Core i7-3720QM: Ivy Bridge Makes Its Mark On Mobility

Benchmark Results: WoW, Call Of Duty, And Battlefield 3

If you want the architectural run-down of the Ivy Bridge architecture's HD Graphics 4000 core, head on over to Intel Core i7-3770K Review: A Small Step Up For Ivy Bridge and check out page three. Briefly, the new engine adds four execution units, which now total 16, and a number of optimizations for performance that yield better benchmark results than Intel's HD Graphics 3000 implementation.

The advantage is more significant on the mobile side than it was in our desktop-oriented measurements. The flagship Core i7-3920XM offers a maximum graphics clock rate of 1.3 GHz, where as the Core i7-3770K's HD Graphics 4000 component tops out at 1.15 GHz. The base clocks for both the mobile and desktop processors are 650 MHz.

The Core i7-3720QM in our notebook sample employs a maximum graphics frequency of 1.25 GHz, putting it just behind the Core i7-2820QM's highest bin. However, the increase in resources dedicated to higher frame rates means HD Graphics 4000 still delivers superior speed.

In our evaluation of the Core i7-3770K, we found that AMD's 100 W Llano-based APUs delivered better graphics performance than Intel's Ivy Bridge architecture. The A8's bigger power budget gives AMD more of an opportunity to emphasize its GPU component. But when you scale all the way back to a 35 W TDP, that's no longer true. Intel's manufacturing advantage more palpably kicks into play, and HD Graphics 4000 is able to shine within a 45 W thermal envelope.

Radeon HD 6620G is the highest-rated graphics implementation in AMD's mobile Llano family. But with only 400 shader cores operating at 444 MHz, the HD 6620G poses no threat to HD Graphics 4000 (or even the HD Graphics 3000 that came before).

In Call of Duty: Modern Warfare 3, AMD's Radeon HD 6620G basically matches the performance of HD Graphics 3000 with anti-aliasing disabled. However, HD Graphics 4000 blows both solutions out of the water, delivering more than twice as much performance. Even  with anti-aliasing enabled, AMD's Radeon HD 6620G is ~30% slower than Intel's HD Graphics 4000.

The older Core i5's vanilla HD Graphics engine is too old to support Battlefield 3. Given the numbers we see from the other systems that do manage to run it, performance would fall in the single-digit range anyway, though.

Again, we see the Radeon HD 6620G match the performance of HD Graphics 3000. But HD Graphics 4000 runs away with the victory, averaging higher (but still only marginally-playable) average frame rates. The speed-up ranges from 70 to 80%, depending on the resolution and quality settings you consider.

  • fstrthnu
    Would there be a noticeable performance gap between the i7-3720QM and the i7-3612/5QM? I'm trying to decide whether the extra 300 Mhz is worth ~$150 more (which I'm guessing not really)
    Reply
  • s3anister
    fstrthnuWould there be a noticeable performance gap between the i7-3720QM and the i7-3612/5QM? I'm trying to decide whether the extra 300 Mhz is worth ~$150 more (which I'm guessing not really)
    There would be a performance difference in applications that could use the extra MHz (Video games, encoding/decoding) and performance would scale accordingly. Otherwise no you'd likely never notice.
    Reply
  • dragonsqrrl
    Wow, it looks like Ivy Bridge is a very compelling option in the mobile market. I had no idea the mobile versions of Llano were so performance constrained by their TDP. The graphics performance results are especially interesting. Just turned my whole world view upside down.

    Great job. Another excellent review Andrew.
    Reply
  • fstrthnu
    It looks like the Geforce GT650M in the notebook I'm looking at would bottleneck faster than the processor would, so I guess I'll save $$ then
    Reply
  • amiame
    A high end desktop plus an ivy bridge ultrabook. Now, THAT works pretty well.
    Reply
  • DjEaZy
    ... why there waz no screenshots of picture quality differences in games between intel's HD4000 and AMD's HD6620?
    Reply
  • blazorthon
    dragonsqrrlWow, it looks like Ivy Bridge is a very compelling option in the mobile market. I had no idea the mobile versions of Llano were so performance constrained by their TDP. The graphics performance results are especially interesting. Just turned my whole world view upside down.Great job. Another excellent review Andrew.
    To be fair, it was a low power APU being bench-marked against higher end, higher power, and newer chips. I would be surprised if it won much of anything, besides power usage, against the Sandy and Ivy i7s. A higher TDP mobile A8 might be able to beat HD 4000 if it had 1600MHz or maybe even 1866MHz memory, granted it still wouldn't win in CPU performance.
    Reply
  • ojas
    Interesting review. But i guess people are likely to point out differences in price (thus affecting performance/$), and RAM speeds, which apparently impact IGP performance.

    IIRC, the IGPs on the mobile chips can be OC'd, right?

    The Core i7-3720QM particularly shines in tests involving:

    Video Transcoding
    DX9 Graphics
    Web Browsing
    Hmmm...wouldn't you agree that "data decryption" should be on this list too? The difference b/w each proc is significant...plus you've got hardware acceleration for AES256 on SB and IB...

    I hope the mobile i3s get HD4000...still wondering why the i5s didn't get it...
    Reply
  • DavidC1
    Andrew, love the review. But there's what seems to be a big error. You said on the power usage tests that AMD defaults to max battery life while Intel goes to balanced? Looking at World of Warcraft results, it looks like all the other results may be running max battery life mode for the AMD A8 chip.

    The i5-460M is faster than A8-3520M, just not that much faster. I have a feeling you need to run the application and gaming tests on max performance all over again. It doesn't matter for the Intel part as Balanced pretty much performs like max performance.
    Reply
  • DavidC1
    Also, you need to do a battery life test. Power usage and battery life tests are hard to connect, because of advanced power management techniques and different usage models.
    Reply