G.SKILL 8GB DDR3 Laptop Memory On A Linux Ultrabook

[phoronix]

Phoronix: G.SKILL 8GB DDR3 Laptop Memory On A Linux Ultrabook

With my purchase last month of the ASUS Zenbook Prime as a new Intel ultrabook for carrying out some development work while traveling, the only two traits of the system that I didn't like were the use of an Ivy Bridge processor over Haswell and that it had only 4GB of DDR3 system memory. Fortunately, the latter can be easily corrected and with the ultrabook order I bought a 8GB G.SKILL DDR3-1600MHz DIMM for this Core i7 system that I dual-loaded with Fedora 20 and Ubuntu Linux.

http://www.phoronix.com/vr.php?view=19540

[przemoli]

Michael.

Aint there any BIOS settings for disabling that on board memory?

Did You checked with ultrabook OEM?



And can You rerun some more 3D tests so we can see how it compare? (If that 2+8 conf. will be used in such scenarios)

[Michael]

Michael.

Aint there any BIOS settings for disabling that on board memory?

No there isn't such memory settings.

[marjancek]

Symetric memory

I'm getting a T440s tomorrow with 4Gb soldered + 4Gb simm.

I was sure I'd get a 8Gb simm right away, but the 50% increase in memory my not pay for a 50% reduction in speed.

I'm surprised that even memory intensive things such as kernel compilation don't come up ahead with more memory.

[stiiixy]

ouch. The lack of dual-channel really is stinging. Still, that increase in total amount is better than the virtual alternative. Maybe do some compilation time tests on using both set ups just for shit's'n'giggles? And the RAM speed affects the Intel 3D side of things, too yeah? I've never even bothered to check because I've never had to rely heavily on the CPU video!

[Michael]

ouch. The lack of dual-channel really is stinging. Still, that increase in total amount is better than the virtual alternative. Maybe do some compilation time tests on using both set ups just for shit's'n'giggles? And the RAM speed affects the Intel 3D side of things, too yeah? I've never even bothered to check because I've never had to rely heavily on the CPU video!

Yes the video is likely affected too. Sadly didn't run such tests for the memory comparison as didn't expect the impact too be so much. As I bought the laptop and RAM, just did this for weekend test and won't keep swapping out stuff on it.

[devius]

I could swear that Intel CPUs since Sandy Bridge when running this kind of asymmetrical memory configuration would try to enable dual-channel operation on the lower end of the memory pool where that is possible (the first 4GB in this case) and then revert to single-channel when accessing the rest of the memory addresses (or whatever).

[schmidtbag]

I'm surprised that even memory intensive things such as kernel compilation don't come up ahead with more memory.

Contrary to what many people believe, there is such thing as too much memory, just as there is also such thing as too high of memory frequency (higher frequency tends to mean lower latency, and latency also affects performance).

I find it a little hard to believe ASUS would actually limit BOTH memory sources to 1 channel, as this makes it much harder for the user to make an upgrade, but I don't see how dual channel could perform that poorly. Dual channel's peak is at around 10-12GB, where once you go beyond that, you start taking a performance hit.


Honestly though Michael, you're probably better off sticking with the old set of memory. I'm not sure what you do on that laptop that requires more than 6GB, but I would think that you're better off letting the OS try cleaning up the buffer than sticking with the 10GB.

[stiiixy]

I find it a little hard to believe ASUS would actually limit BOTH memory sources to 1 channel, as this makes it much harder for the user to make an upgrade, but I don't see how dual channel could perform that poorly. Dual channel's peak is at around 10-12GB, where once you go beyond that, you start taking a performance hit.

Hey mate, could you explain this part in a bit of detail? I'm curious as to how or why this happens, and is it only relevant to the current crop of DDR3-based hardware? And grammatically, can I use a , before an and? =D

[schmidtbag]

Hey mate, could you explain this part in a bit of detail? I'm curious as to how or why this happens, and is it only relevant to the current crop of DDR3-based hardware? And grammatically, can I use a , before an and? =D

I don't fully understand the exact nature of why it happens but I'm guessing its because there's just simply too many areas for the memory controller to focus on. Think of it like this - if you were in a library and had to re-shelve books but you could only use your arms, you can move quickly, but you can only effectively carry 5 books at a time. You yourself in this analogy would be a memory controller, andyour arms would act as maybe 512MB of RAM. Anyway, so you request a cart (a 4GB upgrade). Now, you're able to get the job done much faster since you don't have to keep going back and forth between shelves and the front desk, and the cart isn't really heavy enough to slow you down. But, lets say you're given 3 carts. Not only will the weight drag you down, but it'll take you much longer to find the book you're looking for and you also have to figure out how to manage all 3 carts when you only have 2 arms. So, the best option is to add another person (another memory controller). It is especially relevant to DDR3, because of it's relatively crappy latency. I never got the impression that DDR1 suffered the same problem as easily, on the other hand, most people at the time didn't need more than 8GB.

And yes, you are allowed to use a "," before "and".