Understanding memory benchmarks

rdfiasco · Jul 3, 2013

So I just finished an upgrade. I replaced my board, CPU, and memory, but stuck with my old HDD for now. Here's what I have

Gigabyte 990FXA-UD3
AMD FX-8350
16GB Patriot DDR3 1600
Western Digital Caviar SE16 250GB 7200 RPM
http://www.newegg.com/Product/Product.aspx?Item=N82E16822144701
etc., etc., in my older Apevia X-Pleasure case.
http://www.apevia.com/ProductsInfo.asp?KEY=MX-Pleasure-BK#

My HDD is an obvious bottleneck in this system. I'm just waiting for a good deal on a SSD. However, I ran Passmark's PerformanceTest and I wasn't at all happy with my memory scores. 16GB G.Skill PC3-12800 (apparently the same speed) is outscoring my memory by 60% or more in nearly every category. I'm wondering if this could be a result of my HDD bottleneck, or if the Patriot memory I chose really is that much slower than the G.Skill...

Memory Mark
#1 - 2GB Transcend Information PC2-6400 777
#2 - 4GB Kingston PC2-6400 909
#3 - 8GB Samsung PC3-12800 1882
#4 - 8GB Nanya Technology PC3-12800 2215
#5 - 8GB Corsair PC3-10600 1950
#6 - 16GB G Skill Intl PC3-12800 2299
This Computer 1375

Pinhedd · Jul 3, 2013

Passmark sucks worse than a cheap dustbuster. Run SiSoft Sandra instead for some more objective tests.

One thing to keep in mind is that AMD processors have historically supported ganged, unganged, and interleaved modes of multi-channel memory operation while Intel processors only support interleaved (which is best). I'm not sure if the new FX series processors have this option, but you should set it to interleaved if possible.

Tradesman1 · Jul 3, 2013

What set of sticks do you have model # - there are a number of Patriot models in 1600 -

rdfiasco · Jul 3, 2013

Patriot sticks:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820220734

Tradesman1 · Jul 3, 2013

Actually that's about on a par for AMD...the MC is not nearly as strong as Intel's.....here's a link to what some others have done with different sticks under MaxxMem

http://www.gskill.us/forum/showthread.php?t=10545

06yfz450ridr · Jul 3, 2013

Check your timings in the bios and set them manually as well. I know my 1866 has really high scores and compares to higher end kits that are way over my speed and its only 9-10-9-24. I think I have mine set to 8-9-8-24 atm but I could be wrong

rdfiasco · Jul 3, 2013

Tradesman1 :

I don't think that's it. Comparing my results to a weaker AMD build w/ 8GB PC3-10600, I'm still being outperformed by about 40%. Is it possible I've configured something wrong?

rdfiasco · Jul 3, 2013

06yfz450ridr :

My timings in BIOS are correctly set to the stock settings.

Tabris DarkPeace · Jul 4, 2013

Some (many) of the 'extreme' memory modules are ones that don't support ideal rank in 99% of systems.
They use heat-spreaders to cover the markings on the IC so people can't just look them up.

G.Skill are much better in this respect than other brands, but they do still have some parts (with similar part id's) that lack what I would consider mainstream 'features' of DDR3-SDRAM (the sort of stuff that ALL Kingston ValueRAM modules support for example).

Makes you fugging wonder, eh?

That said, if they're CL1 faster on the CAS it can make up for lacking the rank features, etc.

Kingston HyperX (for example) always scores higher in the Windows Experience Index for RAM compared to Patriot in my experience.

Tradesman1 · Jul 4, 2013

Except WEI doesn't really rate sticks, it's based on mathmatical formulas as far as freq, number of stick, total DRAM etc, it doesn't test and rate performance, WEI is basically a joke among most everybody in the IT field

Tabris DarkPeace · Jul 4, 2013

It does actually test and rate performance, the WEI number is not a linear measure of performance and is widely misunderstood.

Otherwise, do you care to explain why Kingston HyperX at a lower clock speed rates 0.6 higher than an equal amount of Patriot DDR3-1600? (On otherwise identical systems).

Heck, the difference will also show up in SiSoft SANDRA, albeit far less pronounced as their benchmark is coded using inline assembly and far more optimised than any game title ever will be. The memory benchmarks in SANDRA heavily use the ALU and FPU of the processor and all the memory accesses are aligned on eight byte boundaries in their test.

In the 'real world' (vs synthetic highly optimised benchmarks) not all memory access is eight byte aligned, and the rank of the memory plays a HUGE part in real-life system performance.
- TomsHardware used to care (more) about this sort of stuff, more than 10 years ago, back when Tom Pabst was running the show.

You can even monitor the WEI process during the memory test and see it allocating significant amounts of memory.

Anyway, he was asking about Passmark's PerformanceTest.

rdfiasco · Jul 4, 2013

Interesting information...

So I'm gathering from this that the point is that my Patriot memory is just slow? If so, is it something I should worry about and consider exchanging the memory? I would have to pay a restocking fee.

Also, I'm still unsure about my original question: could the bad scores be related to my ancient HDD?

Since you guys brought up WEI, my memory is rated 7.8 there.

rdfiasco · Jul 5, 2013

...any thoughts on my last post?

Tradesman1 · Jul 5, 2013

Shouldn't, no

rdfiasco · Jul 5, 2013

Tradesman1 :

So you're saying there's not a significant performance difference in real application?

maxalge · Jul 5, 2013

rdfiasco :

Unless you are rendering/benchmarking having memory past 1333/1600 is not really noticeable.

Install maxxmem/aida64 and compare your score to other comparable systems.

Tradesman1 · Jul 5, 2013

Bad memory scores have nothing to do with a hard drive, a hard drive is storage like a CD or a DVD it not memory/DRAM

rdfiasco · Jul 5, 2013

maxalge :

I realize that. My problem is that I'm comparing the memory I have to memory of similar specifications and my scores are much lower.

rdfiasco · Jul 5, 2013

Tradesman1 :

Yes, I just wasn't sure how the benchmark scores were attained and if the slow read speeds of the hard drive could cause the memory scores to be lower since it wasn't getting the data fast enough, you know what I mean?

maxalge · Jul 5, 2013

rdfiasco :

maxalge :

I realize that. My problem is that I'm comparing the memory I have to memory of similar specifications and my scores are much lower.

Check your motherboard's booklet it could be as simple as your ram is running in single channel mode because you have them inserted into the wrong slots.

rdfiasco · Jul 6, 2013

maxalge :

Thanks for the suggestion. I was careful to install them into the correct slots when I put everything together, and I rechecked when I first saw the bad scores.

Understanding memory benchmarks

Distinguished

Champion

Legenda in Aeternum

Distinguished

Legenda in Aeternum

Honorable

Distinguished

Distinguished

Honorable

Legenda in Aeternum

Honorable

Distinguished

Distinguished

Legenda in Aeternum

Distinguished

Champion

Legenda in Aeternum

Distinguished

Distinguished

Champion

Distinguished

Share this page