Comments Locked

56 Comments

Back to Article

  • xanxavier - Tuesday, April 2, 2013 - link

    Any chance you can divide the average power used per test into the scores to give us a performance factored by power ratio? That would be interesting to see. I think it was kind of obvious the hd4000 would win, but what if the lower end graphics had a higher TDP....
  • 3DoubleD - Tuesday, April 2, 2013 - link

    +1

    I was thinking the exact same thing. It would be very interesting to see that ratio.
  • dou che - Thursday, April 4, 2013 - link

    why? ipad just sucks, admit it.
  • Sabresiberian - Sunday, April 7, 2013 - link

    Well, frankly the user interested in a Surface Pro right now isn't all that concerned about power/performance ratio, they are interested in running a full-blown OS on a high-quality piece of hardware.

    Not that what you are asking for isn't important or interesting, it certainly is and would be. However, I will be more interested in those numbers when the Surface comes out using Haswell; I think Haswell will be a game-changer for it. The only reason MS released it now (using Ivy Bridge) was because they wanted to get it on the market and in the minds of potential buyers as soon as possible. In my opinion, of course.
  • r3loaded - Tuesday, April 2, 2013 - link

    Clover Trail's showing is frankly pathetic, especially when it's being beaten by a $200 Nexus 7. Then again, I didn't expect much since they stuck with an SGX545 instead of anything remotely modern for the GPU. If I didn't know it was simply a stopgap for Bay Trail, I'd have completely dismissed them from the running altogether.
  • powerarmour - Tuesday, April 2, 2013 - link

    It's mainly the drivers that are pathetic, as weak as the SGX545 is, Intel have done it no favours under DirectX
  • Alexvrb - Wednesday, April 3, 2013 - link

    It's both. In raw shader performance at equal clocks, the 545 has 1/8th the punch of an SGX554MP4. Intel has attempted to make up for lost ground a bit by cranking up the clockspeed, but they really should have used at least an SGX544MP2 at a bare minimum. Then on top of that, as you implied, the drivers are complete and total crap. Intel drivers have a history of being bad, but they've really taken it to the extreme here.
  • rwei - Wednesday, April 3, 2013 - link

    Sure, but that's to be expected. Why invest more than needed in a stopgap platform that you will soon abandon, to benefit only a fraction of your user base?

    I'd rather see them focus their resources on continuing to build the HD 4000 line.
  • Alexvrb - Wednesday, April 3, 2013 - link

    Except they didn't even design the graphics chip in this "stopgap platform". The company that did (Imagination Technologies) had better designs when Intel made this chip, but Intel went ahead with the very outdated SGX545. Again, ImagTech had better solutions on the table, and other companies were using them.

    Intel just didn't care enough, they saved on transistors and made a bigger profit. If you bought one, sucks to be you. Oh wait, isn't that Atom in a nutshell since it was introduced? Every generation someone says me that the next Atom will be worth a damn... I'm still waiting.
  • Lodorand - Friday, April 5, 2013 - link

    What's baffling is how strongly Intel is itching to penetrate the mobile market, yet it refuses (or fails) to put more resources into the Atom line and prefers to use outdated technologies and architectures. Well you can't have it both ways, Intel, you either invest or accept your irrelevancy.
  • zeo - Tuesday, April 16, 2013 - link

    Easily explainable... Intel only recently got serious about penetrating the mobile market. The ATOM was originally on a very slow 5 year product cycle and it's just over 5 years now!

    However, the upcoming Bay Trail marks the switch to a two year tic-toc product cycle... along with the first major architectural update as they're finally switching from In Order Processing to Out of Order and will offer quad cores and support for up to 8GB of RAM with a full 64bit implementation.

    They still have a lot to prove but it does look like they're finally getting serious about this, especially since they're pushing technology first introduced with Ivy Bridge to the ATOM with this release.

