AMD's A10-5750M Review, Part 2: The MSI GX60 Gaming Notebook
by Dustin Sklavos on June 29, 2013 12:00 PM ESTSystem and Futuremark Performance
There's a lot of information that we can cull from our standard suite of system benchmarks when it comes to the MSI GX60 in its stock configuration. We can place its performance against the last generation GX60, featuring an almost identical configuration save a theoretically slower APU, we can determine how much of a performance hit there is from only having single-channel memory operation, and we can get an idea of how much performance is getting left on the table by using an AMD A-series APU instead of an Intel i7 quad core CPU.
PCMark 7 will always favor systems with SSDs, so there's not a whole lot to learn here unless you compare the GX60 with the 7970M to the IGP and dual-channel memory (listed as the A10-5750M + HD 8650G). On the CPU side there's either almost nothing lost, or the 7970M is picking up slack.
MSI was right in that the single-channel memory operation doesn't really impact the GX60 in CPU-isolated tasks. It's consistently slower, but just barely so. Unfortunately we also get to see just how rough the drop down to an A10-5750M from an entry-level Ivy Bridge quad core really is; the i7-3630QM is never less than twice as fast as the A10. It's true that the i7 is rated to draw ~30% more power, but it gets at least 100% more performance.
Even when trying to isolate GPU performance with 3DMark, it's pretty clear the AMD APU is holding back the Radeon HD 7970M. The newer generation GX60 is able to eke out a fairly consistent, measurable lead over its predecessor, but a substantial amount of the 7970M's performance is clearly being left on the table. I'd say the 7970M is just too much GPU for the A10-5750M, but unfortunately AMD doesn't have any good substitutes for the 7970M/8970M; the 7870M and 8870M are both based on Cape Verde, which has half the GCN cores and memory bus width the Pitcairn-based 7970M/8970M does. And there's no 7950M/8950M based on the cut-down Pitcairn used for the desktop 7850, nor even any mobile chips based on Bonaire (HD 7790).
This highlights a glaring hole in AMD's mobile lineup; NVIDIA's only using the GK104 in their top end chips (GTX 680M and 780M), but they have the beefy GK106 to fall back on. Bonaire needs to make its mobile debut in a hurry.
Facebook
Twitter
RSS
69 Comments
View All Comments
JarredWalton - Sunday, June 30, 2013 - link
My testing suggests otherwise. Despite having 20 EUs vs. 16 EUs, HD 4400 and HD 4600 are generally about the same performance as HD 4000. I've even got a quad-core standard voltage i7-4700MQ system (MSI GE40), and surprisingly there are several games where the HD 4600 iGPU fails to be a significant upgrade to an ULV HD 4000 in and i5-3317U. I'm not sure if Intel somehow changed each EU so that the Haswell EUs are less powerful relative to IVB, but outside of driver optimizations (it's still early in the game for Haswell) I have no good reason for the lack of performance I'm getting from HD 4600.sheh - Monday, July 1, 2013 - link
The i7-4500U review showed not much of a difference at times, and even slightly lower performance, but +20% in other cases. i7-4770 vs i7-3770 shows more improvement. I'm guessing the improvement in a 37W CPU would be more like the desktop parts rather than the 15W CPU. But, well, all theoretical anyway.TheinsanegamerN - Saturday, June 29, 2013 - link
its a little problematic, i think. they are ultrabook cpus, and bga to boot. it's cheaper to make laptops with rpga connectors than soldering the cpu to the motherboard. until rpga models appear (not likely) we might not see that many. they are also $100 more expensive than ultrabook models from last year, so a lot of maufacturers are still using ivy. toshiba just brought out 3 new laptops, two are ultrabooks, all which use ivy instead of haswell.TheinsanegamerN - Saturday, June 29, 2013 - link
never mind, just found list that included socketed models. guess the oems are just slowIntelUser2000 - Saturday, June 29, 2013 - link
The SV dual cores are actually Q4, so its quite a few bit left. They are planning on phasing it out in favor of U chips, where it Broadwell it disappears entirely.sheh - Sunday, June 30, 2013 - link
On the forum you said Q4 is rumors, any more concrete info since then?arthur449 - Saturday, June 29, 2013 - link
I wouldn't sell or buy a laptop if it didn't have all its available memory channels populated. This entire review is pointless to me.JarredWalton - Saturday, June 29, 2013 - link
Because buying an extra 8GB DIMM for $75 or whatever is too hard? And it will only help with iGPU performance?arthur449 - Saturday, June 29, 2013 - link
Mainstream CPUs have used a dual channel memory controllers for so long that simply disabling access to one channel can have (as we've seen in this case) a not insignificant effect on benchmarks that do not directly stress the GPU portion of the chip. CPUs have had uses for more memory bandwidth before there were on-chip GPUs, afterall.Furthermore, in scientific terms, due to the author changing two variables in this experiment (var1a: iGPU. var 1b: dual channel. var2a: dGPU. var2b: single channel.) to compare the iGPU to the dGPU, we cannot be certain that the subtraction of the second channel of memory has a non-zero effect on dGPU performance.
If the author's intention was merely reviewing the out-of-the-box performance of this hardware configuration of the MSI GX60 Gaming Notebook, this would be fine. Instead, the article is titled: "AMD's A10-5750M Review: ..."
I'm not trying to be unnecessarily harsh. And the subject of this review could very easily apply to my interest. (I *just* purchased a laptop for a family member with an AMD A8-5550M.) But if the author is going to test the processor's gaming performance: don't change more than one variable.
TheinsanegamerN - Saturday, June 29, 2013 - link
the single channel ram should be fine in this case, since its not used for video at all. at least its 1600mhz