Core i7-3970X Extreme Review: Can It Stomp An Eight-Core Xeon?

Power Consumption And Efficiency

Logging power consumption during the entire benchmark run gives us a window into the behavior of each processor and platform.

We’ll start by looking at the Core i7-3970X’s green line. A 150 W TDP should have been enough for us to know that high power consumption would be measurable. Indeed, the -3970X spikes up to use the most power under load.

What about the eight-core Xeon E5, which is also a 150 W processor? Despite that rating, its consumption is far below the new Core i7s, and in fact looks to be even more conservative than FX-8350’s power use.

Although the Core i7-3960X is only just slightly slower than the -3970X, the yellow line is quite a bit lower on our chart. And of course, the Core i7-3770K is downright miserly.

Averaging those power numbers out reveals that Intel’s Core i7-3970X fares worst. And although the Xeon E5 bears the same 150 W rating, it ends up using even less power than the 130 W Core i7-3960X.

Calculating efficiency requires that we know two things, though. First, we need the consumption figures above. Then, we need some measure of performance.

We capture this data by logging our scripted benchmark suite in two-second intervals. Multiplying out the number of resulting samples by two, dividing by 60 (minutes), and dividing by 60 again (hours) tells us how long each processor takes to wrap up testing.

Eight cores reign, it appears, though the Core i7-3970X and -3960X aren’t far behind. Opting for a Core i7-3770K gets you significantly lower average power use; however, it also takes 10 minutes longer than the -3970X to complete testing.

We’ll leave the FX-8350 alone in all of this. Its high power consumption and last-place performance are countered only by a $220 price tag.

Not surprisingly, Intel’s 77 W Core i7-3770K is the most efficient choice. More interesting is the fact that the 150 W Xeon combines brutal speed with lower-than-expected consumption to finish second in our efficiency metric.

Core i7-3970X isn’t able to add enough performance to make up for dramatically higher power use, and this reflects in its loss to the -3960X.

Then there’s the FX-8350, which, again, combines modest performance and high average power use, hurting its efficiency story.

Create a new thread in the US Reviews comments forum about this subject
This thread is closed for comments
74 comments
    Your comment
    Top Comments
  • tumetsu
    I've recently started facepalming every time I see BF3 in CPU benchmarks. "Boy oh boy, this hasn't been confirmed like a hundred times already but the single player is decidedly graphics-bound, so here, have these charts with identical results anyway."
    31
  • jaquith
    Boo on Intel for not enabling all 8-cores especially at that price!
    28
  • amuffin
    100mhz faster than the 3960X, not worth the extra premium.

    Same thing goes for the 3960X compared to the 3930K....not worth the extra 100mhz for $400....
    26
  • Other Comments
  • amuffin
    100mhz faster than the 3960X, not worth the extra premium.

    Same thing goes for the 3960X compared to the 3930K....not worth the extra 100mhz for $400....
    26
  • jaquith
    Boo on Intel for not enabling all 8-cores especially at that price!
    28
  • tumetsu
    I've recently started facepalming every time I see BF3 in CPU benchmarks. "Boy oh boy, this hasn't been confirmed like a hundred times already but the single player is decidedly graphics-bound, so here, have these charts with identical results anyway."
    31
  • dragonsqrrl
    jaquithBoo on Intel for not enabling all 8-cores especially at that price!

    They don't have much of a choice when it comes to the i7's. With the 32nm Sandy Bridge-E Intel has to make a choice between prioritizing clocks or core count within a 150W TDP, based on the target workload for a particular processor. For Xeon's the choice is easy, more cores. For desktop applications the choice isn't as clear, but I think most users would still benefit more from a higher clocked 6-core than a lower claocked 8-core. That's slowly changing though.

    Intel also doesn't want a situation where their LGA 1155 processors outperform their $1000 extreme edition in lightly threaded workloads, which is yet another reason to favor 6-core for now.

    I'd personally like to see an 8-core i7, even if it means lower clocks, but I don't think that'll happen until Ivy Bridge-E. At 22nm Intel probably won't have to make a choice, we'll get the best of both worlds.
    15
  • samuelspark
    So much money...
    19
  • nebun
    jaquithBoo on Intel for not enabling all 8-cores especially at that price!

    why would they....they don't need to do it at this time....amd's top cpu is still very slow when compared with even intels mid rannge cpus
    -7
  • unknown9122
    Why do people still benchmark on itunes 10.4? 10.7 is out... as for the 8 cores as said above^, there is no need to have more than 6. Because if it had 8, then xeons would not sell to pros.
    9
  • dark_wizzie
    Why are we not manually overclocking this expensive CPU? Why do we do benchmarks against stock ig 2500k?
    0
  • A Bad Day
    You also forgot something when comparing against Xeon:

    Stability test.