    In the meantime, Intel is upping the graphical performance a bit with the Clover Trail+ as that one will use a GMA based on dual 544 PowerVR GPU at 533MHz... should provide much more decent graphical performance until Bay Trail starts coming out by the end of this year, and on through early next year.
  • lorribot - Tuesday, April 2, 2013 - link

    Would be interesting to see how AMD's lowest power offering does.
  • Ewram - Tuesday, April 2, 2013 - link

    Agreed, need some AMD numbers!
  • kyuu - Tuesday, April 2, 2013 - link

    +1
  • duploxxx - Wednesday, April 3, 2013 - link

    agree, but OEM will never use it that much, they prefer to stick to fooling consumers with lots of marketing djingles that they buy the best brand in blue, Temash would be a way better fit then atom. It has low power usage , cpu power and gpu power and probably be cheaper in total...
  • zeo - Tuesday, April 16, 2013 - link

    Cheaper, no... Intel SoCs are barely more costly than ARM SoCs and that's about as low as they get.

    Temash, though, should offer more performance, especially graphically... However, it's still a generation longer before AMD can really compete for mobile device range power efficiency.

    Like aside from S3 suspend state, there's no support for features like always connected standby and idling doesn't go as low as ARM and ATOM SoCs can go.

    The dual core Temash is the only one that can go fan-less but it will still consume more power than the average mobile tablet. While the quad core will still consume significantly more and will require a fan but it will also offer a higher performance range, especially when docked as then it can enter a Turbo mode of 14W max TDP for more Ultrabook range performance.
  • oaso2000 - Wednesday, April 3, 2013 - link

    Here's what I got on my computer:

    AMD E-350
    Windows 7 x64

    3DMark Score: 13143 3DMarks
    Graphics Score: 15173
    Physics Score: 8953
    Graphics Test 1: 71.9 FPS
    Graphics Test 2: 61.0 FPS
    Physics Test: 28.4 FPS
  • kyuu - Thursday, April 4, 2013 - link

    Thanks for posting that oaso2000. I also went to the 3DMark website and looked at one of the results for a C-60 system, as the scores should be similar for the Z-60 (AMD's tablet APU). These are the results:

    3DMark Score: 10880 3DMarks
    Graphics Score: 13461
    Physics Score: 6511
    Graphics Test 1: 63.0 FPS
    Graphics Test 2: 54.7 FPS
    Physics Test: 20.7 FPS

    So, assuming those numbers are accurate, it looks like the Z-60 lags behind the newest ARM SoCs in terms of CPU performance, though not by a whole lot. GPU-wise, it's still ahead everything in the charts, bar the HD4000 in the Surface Pro.

    This confirms my suspicion that even the Z-60 (which should be greatly improved upon by Temash) is a much, much superior tablet solution than Clover Trail. God damn OEMs going for Intel silicon when the AMD solution is better *and* almost certainly cheaper.
  • 63jax - Thursday, April 11, 2013 - link

    i ask the same question! even if AMD will pull out a monster, the OEMs will still not adopt it, something is still very fishy there.
  • cmdrdredd - Tuesday, April 2, 2013 - link

    Does anyone else think that if battery tech would keep up then we wouldn't need to be worrying about power usage so much?
  • ssj4Gogeta - Tuesday, April 2, 2013 - link

    I agree. You'd still have to deal with the heat, but once you have a lot of power, you can find ways to dissipate heat too.
  • Zan Lynx - Tuesday, April 2, 2013 - link

    If battery tech "kept up" then it'd be illegal to carry your devices on airplanes. And probably most places in public.

    Unless an energy storage device has some rate limit on output, it is equivalent to a bomb. Fuel cells can be pretty easily hacked to explode. A superconducting loop storage could dump all its energy at one time like a lightning bolt. The effects of an energy storage flywheel failure are really spectacular.
  • whyso - Tuesday, April 2, 2013 - link

    Legally, manufacturers cannot sell a laptop with a battery over about 100 watt hours. Thats why the ultralong battery life notebookts (with the slice battery) have two smaller batteries.
  • iuqiddis - Tuesday, April 2, 2013 - link

    Is it possible to put up a comparison with a current generation desktop PC platform? Maybe a i7-3770 series or even an AMD APU. I know it's an apples to watermelons kind of comparison, but it would be fascinating to see where today's mobile offerings stand compared to old school desktops.
  • takeship - Tuesday, April 2, 2013 - link

    I can give my results at least, run this afternoon.

