RAM bin performance

[Letgomyleghoe]

Is their a way to measure each BIN's performance? i know about total latency, and the formula for it, but is that a proper way to measure the performance of the bin?

[Grabhanem]

I'm not an expert but I'd assume it depends too heavily on workload & CPU to really have a universal standard-- for example Cinebench barely cares about memory at all, while APU graphics have a massive performance boost from memory speed.

[Letgomyleghoe]

10 minutes ago, Grabhanem said:

I'm not an expert but I'd assume it depends too heavily on workload & CPU to really have a universal standard-- for example Cinebench barely cares about memory at all, while APU graphics have a massive performance boost from memory speed.

thats comparing a workload to a component, not two different workloads. 

 

cinebench not using memory doesnt mean its different for different workloads, if it were on the other to benefit more from CL latency's than speed that would be different.

[Jurrunio]

performance, as in actual bandwidth and latency? The CPU makes a difference here and I dont think there are calculators for the whole thing.

[Letgomyleghoe]

Just now, Jurrunio said:

performance, as in actual bandwidth and latency? The CPU makes a difference here and I dont think there are calculators for the whole thing.

damb i was gonna make tier list based off first word latency,

 

so cpus benefit from CL latency and bandwidth differently?

 

ik ryzen for instance benefits off better ram a lot more than an intel system, but do they benefit off CL and bandwidth differently than each other?

 

 

[Jurrunio]

16 minutes ago, Letgomyleghoe said:

so cpus benefit from CL latency and bandwidth differently?

Yes. Intel's ring bus CPUs care less about them than mesh bus CPUs (Skylake X) or Ryzen. as a result improving bandwidth by x% and performance gain of y% on Ryzen does not mean x% gain on bandwidth on Intel Comet Lake will also give you y% performance gain.

[Letgomyleghoe]

23 minutes ago, Jurrunio said:

as a result improving bandwidth by x% and performance gain of y% on Ryzen does not mean x% gain on bandwidth on Intel Comet Lake will also give you y% performance gain.

could the first word latency still be used as a measure for performance, or is their something else you recommend?

[Jurrunio]

1 hour ago, Letgomyleghoe said:

could the first word latency still be used as a measure for performance, or is their something else you recommend?

Not really. You can use it to judge a memory kit and say X is generally better than Y, but that doesnt mean X has to beat Y in an actual workload. I dont think a tier list make sense for memory.

[Letgomyleghoe]

Just now, Jurrunio said:

Not really. You can use it to judge a memory kit and say X is generally better than Y, but that doesnt mean X has to beat Y in an actual workload. I dont think a tier list make sense for memory.

it ticks me off every time i see someone trying to match a 5600x or whatever with 3000mhz cl18 it genunely bugs me how much i see it

[Jurrunio]

5 minutes ago, Letgomyleghoe said:

it ticks me off every time i see someone trying to match a 5600x or whatever with 3000mhz cl18 it genunely bugs me how much i see it

I doubt that will happen nowadays when 2x8gb 3200c16 is a few percent more expensive than the cheapest 2x8gb set in the US, tho harder to say for other countries.

[Letgomyleghoe]

Just now, Jurrunio said:

I doubt that will happen nowadays when 2x8gb 3200c16 is a few percent more expensive than the cheapest 2x8gb set in the US, tho harder to say for other countries.

i just think alot of people look at the speed and call it good, and dont realize theyre about to throw some not so good ram in their system