DDR3 1066 vs 1333 on 1647 (i5) - some findings

So, I received a replacement 1647 today due to overheating issues, one of the different components between the 2 was 1333mhz DDR3 in the replacement vs. DDR3 1066 in the original. I decided to do some testing between the RAM types knowing that the i5 is only capable of running the 1333mhz ram at 1066.

First, a quick summary of my findings: The 1066Mhz ram performs better all around paired with the 1647/i5.

You: Say what?

To check and compare RAM timings I used 2 programs, CPU-Z and CPU-Tweaker. CPU-Z showed only one timing different (tRFC @ 89 on 1333mhz RAM and tRFC @ 60 on 1066mhz RAM)

First I tested with all timings as the BIOS would set them up (no changes by me) - the 1066 RAM scored 9.25GB/s on SiSoft Sandra Memory Bandwidth test, 1333 RAM scored 8.88GB/s. I decided this probably was due to the difference in tRFC timings, so I used CPU-Tweaker to lower the timings on the 1333Mhz part to match (tRFC to 60) - the difference it made was nearly statistically within the margin of error.

Now my 2 test laptops were similar, but not exactly the same (1647 systems, i5 430m on one, 540m on the other, 5730 on one, 4670 on the other, broadcom 1520 vs. intel 6200, no bluetooth vs. bluetooth)

So i swapped the memory modules out, to make sure everything else was the same this time around..

And the result was once again the same. The 1066Mhz RAM consistently scored better than the 1333Mhz RAM. Even at the same exact timings. Furthermore, I was able to use CPU-Tweaker to lower the timings a bit more, and get the bandwidth test to eek out 9.45GB/s. The highest I was able to get the 1333Mhz RAM was 8.92GB/s.

Note that 1066Mhz RAM is PC3-8500 (8.5GB/s) so even at the default BIOS set timings it's performing above expectations. But the question remains, given equal timings and speed, why the heck is 1066Mhz RAM edging out 1333Mhz RAM by a statistically significant margin? I'm sure that on an i7 platform this would not be the case (as the 1333Mhz RAM would be running at 1333Mhz) - but here on the i5 side of things, that's an interesting result!

 

Just out of curiousity, does the Windows Experince Index change with the type of ram? I have 5,9 in ram-score. Its an 1647 with i5-540M and 4 GB 1066 ram.

 

I have 8GB 1066 ram on i5 520M and the score is 6.7...

 

I just re-ran the test and i've still got a 5,9 score. This seems pretty low? Although the SiSoft Sandra Memory Bandwidth test shows 9,1 GB/s, so this should be pretty normal?

 

Windows ratings have more to do with what it reads as the amount and the specs, than any actual speed test. On my desktop, I upgraded from 2GB to 4GB ram, and my score jumped from 5.9 to 6.4 without even running the assessment.

On my laptop, I got 7.4 from the specs on my sig line, but I doubt it's that much faster than yours...

 

Yep, the WEI scores seem to be mostly dependent on what components are in your configuration and not on actual measured performance. For example, the highest hard drive score possible without a solid state drive is 5.9. The highest RAM score possible with the 1333 DDR3 RAM in a 1645 laptop appears to be 7.4 and going from 4 GB to 6 GB or even 8 GB will not raise the score above 7.4. Also, various Windows updates can change what scores are available even if there were no changes in actual hardware. With one of the Windows updates a few months ago, the score for the i7 720 processor on my 1645 dropped from 7.0 to 6.9. With that same update, the processor score for my i7 920 desktop processor dropped from 7.7 to 7.6.

Also, while I don't know about overclocking the GPU, but, when the CPU is overclocked, it will have absolutely no effect at all on the WEI score.

 

If you notice, the graphics portions take up like 90% of the time of the assessment, so I think that it is the only thing that actually is tested. I just ran a GPU overclock (from 650/800 OCd to 825/1100) and my WEI for graphics increased from 6.4 to 6.9, not much, but still something at least.

 

As far as the Sandra benchmark in the OP goes, maybe the 1333 RAM just doesn't run as well when not running in its native speeds, even when running at identical settings as the theoretically slower 1066 RAM?

That is kind of weird though, you'd think that the same settings would have the same speeds, as long as the hardware was capable of running the settings at all.

 

Another place where I've seen reports of changes in the WEI scores that are apparently based on performance is when people have reported drops, sometimes quite large, in their hard drive performance with a sold state drive, and those drops would appear to have been related to issues with the drive and the TRIM technology.

 

Uhhh.. you kinda just wasted a bunch of time. Unless you overclocked the CPU with it, arrandales support only 1066 not 1333. You'd have to get a clarksfield, aka mobile quad, to support 1333. non of the i3/i5/i7s(dual core) support 1333. This is why you're seeing these results. You may be running the ram higher but the bus on the CPU can't take it, and therefor isn't sychronous with the memory, hence the performance penalty. But even before that, the cpu should have underclocked the memory to 1066 anyway..

You coulda saved yourself a whole lot of time had you just read up on the CPU/chipsets themselves.

 

Uhh.... Or you could have read the entire OP, and saved your self a lot of time writing a rebuttal that isn't relevant.... Just sayin'.

 

Lol well said.

And I have the 1333 MHz RAM in my Arrandale because it was offered as standard! And the Dell Rep said I should get 1333 MHz just in case I fancied a CPU upgrade in the future! But yeah that's not gonna happen...

 

i think you should re-read what i said bro before you get ahead of yourself. if you had, you probably wouldnt have posted trying to call me out...just sayin'

 

seeker_moc was correct to question your post, the points you made are invalid. I acknowledged that Arrandale doesn't support 1333 twice in my post. Also, communications with the RAM are no longer done over the bus on the i5/i7 platform, the DRAM controller is integrated on-die and the speed is determined by bclk * RAM multiplier. So the point about being synchronous with the bus is also invalid.

More to the topic, the clocks, timings, and subtimings being equal on both types of RAM, and all else being equal, they should output the same exact scores. But they don't. I suspect the memory controller may be able to take advantage of the 4 bank layout on the 1066 chip vs. the 2 bank layout on the 1333 chip to speed up reads/writes by spreading them amongst the 4 banks. Either that or the chipset is using less aggressive timings to ensure stability with the higher clocked RAM (1333). Curious also, that when booting linux the intel ips driver kicks back a warning that the expected TDP value is 35 and the BIOS reported TDP value is 25.

Code:

intel ips 0000:00:1f.6: Warning: CPU TDP doesn't match expected value (found 25, expected 35)

I saw a post from an intel engineer that this is expected behavior on an OEM system that is designed to operate below maximum TDP (i.e. it can't handle the heat dissipation of a fully loaded CPU) - i.e. our XPS 16s