GTX 480M SLI Takes on Desktop SLI and CrossFire

This is a desktop replacement, so what better way to show performance than to compare it with actual desktops? We don't really have any other notebook results at High 1080p settings, let alone our Ultra 1080p metric, so we'll compare the Clevo X7200 with our most recent desktop reviews. We've got the iBUYPOWER XLC (an i7-930 running at 3.5GHz and SLI GTX 470), a CyberPower Gamer Xtreme 8500 (i7-875K at 3.85GHz with EVGA Superclocked GTS 450 SLI), and an AVADirect Nano Cube (i5-750s at 3.0GHz with a single HD 5870 stuffed into a mini-ITX chassis). I've also added results for my own gaming system: an i7-920 (stock; still working on a CPU upgrade so I can actually overclock) with CrossFire HD 5850. Pricing for the systems is heavily against the X7200, obviously: the Nano Cube checks in at around $1600, the CyberPower costs close to $1700, and the iBUYPOWER (as well as my custom system, thanks to the two 120GB SSDs) will set you back roughly $2300.

Battlefield: Bad Company 2

DiRT 2

Left 4 Dead 2

Mass Effect 2

STALKER: Call of Pripyat

StarCraft II: Wings of Liberty

At our "High" settings, the Clevo X7200 falls just shy of the GTS 450 SLI results but manages to beat the single HD 5870. Considering the latter is easily able to churn out playable frame rates in nearly every game at maximum detail and a 1080p resolution, the X7200 is no slouch. However, look at higher end dual-GPU options and it's clear the GTX 480M is nowhere near the performance of the desktop GF100 cards.

Battlefield: Bad Company 2

DiRT 2

Mass Effect 2

STALKER: Call of Pripyat

StarCraft II: Wings of Liberty

Moving up to "Ultra" detail, the X7200 manages to post a lead over SLI GTS 450 in several titles, while DiRT 2 and ME2 still give the lead to the desktop system. We don’t normally test beyond 4xAA, but with Left 4 Dead 2 already at 4xAA and running smooth, we decided to max it out with 16xQAA and still pulled 105FPS! Anyway, back to the desktop comparison. Without identical CPUs, we can't say with certainty how much of the difference lies with the GPU and how much comes from the CPU. More likely than not we're GPU limited in the CyberPower, iBUYPOWER, and Clevo X7200; however, the stock i7-920 and 5850 CrossFire as well as the Nano Cube HD 5870 with an overclocked i5-750S appear to be hitting some CPU bottlenecks.

What that really means is that the best we can do in mobile graphics is still about one-half to two-thirds of the top-performing desktop solutions. You can get a rough idea of how the Clevo X7200 stacks up against desktops in our GPU Bench; we run the same test in BFBC2, DiRT 2, and almost the same in Mass Effect 2and STALKER: Call of Pripyat, though the desktops were tested at 1920x1200 instead of 1080p.

Clevo X7200: This is your games on 480M SLI Clevo X7200: Breaking Application Performance Records
Comments Locked

38 Comments

View All Comments

  • Candide08 - Saturday, October 9, 2010 - link

    Some companies, like mine, are no longer buying workstations. We are issued "desktop replacement" laptops, like HP 88740W with dual i7 quad cores and 8GB of memory. Throw in Virtual-Box and run one or two other VM's and its workable as a full function portable desktop with 1920 x 1200 screen resolution and a 5.9 Win 7 rating.

    The 8740W is not quite a spec'd-out as this, but its in the same league.
    Yes, its heavy. Yes, the power brick is huge. Yes, I bought a rolling laptop case.
  • crackedwiseman - Thursday, October 7, 2010 - link

    Honestly, why not go for a desktop GPU with those sort of thermals? You could quadfire a mobile Juniper GPU, or use a vanilla, normally binned 5870 and still come in with the same power draw.
  • Meaker10 - Sunday, October 10, 2010 - link

    Because:

    A: It's easier to cool two separate sources of heat.
    B: Moduals are of the smaller size are made already with mobile GPUs.
    C: Quadfire would take up more space and would suck for efficiency.
    D: Desktop drivers do not have some of the mobility options.
    E: Mobile 480M crossfire is faster than a single desktop 5870.
  • rsgeiger - Thursday, October 7, 2010 - link

    I forgot if you guys ever got a review unit from Alienware, or are going to in the near future, but a comparison of the dual GPU notebooks out there would be fun to see.

