwide political focus on ‘Greener’ standards and practices. I wonder what their 6core Processor(Thuban) will be like(edit: as in newer power efficiency), as well as AMD’s 890 FX/GX (edit:as in new “Cool n Quiet” management) chipsets when they release them in the upcoming months…perhaps we will something then?

    Good chart data and the ‘simplified’ build configurations is a good feature as well.

    (I think these edit will help clarify routes of thinking/approach)

      • OneArmedScissor
      • 10 years ago

      TDP has practically nothing at all to do with power efficiency.

      Even measuring a “100% load” doesn’t tell the full story. That could mean all sorts of things and doesn’t even remotely apply to general use scenarios.

      CPUs with more cores are generally more efficient because they can run lower clock speeds and still do enormous amounts more work. That’s exactly why AMD are doing 12 core CPUs while still at 45nm.

      But of course, that only matters if you’re using those cores. It’s all about what you’re doing, not a theoretical heat output figure.

        • KGA_ATT
        • 10 years ago

        Not according to AMD’s website. In their release (I also get emailed updates from them like many) of the Phenom II 965 X4 at 125w which brings “better efficiency”. In fact it is better. That’s just one of many things that can assist with power management/efficiency.

        Also here’s additional reading/reporting from other sources:

        §[<http://techgage.com/article/quick-gage_amds_125w_phenom_ii_x4_965_black_edition/<]§ §[<http://vr-zone.com/articles/amd-releases-revised-phenom-ii–x4-965-fits-to-125w/7986.html<]§ and of course: §[<https://techreport.com/articles.x/17894<]§ You seem quick to pull the trigger. Interesting.

          • Voldenuit
          • 10 years ago

          You shouldn’t be so quick to criticize, especially since most reviews have found negligible improvement in power draw on the 125 W C3 stepping vs the original 140W PhII 965.

          §[<http://www.xbitlabs.com/articles/cpu/display/phenom-ii-x4-965-c3_3.html#sect0<]§ §[<http://www.techreport.com/articles.x/17894<]§ TDP is a metric for OEMs to validate thermal solutions. It doesn't say anything about how much work a processor performs at that rating, nor does it tell you how much (or how little) the system consumes at idle. Hence I agree with OneArmedScissor's statements. Cyril: another approach might be to consider "performance per X dollars" eg, on a $600 build, an AM3 builder might be able to get a faster GPU than a 1156 builder, then they might have disparate performance (or not, hence my desire to see the testing done) on gaming benchmarks. EDIT: Ex: the price delta between the X58 and the 785G system is enough to upgrade the 5670 to a 5770, 4890, 4850 X2 or GTX260, all of which would absolutely demolish the 5670 in gaming benchmarks. We're not talking incremental improvement here (as is the case with going up speed grades of CPUs), this is night and day improvement.

            • KGA_ATT
            • 10 years ago

            Voldenuit, great point on the price difference between the X58 platform and the 785G. It has been stated by many others(myself included) that one can configure a mighty powerful system if such things as total platform and CPU costs are factored in. Which of course has been done in the review. The 785G,G43 and H55 are true competitors in price build, however the AMD Phenom II X4 965 loses out when it comes to efficiency.

            As far as your point about task/work a CPU does, it’s just as my post(my main point that is…). And a bit more. I was merely focusing on the lowering of the power consumption as ‘a start'(not heralding it as the ‘end all-be all solution), then followed by increased power management in the CPU, and chipset technologies to further increase power efficiency. Again pointing out that AMD may release such changes in their upcoming products(via new CPU’s or Chipsets).

            Now ‘multi-core/multi-cpu’ platforms(motherboards), with Server OS’s have superior efficiency for work/tasks that are handled. But that is an entirely different subject. Desktop OS’s cant even scratch that efficiency.

            • Voldenuit
            • 10 years ago

            IIRC, xbitlabs measured the idle power consumption of the 965 X4 on the 12V ATX lines, and it only drew 16W (DC) at idle. So it’s actually not too bad, despite its sky-high load draw.

            Would love to see AMD regain competitiveness, I think it’s doing very well at price/performance at the moment, but I’m sure their margins are a lot lower than intel’s.

      • JustAnEngineer
      • 10 years ago

      32nm vs. 45nm manufacturing should give Gulftown decent power efficiency numbers.

    • Usacomp2k3
    • 10 years ago

    Thanks for taking a 2nd look at it.

      • JustAnEngineer
      • 10 years ago

      Thanks for the update, Cyril.

      It’d still be interesting to see the Core i7-860 and the Phenom II X4 955 Black Edition in there.

      • flip-mode
      • 10 years ago

      Bump. Thanks for that Cyril… an unexpected treat for you to do that…

      FWIW, I think in desktop computers, idle power consumption is -by far- the more important measure, as long as load consumption is reasonable. Either way, Intel has done impressive things.

    • DrDillyBar
    • 10 years ago

    Q9300 a year ago FTW!

      • ryko
      • 10 years ago

      same here…got mine at microcenter for like $160

      • indeego
      • 10 years ago

      I should hope that anyone buying a processor a year ago wouldn’t be feeling pain yetg{<.<}g

Pin It on Pinterest

Share This