    Run the i7 for one month under Prime95. It will crash. Run the Xeon for one month under Prime95. If it crashes, then you got a defective Xeon because they're not suppose to crash under 24/7 workload.
    12
  • anthonyorr
    nebunwhy would they....they don't need to do it at this time....amd's top cpu is still very slow when compared with even intels mid rannge cpus


    Why would you even include the 8350? It is 1/6th the price of this CPU. I couldn't imagine what a modern AMD desktop CPU would consist of at the $1000+ price range.
    14
  • A Bad Day
    dragonsqrrlI'd personally like to see an 8-core i7, even if it means lower clocks, but I don't think that'll happen until Ivy Bridge-E. At 22nm Intel probably won't have to make a choice, we'll get the best of both worlds.


    Or set the TDP to 195W and add a warning stating that Intel's stock coolers won't be sufficient for the thermal load.
    -4
  • merikafyeah
    A Bad DayYou also forgot something when comparing against Xeon:Stability test.Run the i7 for one month under Prime95. It will crash. Run the Xeon for one month under Prime95. If it crashes, then you got a defective Xeon because they're not suppose to crash under 24/7 workload.

    The i7 isn't supposed to crash under 24/7 workloads any more than a Xeon is. The footnote to this statement however has everything to do with thermal envelopes. The primary reason why Intel charges $1000 more for a Xeon is because Xeons can operate stably at higher temperatures. This is very, very important because cooling costs lots and lots of money when you need to cool hundreds of server racks and rendering farms all year round and there can't be any downtime. Ever. Because lost time equals lost money. They say every minute of downtime means a million dollars lost, so you pay extra to ensure a sudden heat wave doesn't wipe out your business. (Though some companies could certainly use better flood protection as to not rape their customers for lost profits, but I digress.)
    15
  • nebun
    anthonyorrWhy would you even include the 8350? It is 1/6th the price of this CPU. I couldn't imagine what a modern AMD desktop CPU would consist of at the $1000+ price range.

    lol....here is the answer...a lsow cpu, lol
    -9
  • apache_lives
    A Bad DayYou also forgot something when comparing against Xeon:Stability test.Run the i7 for one month under Prime95. It will crash. Run the Xeon for one month under Prime95. If it crashes, then you got a defective Xeon because they're not suppose to crash under 24/7 workload.


    Not correct

    Your talking about a processor (i7) vs a platform (Xeon, since the Xeon's usually require ECC memory, server boards, usually server OS etc) -- 99.999% of the time crashes/issues are NOT processor related.

    Crashes are usually from things like non JEDEC standard spec ram, poorly written SSD firmware, bad drivers and so on - not a processors fault.
    11
  • A Bad Day
    merikafyeah(Though some companies could certainly use better flood protection as to not rape their customers for lost profits, but I digress.)


    I recall seeing a picture of a tiny server room flooded with raw sewage from a busted pipe. The management of the small business attempted to ignore the laws of physics and told the IT that there will be no shutdowns.

    I'm pretty sure there was a smell of magic smoke accompanying the sewage odor shortly afterwards.
    5
  • clonazepam
    Great article.

    "Complete Tom's Hardware Suite" Chart... ugh I hate being color blind, must concentrate harder.

    The 3DMark 11 benchmark was a "Performance" run I'm guessing. Couldn't find that anywhere.
    1
  • blazorthon
    A Bad DayYou also forgot something when comparing against Xeon:Stability test.Run the i7 for one month under Prime95. It will crash. Run the Xeon for one month under Prime95. If it crashes, then you got a defective Xeon because they're not suppose to crash under 24/7 workload.


    If the i7 fails, then it's no less faulty than the Xeon. They're the same chips, just with different feature sets.
    6
  • TheBigTroll
    nah. xeons are binned for lower power consumption and lower heat output compared to the i7 counterparts
    7
  • blazorthon
    TheBigTrollnah. xeons are binned for lower power consumption and lower heat output compared to the i7 counterparts


    Same chips still. Binning doesn't change what they're made of. Similar to how the Tahitis in the 7970 GHz Edition are better binned than those in the regular 7970, but they're still the same chips.
    7
  • TheBigTroll
    true. but they are different in their own way
    6