    Smartphone is a Tmob Sensation 4G (Dual core Snapdragon S3, with Adreno 225 graphics)

    Test 1: 24 fps
    Test 2: 21 fps
    Physics: 4.5 fps

    I'm running Android Revolution HD 7.1, so a bit better than what a stock HTC rom will give you.

    Desktop is a i7 3770 on ROG Gene V board with Direct CUII Radeon HD 7970

    Test 1: ~1540 fps
    Test 2: ~1450 fps
    Physics: ~140 fps

    So, summation: a top end modern GPU & CPU is roughly 64x the performance of last year's ARM v7.
  • WagonWheelsRX8 - Wednesday, April 3, 2013 - link

    Precisely what I was also curious about! With all of the focus on the march of progress of these SoC's and the availability of a cross-platform benchmark I wanted to know what the delta was...its still a big one, thanks for the post!
  • bengildenstein - Tuesday, April 2, 2013 - link

    "In an average frame, 530,000 vertices are processed leading to 180,000 triangles rasterized either to the shadow map or to the screen."

    I don't think this is correct. Each vertex in a strip (which then make up objects) imply a whole new triangle, so an example object with 530K verticies would have 530K - 2 triangles. I don't think a simple divide by 3 would yield the correct number of triangles rasterized for a scene like this. It would likely have far more triangles; probably very close to the number of total transformed vertices.

    Look up "Triangle Strip" or "Triangle Fan" on wikipedia for more info. Corrections welcomed.
  • 3dcgi - Tuesday, April 2, 2013 - link

    Triangle lists are used more often than strips and even if a strip is used it will be a bunch of strips not one giant strip. That said, the number is still weird as it implies there's no vertex reuse.
  • bengildenstein - Wednesday, April 3, 2013 - link

    Thanks for the added insight!
  • piroroadkill - Tuesday, April 2, 2013 - link

    Good idea.
    When you said "old school desktops" that made me think of actual old desktops.
    I'd love to see what old desktop configuration comes closest to some of these devices.
    Athlon XP with a Radeon 9700? Who knows!
  • WagonWheelsRX8 - Wednesday, April 3, 2013 - link

    I second that! My phone has a Tegra 2 in it and I guesstimated it was about the equivalent of a GeForce FX 5200 (if that)...but I don't have a PC with a configuration that old to test it. Of course Tegra 2 is long in the tooth now but I think it'd be really awesome to be able to relate a current SoC with an equivalent oldschool PC, just as a frame of reference.
  • Chubbbs - Tuesday, April 2, 2013 - link

    Out-of-order execution is not likely to help Atom with physics and other floating-point code. These workloads are all about memory bandwidth and the ability to simultaneously issue FP load/store and FP compute instructions.

    Remember that Intel spent most of the 1990s trailing the floating-point performance of the PowerPC 604/G3/G4, even though these microarchitectures had much smaller reorder buffers than the P6-based Pentiums.

    If Atom is to improve its floating-point, vector, and graphics performance, it needs a major overhaul of its load/store units, cache hierarchy, and memory controllers. There's just not a huge demand for better integer performance on mobile devices.
  • Phelerox - Tuesday, April 2, 2013 - link

    The Snapdragon 600 has an Adreno 320 GPU, right? So this is a typo: "HTC One (Snapdragon 600/Adreno 330)"

    Great article, as always.
  • lmcd - Tuesday, April 2, 2013 - link

    For some more context:

    "All of these tests are run at the default 720p settings. The key comparisons to focus on are Surface Pro (Intel HD 4000), VivoTab Smart (Intel Clover Trail/SGX 545) and the HTC One (Snapdragon 600/Adreno 330). I've borrowed the test descriptions from this morning's article."
  • Kidster3001 - Tuesday, April 2, 2013 - link

    You have the Atom part wrong for the VivoTab Smart. Z2760 is what's in it.
    Z2760 is Clovertrail (no +) with SGX545.
    Z2560 is Clovertrial+ with SGX544MP2
  • watersb - Tuesday, April 2, 2013 - link

    Thanks as always for a great article.

