T420, HD3000, and RAM speed

I've read in a few places that upgrading the RAM in a T420 (or any Sandy Bridge laptop) can provide a speed boost for the Intel HD3000 IGP. Specifically, faster RAM will show a noticeable improvement. Can anyone here sound off on this? My Google-Fu hasn't come up with much on the subject, so I figured I'd ask around.

I'd like to upgrade the RAM in my T420 to 8GB eventually.

 

It is true on a couple of different levels. First, more RAM will benefit the overall environment if using a 64 bit OS. Today most people realize Windows 7 x64 does a great job in this regard. However, most people only need 4GB of RAM.

The integrated Intel iGPU will take advantage of more RAM because it uses shared memory. But keep in mind there are different versions and clock speeds of the Intel HD 3000 iGPU. This isn't commonly known. For example, take a look at this Intel proc comparison.

 

In theory, faster system RAM will result in better Intel 3000 IGP performance. I believe some NBR members have tested this out on the X220 (and other SB laptops with the Intel 3000 IGP), and although some have reported some gains, they're generally rather insignificant.

Bottom line: if the Intel 3000 IGP couldn't run the program satisfactorily before, using faster RAM won't change that.

 

I was referring to more RAM, not faster RAM. You should see a better result on a 8GB system compared to a 2GB system.

 

significant?   let's test that hypothesis.

here's how fast 1333MHz memory runs in an X220:



let's see some 1600MHz and 1833MHz memory results in comparison.

 

Now I'm tempted to get even more RAM for caching.

 

i tested various apps using a ramdisk for scratch/swap/page and didn't notice a difference over an SSD scratch disk.   by manually measuring the time it takes to update a temp file, it seems like most apps are bandwidth-limited to around 500MB/sec even on a ramdisk.   it's safe to say that software is a larger limitation than CPU, GPU, memory, storage device, or operating system, yet everyone focuses on hardware.

but, i'd highly recommend testing it just to see what you find.   it's always fun to experiment.

 

Get more ram, my E420 had 4gb or ram, I monitored the ram while playing civ 5, the intel hd3000 was using between 1400 and 1600mb of ram, it could not always get the full use of the ram because... the system itself was using the ram, I gave a long explanation in the E420 thread, but basically by upgrading to 8gb of ram, the gpu maxes and then the laptop uses the rest to play the game, the game is more stable, plays better, faster, etc........

 

Any two modules are better than one. The amount of RAM is less of a factor. The 1600MHz RAM gives a modest boost to graphics performance compared to 1333MHz RAM. Draw your one conclusions from this table of resuots which I've been wanting to analyse for some time when work stops getting in the way.



Sandybridge's built-in memory controller provides a major boost to integrated graphics performance compared to the previous Intel platforms.

John

 

Was the same CPU used in all of the tests? Which one?

Thanks

 

The graphics core will be the same for all the Core i5 and Core i7 CPUs offered on the X220, T420, and T420s (see Intel Comparison here). The only difference is present in the Core i3 CPUs, where the max GPU frequency is slightly lower (1.1 GHz versus 1.3 GHz).

That said, though, some of the benchmarks (ie: 3DMark06) places a pretty heavy emphasis on CPU performance, so different CPUs would impact the scores. I would assume John performed the tests on machines with the same configuration.

 

i5-2520M as mentioned here. IIRC, changing the CPU in the T420s isn't easy.

John

PS: And all tests used the same Intel 320 SSD: The SSD along makes a significant boost to PCMark scores.

 

Sorry, but I don't read minds. I didn't know if the same machine was used in all of the tests, that the machine was yours, and shockingly I don't have your review URL committed to memory or bookmarked.

 

1. It didn't cross my mind that anyone would think I had more than one T420s to play with. 2. Links to all my reviews are in my signature but maybe you use a forum skin that doesn't show the signature.

Anyway, does anyone have any further conclusions about what my test results show?

John

 

john - test your PC3-12800 in a ramdisk and let's see what you get for throughput in a storage environment.

http://members.fortunecity.com/ramdisk/Download/download_and_006.htm

download the x64 enterprise eval.

 

Interesting suggestion. I'll give it a try if I can get the file down through my poor internet connection. However, I should note that I've only got 4GB + 2GB modules to hand, which is not the optimum RAM configuration for performance (but, as I said more than once, 8GB results in a noticeable increase in hibernation / resume time).

John

 

OK, erik. Here's new results for the 1866MHz RAM using the Ramdisk tool you suggested. Looks a lot faster than 1333MHz...

 

which ramdisk software did you use?   your 4K/4K-threaded results are significantly worse, which would be a concern if you did indeed use the same software.

check the link two posts above yours for the one i used.

 

That members.fortunecity.... link does not work for me.

I used RAMDisk - Software - Server Memory Products & Services - Dataram

 

try this link: Download and/or Buy RAMDisk

they must have changed URLs in the meantime.

also, full disclosure, my CPU is an i5-2537M and i'm not sure if it truly has the same memory bandwidth capability as a normal-voltage processor.   intel may rate them the same but that doesn't mean results will be the same on ULV and NV procs.

 

Here's the long-awaited RAMdisk result for the Kingston 1600MHz HyperX(timings of 9-9-9-27) in my T420s (i5-2520M CPU).



When time permits I will test other RAM combinations on the same hardware platform.

John

 

That's probably right on for the DDR3-1600 being faster than DDR3-1866 on the 4K benchmarks due to the looser (high) timings on DDR3-1866.

Which all means DDR3-1600 with tight (low) timings is better than both DDR3-1333 and DDR3-1866 for the OS and most (smaller) applications but DDR3-1866 is better for the HD3000 IGP and applications that require more intensive memory requirements example: games like BF3, Photoshop, video editing and benchmark software from FutureMark.

I've been contemplating for quite some time now if the memory bandwidth of DDR3-1866 is worth the price premium over DDR3-1600 with lower timings but for personally for my use I think DDR3-1600 is the way to go since the there's still a pretty high price premium to get DDR3-1866 and the bandwidth combined with looser timings doesn't always result a performance gain in all situations.

But this thread is about " T420, HD3000, and RAM speed", so in short the answer is if it's just for the HD3000 IGP performance than DDR3-1866 is the best choice.