Did I buy slow RAM? (Solved)

[Agentbb007]

So I bought these G.Skill Ripjaws V's (F4-3000C15D-16GVRB) for $102.99 last Wednesday.
I did some AIDA64 Benchmarks and got 41166 MB/s Read, 43424 MB/s Write, 37763 MB/s Copy 52.0 ns Latency. Screenshot below.

Then today I read the review on the Crucial Ballistix Elite 3200 MHz DDR4 16 GB and see 71289 MB/s Read, 73537 MB/s Write, 72661 MB/s Copy and 56.2 ns Latency in AIDA64.

So should I return these and pay the 15% restocking fee and buy these Ballistix Elite for $139.99? Or these 2x8GB sticks that are DDR4 3000? Or feel free to let me know if there's some other RAM I should be buying. I originally replaced my Crucial Ballistix Sport LT DDR4 2400 32GB for these Ripjaws just to see if they are causing my random crashes every so often during long gaming sessions.

Thanks
BB

 

[yotano211]

THere must be something that is slowing down the RAM speeds. Is it mixed with another set of Ram sticks.

 

[Agentbb007]

Nope just the two Ripjaws V Series 16GB sticks in there.

 

[Kursah]

Was that review with a quad channel X99 system? If so that bandwidth makes sense. Your speeds seem close to what I'd expect from them on your platform.

 

[Agentbb007]

Was that review with a quad channel X99 system? If so that bandwidth makes sense. Your speeds seem close to what I'd expect from them on your platform.

Ah ha! That's what I was thinking might have been why those were so fast. Yes it was a X99 system. This is the review:
https://www.techpowerup.com/reviews/Crucial/Ballistix_Elite_3200_DDR4/5.html

 

[jboydgolfer]

Can we just call it "special" RAM?

It's not necessarily slow it's just different

 

[EarthDog]

So I bought these G.Skill Ripjaws V's (F4-3000C15D-16GVRB) for $102.99 last Wednesday.
I did some AIDA64 Benchmarks and got 41166 MB/s Read, 43424 MB/s Write, 37763 MB/s Copy 52.0 ns Latency. Screenshot below.

Then today I read the review on the Crucial Ballistix Elite 3200 MHz DDR4 16 GB and see 71289 MB/s Read, 73537 MB/s Write, 72661 MB/s Copy and 56.2 ns Latency in AIDA64.

So should I return these and pay the 15% restocking fee and buy these Ballistix Elite for $139.99? Or these 2x8GB sticks that are DDR4 3000? Or feel free to let me know if there's some other RAM I should be buying. I originally replaced my Crucial Ballistix Sport LT DDR4 2400 32GB for these Ripjaws just to see if they are causing my random crashes every so often during long gaming sessions.

Thanks
BB

AIDA is a multi threaded memory test. Meaning, its bandwidth will scale with the number of cores/threads you have available. I am imagining that review used an octo/deca versus your quad and its of course quad channel memory on that platform.

EDIT: Ha, yeah... X99 and a deca...

Put the hex on there and watch that bandwidth drop... not to dual channel quad levels, but...

 

[arbiter]

Ah ha! That's what I was thinking might have been why those were so fast. Yes it was a X99 system. This is the review:
https://www.techpowerup.com/reviews/Crucial/Ballistix_Elite_3200_DDR4/5.html

That would be why, x99 system is quad channel memory where as the 6700k you have only dual channel.

 

[EarthDog]

Again... the cpu thread count has a lot to do with it as well.

 

[Aquinus]

It's all about memory channels by the time you get to a quad. My SB-E 3820 has a whole lot more memory bandwidth than a SB quad.

 

[EarthDog]

It's all about memory channels by the time you get to a quad. My SB-E 3820 has a whole lot more memory bandwidth than a SB quad.

...and the hex and octo core would have way more bandwidth than that.

I went through this when the 5960x was released and I had a 5820k. Even though the other reviewer and I had the exact same sticks but a different board, his octo core bandwidth was WAY more than my hex...

 

[Aquinus]

...and the hex and octo core would have way more bandwidth than that.

