Alienware's M18x, Part 2: AMD's Radeon HD 6990M in CrossFire
by Dustin Sklavos on October 14, 2011 12:00 AM ESTGaming Performance
We've seen that as far as Futuremark is concerned, the AMD Radeon HD 6990M and NVIDIA GeForce GTX 580M are essentially comparable. 3DMark doesn't tell the whole story, though, and we've seen graphics hardware perform well in it and then go belly up in actual benchmarking situations. Maybe our game testing will break the stalemate?
At least in our "high" preset, the two tend to trade blows, though when the GTX 580M wins it wins big, blowing past the Radeon in STALKER, Mafia II, and surprisingly DiRT 2, where the Radeon puts in an unusually poor performance. The 6990M is still totally playable in every case, but the GeForce is always at least on par with it.
The gap closes a bit in our "ultra" testing suite, but still generally favors the GeForce in multi-GPU settings while the single GPU solutions perform roughly on par with each other. Meanwhile, our multi-GPU solutions are oftentimes faster than our gaming desktop, and that's not too shabby. For users that want a desktop-level gaming experience in a notebook and are willing to pay for it, SLI and CrossFire look ready and able to provide it.
Facebook
Twitter
RSS
24 Comments
View All Comments
Alexvrb - Sunday, October 16, 2011 - link
Meaker is right. You can overclock it 100% and it doesn't mean beans if its throttling the heck out of it.Meaker10 - Friday, October 14, 2011 - link
You have the 6990M coming out behind the 6970M results....JarredWalton - Friday, October 14, 2011 - link
There are several factors at play. First, different drivers -- newer may not always be better, but without having both laptops and retesting, we can't say for sure. Second and more important by far is that the X7200 has a hex-core i7-990X. Even overclocked, the i7-2920XM can't always match it. Third, there's a difference in chipsets; the X7200 uses the X58 while the M18x uses the HM67. The X58 has tri-channel memory with two full x16 PCIe slots, where the mobile platforms go with dual-channel and two x8 PCIe slots.While individually, each of these seems minor, taken together it's not too surprising to see the X7200 win some of the gaming benchmarks. Also notice that in more GPU-limited tests (Metro 2033, Mafia II, and STALKER at our Ultra settings), the 6990M CF setup outpaces the 6970M CF by a fairly large margin. Most of our other titles, even at max settings, may not completely saturate the GPUs.
Meaker10 - Friday, October 14, 2011 - link
While I agree for the most part, if we were CPU limited we would see a hard wall, the CPU utilisation is close between AMD/Nvidia so that while there can sometimes be gaps they are not large.Well looking at the highest setting Dirt 2 benchmarks we see:
580M in the lead on the back of the 2920XM over the x7200 setup by 15%, a lead that suggests no GPU bottleneck.
Now looking at the M18X the 6990M crossfire is getting 77% of the FPS of the 6970 setup. Usually crossfire is less limited by PCI-E lanes than SLI.
Usually drivers can alter results, but if we project where the 6990M crossfire results should be then they are underperforming by around 30%, not something you would expect in newer drivers.
Have you checked the card's were not throttling during this run? Looks a bit suspect to me.
Meaker10 - Friday, October 14, 2011 - link
Sorry I meant to say it indicates no CPU bottleneck.JarredWalton - Friday, October 14, 2011 - link
I'll ask Dustin to check on the clocks and thermals of the AMD 6990M, as well as the i7-2920XM -- both could potentially throttle. I'll ask him about AMD driver version as well. Also, while on the desktop the SLI and CF results are often similar in terms of CPU utilization, on notebooks all bets are off. Every time I've played with an SLI or CF laptop, I've always felt like performance was never quite where I'd expect for the given hardware.For instance, SLI and CF scaling on desktops compared with scaling on notebooks seems like it usually doesn't do nearly as well on notebooks. I'd have to go analyze some hard numbers, but it's just been my impression. Another example is when you launch a game on an SLI or CF notebook, the display usually seems to flicker on and off for 10 seconds. Maybe that's been fixed now, but the last time I tested it I seriously thought, "WTF is going on!?"
In theory, everything should be the same, but when you go mobile it rarely feels that way. Yet one more reason to recommend the M17x or ASUS G74SX over the M18x.
Meaker10 - Saturday, October 15, 2011 - link
I have a 16F2 barebone myself (GT683R based).It would be interesting to face the x7200 and M18x against each other. Recording a baseline one card performance and looking at the mobile vs desktop chipset scaling with the same drivers and cards.
ik9000 - Saturday, October 15, 2011 - link
re your comment that the premium for a single 580 isn't worth it over the 6990. While dual GPU set-ups don't offer optimus, with a single card isn't this an option? Would Optimus give a single GTX580 machine a better battery life than a single 6990?iamlilysdad - Saturday, October 15, 2011 - link
It appears that Dell now lists a 256gb + 750gb option for custom configuration for the M17X R3. Guess that's another plus in the column for the "little" brother.Akv - Sunday, October 16, 2011 - link
Another "gaming laptop". Yawn...Sorry to repeat that every six months, but I am part of the population who have all the necessary large gaming equipment at home (screens, cases, fans, mice, keyboards...) and who use their laptop when traveling.
In that sense the prospect of gaming with a laptop seems to me appalling compared to what my desktops can offer, and the prospect of using a low-power, low-noise, light-weigth laptop with excellent screen and excellent speakers seems highly appropriate.