    Suggestion for Anand: Please remind us in the first few paragraphs that the VivoTab is the one with Clover Trail -- my very small brain hasn't memorized all the product specs yet. :-)
  • WagonWheelsRX8 - Wednesday, April 3, 2013 - link

    Concur...great article. Also haven't memorized which SoC is in which device, would be nice for the charts to note those next to the device name.
  • TiGr1982 - Tuesday, April 2, 2013 - link

    Anand,

    Your MSM8960 (Galaxy S3 AT&T) posted really weak scores in 3dmark. Mine (Motorola RAZR HD XT925, Canadian Rogers/Fido, stock Motorola Android 4.1.2, stock SoC frequency, unrooted and untouched) even with 1 GB RAM (in contrast to 2 GB in Galaxy S3) posted the following Ice Storm scores, which are way bigger:

    Regular Ice Storm:
    - Ice Storm score: 6293
    - Graphics score: 6688
    - Physics score: 5215
    - Graphics test 1: 30.7 FPS
    - Graphics test 2: 27.6 FPS
    - Physics test: 16.6 FPS

    Ice Storm Extreme:
    - Ice Storm score: 3536
    - Graphics score: 3220
    - Physics score: 5386
    - Graphics test 1: 18.1 FPS (18.1 FPS)
    - Graphics test 2: 11.4 FPS (11.2 FPS)
    - Physics test: 17.1 FPS (16.4 FPS)

    Sony Xperia SX and Motorola Droid RAZR M, based on the same SoC with 1 GB of RAM (unlike North American Galaxy S3 with 2 GB RAM) score mostly like my XT925.

    So, Galaxy S3 is severely underperforming for its SoC. Hence, it's better to look at other MSM8960-based devices (see above).

    Thanks.

    P.S. The "Post Your Comment" window in the new website design has ridiculously small vertical size. This window should be increased 4X vertically.
  • ChronoReverse - Tuesday, April 2, 2013 - link

    It looks like his GS3 is only activating one core when running the physics test which destroyed the overall score. The graphics score doesn't look too off.
  • yannigr - Wednesday, April 3, 2013 - link

    HD4000 looks like a Titan here. Funny.
  • dealcorn - Wednesday, April 3, 2013 - link

    I thought Intel disclosed what the next gen 22 nm atom's graphics engine is "comparable to". I wonder where that "comparable to" lands on the charts. 22 nm Atom will not have HD 4000 but it may be possible to speculate sorta insight-fully whether 22 nm Atom may have the right stuff. There is a place for properly labeled, insightful speculation about the upcoming battle. I liked the articles where you you used lots of wires to measured voltage drop and calculated power consumption by functional unit (e.g. graphics engine) within a SOC. The appeal of applying that technique to the "comparable to" SOC is it should provide an upper bound to next gen Atom's GPU power consumption. That technique provides wrong results because next gen Atoms will be built with a different (more power efficient) fabrication technology. However, it may hint how much (or little) Intel needs out of its low power fab tech to be competitive. If the din from the crowd at the coliseum clamoring for more red meat gets too loud, I might throw that into the ring.

    Second, I think in the first Chart, it is a boo boo to color the HPC item red.
  • scaramoosh - Wednesday, April 3, 2013 - link

    Why do people keep comparing this to ARM tablets? This is a Laptop replacement, those other devices are basically just large mobiles without the mobile capability.

    Two very different markets.
  • meacupla - Wednesday, April 3, 2013 - link

    Because ivy bridge needs to lower power consumption, while ARM needs to improve computing and graphics capacity?

    These two technologies are in convergence, so it helps to see where they are at now.
  • krumme - Wednesday, April 3, 2013 - link

    I think its relevant to compare a Porche to a Toyota. That gives some insigt, and the test here gives some surprising results.

    But the presentation of the test and the context is wrong, because its presented as if IB is a competitor to Arm. And thats just way off, and done in that perspective, it would be the same as if i compared my Samsung 9 series x3c to my doughters HP 11.6 AMD e450 APU. Thats just different markets and needs.

    When you read this and other articles you get the impression its interesting if broadwell or successor can reach AMD power levels. How on earth did that become interesting. Its excactly like comparing a cheap tn low res screen to the excellent pls. Or cheap wiskey to the most expensive. The comparison have its strong sides, and also the obvious limitations. I think its good journalism to present at write that explicit.