    Otherwise great review! I really read closely this time the Powerdraw, heat, and noise comparisons. This always the most important buying decision I have when buying DTRs.
  • JarredWalton - Thursday, October 7, 2010 - link

    We looked at the previous generation M980NU with QX9300 and SLI GTX 280M, which is pretty similar to the older M17x. Now M17x is doing up to i7-940XM, with either HD 5870 CF or GTX 285M SLI. (5870 CF is going to be faster than 285M SLI by around 20-25% I think, but no PhysX or CUDA.) Anyway we updated our gaming benchmarks and don't typically have the opportunity to rerun new tests on previous reviews, which is why we focused on comparisons to desktops.

    You can always get a rough estimate by comparing 3DMarks, but it's pretty safe to say the M17x needs a new revision before it's going to come anywhere near the performance of the X7200. Pricing for a maxed out M17x is currently $3900 without the RAID 0 C300 SSDs, so $5000 total with 5870 CF and 8GB RAM. But you have to look at what you're missing: 480M SLI is quite a big jump from 5870 CF, and the i7-980X is, as I mention in the text, about 50% faster than even the i7-940XM. Alienware does have a much nicer LCD panel, though: 1920x1200 with RGB LED backlighting.
  • Rasterman - Thursday, October 7, 2010 - link

    CPU wise the 980x is 136% faster than the 940xm.
    http://www.cpubenchmark.net/high_end_cpus.html
  • 5150Joker - Thursday, October 7, 2010 - link

    It's not "much faster" than an M17x with 5870m Crossfire + 940xm. If you look at NBR, you'd see that the 480M SLi barely outpaces the 5870s by a mere 500-800 3dmark vantage points. That's basically nothing. When both are overclocked, the X7200 again does not surpass the M17x, they are about even.

    The biggest advantage the X7200 has is the desktop processor but there's a program called ThrottleStop that allows end users to manipulate TDP/TDC + multiplier settings for the 940xm in the M17x. I've gotten the 940xm as high as 3.8 GHz on all 4 cores (8 threads) so again, the X7200's desktop processor advantage is diminished even if it is 6 cores.

    You have to keep in mind the X7200 is severely limited by it's 300W PSU as you discovered so there's no real room for overclocking. One of the first users to purchase the X7200 on NBR discovered that his 480m SLI + 980x setup was shutting off during the Mafia 2 benchmark and he was running the system at stock!
  • JarredWalton - Thursday, October 7, 2010 - link

    I'm out of town now so I can't test other scenarios, but I ran every benchmark several times and the only thing to ever overload the power brick was Furmark. All the other tests I saw topped out at power draws of around 310W, but it's possible some combination could get that higher. I'd like to know exactly what others have seen overload the PSU, though. Note that I did get beta drivers from NVIDIA for testing (260.80 I think -- the WHQL drivers should release "soon" according to NVIDIA), and that might be why I didn't experience other overloads.

    As for performance comparisons, you can't cite 3DMark Vantage as a meaningful item. I include it as a quick point of reference, but it's not a game and it doesn't really behave like most games. Simply getting GPU PhysX to work should boost the score I posted by 1000 or more, but it's ultimately the games that matter.

    For gaming, the X7200 with 480M SLI beats the tar out of everything else in the notebook world. If you had a similar X7200 with HD 5870 CF, the difference would be smaller (thanks to having a fast CPU), but with a large cross section of games it's pretty clear 480M is faster. It's also more power hungry by a large amount, though, so I can understand going the 5870 CF route. It's too bad that the only way to do get 5870 CF is the M17x or Clevo X8100, since that also gives up the desktop CPU and you get areas where you're CPU limited.
  • Meaker10 - Sunday, October 10, 2010 - link

    PhysX scores are not valid results for comparisons.

    The original intention of the test was for 3rd party physX cards at showing the potential in the future, even then the scores were not valid.

    Nvidia twisted this and while it has not been removed, PhysX scores will not appear in comparison searches unless you specifically choose them to. They will never appear in the hall of fame.

    Futuremark recommends that reviewers keep PhysX off when using 3dmark to review graphics cards.
  • silverblue - Thursday, October 7, 2010 - link

    Ouch... :) I'd be surprised if mine is over 25lb, though it's hardly a power machine. I'd love to have a play with a 50lb desktop... see if the power/weight ratio is favourable. ;)

Log in

Don't have an account? Sign up now