Actually, that hasn't been the experience that I've had, at least with SB-E. Usually bandwidth is very similar but, my one observation was that latency on the 3820 was slightly lower. All in all, more cores doesn't get you more memory bandwidth. It's more likely that it only gets you more resources to use the bandwidth you couldn't use before and even then, cache hit rates are so high that more memory bandwidth almost never yields reasonable gains so it's kind of a moot point in the first place.

Also, I'm sure that @cadaveca would tell you that the motherboard can have an impact on memory bandwidth.

 

[EarthDog]

Sure it does.... but not by much. A couple percent at best. Clearly here the difference is in the quad channel memory, and using the deca core on top of it. I don't care if it's pushing or pulling more cores give you/unlock more bandwodth. You run maxxmem on it, bandwidth won't be different... its single threaded.

When I have a moment, I'll post up some aida SS and drop the cores.

 

[Aquinus]

Sure it does.... but not by much. A couple percent at best. Clearly here the difference is in the quad channel memory, and using the deca core on top of it. I don't care if it's pushing or pulling more cores give you/unlock more bandwodth. You run maxxmem on it, bandwidth won't be different... its single threaded.

When I have a moment, I'll post up some aida SS and drop the cores.

I'm not saying that AIDA won't show a little more throughput if you have more cores. All I'm saying is that more cores doesn't suddenly make your memory go faster, it's just in a better position to utilize the bandwidth you already had.

 

[EarthDog]

We agree on that point. As I said, I don't care if its pushing or pulling, the cores, the point is, its more, mostly because of quad channel, and some because of the cores.

Now, that said, I stand partially corrected... Reads drop off a ton (not a little), otherwise, everything else is lower, but not quite what I was thinking (was closer to the reads drop) My apologies. Here are the SSs...



[/S]

 

[Aquinus]

We agree on that point. As I said, I don't care if its pushing or pulling, the cores, the point is, its more, mostly because of quad channel, and some because of the cores.

Now, that said, I stand partially corrected... Reads drop off a ton (not a little), otherwise, everything else is lower, but not quite what I was thinking (was closer to the reads drop) My apologies. Here are the SSs...

View attachment 83185

View attachment 83186[/S]

I was talking about my experience with SB-E which had less memory bandwidth than DDR4. I did some searching on old posts and found this from 3 years ago. I don't really run Windows anymore because Microsoft has to stop screwing up my machine but:

The read on yours seems rather low but, SB-E behaves a bit differently when it comes to overclocking. A newer result would require Microsoft being less douche about their update policy.

 

[Arctucas]

So I bought these G.Skill Ripjaws V's (F4-3000C15D-16GVRB) for $102.99 last Wednesday.
I did some AIDA64 Benchmarks and got 41166 MB/s Read, 43424 MB/s Write, 37763 MB/s Copy 52.0 ns Latency. Screenshot below.

Then today I read the review on the Crucial Ballistix Elite 3200 MHz DDR4 16 GB and see 71289 MB/s Read, 73537 MB/s Write, 72661 MB/s Copy and 56.2 ns Latency in AIDA64.

So should I return these and pay the 15% restocking fee and buy these Ballistix Elite for $139.99? Or these 2x8GB sticks that are DDR4 3000? Or feel free to let me know if there's some other RAM I should be buying. I originally replaced my Crucial Ballistix Sport LT DDR4 2400 32GB for these Ripjaws just to see if they are causing my random crashes every so often during long gaming sessions.

Thanks
BB

I am using
F4-3600C16D-8GTZ
which is closest to these.

 

[EarthDog]

I was talking about my experience with SB-E which had less memory bandwidth than DDR4. I did some searching on old posts and found this from 3 years ago. I don't really run Windows anymore because Microsoft has to stop screwing up my machine but:

The read on yours seems rather low but, SB-E behaves a bit differently when it comes to overclocking. A newer result would require Microsoft being less douche about their update policy.

Oh shoot.. well, the thread was about DDR4 so, I was sticking within the context.

EDIT: I bet if I clock to 4.4GHz, I would be a lot closer in reads... assuming I remember correctly that its reads that goes up with clocks... LOL

 

[eidairaman1]

Use These Timings and Voltage on that ram with the same Clock Rate (fastest for the Kit)

http://gskill.com/en/product/f4-3000c14d-16gvk

http://gskill.com/en/product/f4-3000c14d-16gtz

http://gskill.com/en/product/f4-3000c14d-16gtzr