    Intel have huge capex, and that means its a dream to compare IB to ARM. Even Atom is a stretch, but thats the same market as huge high-end ARM, so its fine.
  • krumme - Wednesday, April 3, 2013 - link

    ARM not AMD power levels - reaching BD levels should be easy :)
  • zeusk - Wednesday, April 3, 2013 - link

    uhm, there is a big mistake in the article, QSD600 carries adreno 320 with higher AXI bus speed, not the 330 which is slated for QSD800
  • Drazick - Wednesday, April 3, 2013 - link

    Why don't you publish it on Google+?

    Anyhow,
    Could you compare the same hardware under Android and WP8 so we could see which OS is more efficient?
    Thanks.
  • kyuu - Wednesday, April 3, 2013 - link

    I don't believe they released the benchmark for WP8.
  • Krysto - Wednesday, April 3, 2013 - link

    Who cares about Surface Pro? Why are you comparing $1000-$2000 tablets with $200-$500 tablets? You know that's not a fair comparison.

    A much more fair comparison is with Surface RT, and then we can actually see what's the difference in scores for the cross-API/cross-platform benchmark, while using the same hardware (Tegra 3). So do one with Transformer Prime vs Surface RT. You may even include the Nexus 7 (although that may have a slightly lower clock for both CPU and GPU).
  • meacupla - Wednesday, April 3, 2013 - link

    It's more of a comparison of ARM vs. x86
    Seeing as x86 only has 3 or 4 CPU choices, ULV i3/i5 and clovertrail atom, it's quite obvious that these would be included.

    And why would I want a $500 tablet, when I'll need a $500 laptop to go with it? I might as well buy myself a $1000 x86 tablet.
  • Crazy1 - Wednesday, April 3, 2013 - link

    This comparo was done between new benchmark suites and whether they give comparable results across platforms and API's. It's just numbers. There is no need to get upset about it.

    Anand stated in the article that the new 3dMark has not been released for Windows RT yet. This article is clearly a piece of a much larger article that will thoroughly analyze 3dMark results across all supported platforms. We just have to wait for 3dMark to be released for all the planned platforms. Give it time.
  • Pinscher - Wednesday, April 3, 2013 - link

    would it be possible to add a testing of an amd based tablet. considering the superior graphics over intel hd 4000... very interesting times ahead, im sure of it.
  • BabelHuber - Thursday, April 4, 2013 - link

    Here is my OC'd Tegra 3 Score:

    Asus TF700T modified stock ROM (Android 4.1.1 CROMI 3.4.6) _that OC Kernel v2 CPU 4x1800MHz GPU 650MHz

    Ice Storm score 5150
    Graphics score 4600
    Physics score 8861
    Graphics test 1 18.2FPS
    Graphics test 2 22.2FPS
    Physics test 28.1FPS
    Demo 24.6FPS

    The Nexus 7's score of 3539 is anemic in comparison, even worse is the Physics test.

    The Tegra 3 has aged already, but a properly tuned device is not as slow as this Nexus 7 in this test here - what have you done to the poor thing?
  • kyuu - Friday, April 5, 2013 - link

    ... Are you seriously asking why they didn't use an OC'd Tegra 3 coupled with a non-standard ROM to benchmark with?
  • BabelHuber - Friday, April 5, 2013 - link

    No. I'm asking why the Nexus 7 is so slow. 8.9fps in the Physics test is a devastating result, even when compared to other Tegra 3 devices.

    My numbers only show what Tegra 3 is capable of in other devices. 28fps vs. 8.9 fps cannot only be the result of a 20% overclock...
  • Jraptor59 - Friday, April 12, 2013 - link

    Wow, a clunky, 50 pound tablet running a hostile OS that MS is trying to shove down our throats, can run 3DMark better the the light and elegant tablets! I must purchase this immediately!
    The Surface is running a faster processor but trade this off with the above caveats that remove it from the real tablet catagory (light, portable, etc.). I thing the Surface is more in the laptop class because of its weight. Compare it there for price and features for a real comparison. This is a stupid comparison. MS much?

Log in

Don't have an account? Sign up now