MBP 2016 owners: RAM bandwidth

[leman]

As our new MBPs have not arrived yet, could some MBP 2016 owners (doesn't matter which model) run a RAM benchmark using http://www.memtest86.com and report the results (and your machine)? The memtest86 is a bit more cumbersome to run (you need to restore the image to a USB stick and boot from there), but it provides accurate results. It would be really cool of some of you guys would find the time to do it!

Explanation: I am confused about the RAM benchmarks results in Geekbench — they seem to be significantly worse than the previous-gen MBP. I'd like to know whether its a problem with Geekbench or whether the RAM is indeed slower.

 

[monkeydax]

As our new MBPs have not arrived yet, could some MBP 2016 owners (doesn't matter which model) run a RAM benchmark using http://www.memtest86.com and report the results (and your machine)? The memtest86 is a bit more cumbersome to run (you need to restore the image to a USB stick and boot from there), but it provides accurate results. It would be really cool of some of you guys would find the time to do it!

Explanation: I am confused about the RAM benchmarks results in Geekbench — they seem to be significantly worse than the previous-gen MBP. I'd like to know whether its a problem with Geekbench or whether the RAM is indeed slower.

Yeah i'm kinda wondering also, what's up with the RAM bandwidth scores we're getting on Geekbench??

 

[Ries]

Yeah i'm kinda wondering also, what's up with the RAM bandwidth scores we're getting on Geekbench??

No Iris Pro aka embedded RAM to cache the iGPU. The iGPU needs bandwidth to work now.

 

[sleim031]

Taking this from another thread, it seems that HD 530 performs better in benchmarks than Iris Pro from the previous 15"

https://browser.primatelabs.com/opencl-benchmarks

 

[fokmik]

Taking this from another thread, it seems that HD 530 performs better in benchmarks than Iris Pro from the previous 15"

https://browser.primatelabs.com/opencl-benchmarks

i dont find the HD 530
the Intel(R) Iris(TM) Pro Graphics 6200 has a score 25939

 

[monkeydax]

i dont find the HD 530
the Intel(R) Iris(TM) Pro Graphics 6200 has a score 25939

Intel HD Graphics 530 - 17238

Intel(R) Iris(TM) Pro Graphics 5200
- 12722

Do a cmd+f to find these. The Haswell processors only have the Iris Pro 5200 and not the 6200. So hmm. HD Graphics 530 > Iris Pro 5200?

 

[fokmik]

Intel HD Graphics 530 - 17238

Intel(R) Iris(TM) Pro Graphics 5200
- 12722

Do a cmd+f to find these. The Haswell processors only have the Iris Pro 5200 and not the 6200. So hmm. HD Graphics 530 > Iris Pro 5200?

you own already an 15" MBP? if yes, how is your battery? on surfing the web for example

 

[leman]

Intel HD Graphics 530 - 17238

Intel(R) Iris(TM) Pro Graphics 5200
- 12722

Do a cmd+f to find these. The Haswell processors only have the Iris Pro 5200 and not the 6200. So hmm. HD Graphics 530 > Iris Pro 5200?

There is obviously either an issue with the driver or an issue with geekbench. There is no way that hd 530 is faster than 5200

 

[monkeydax]

you own already an 15" MBP? if yes, how is your battery? on surfing the web for example

Nope I don't, still trying to figure out if the 2016 model is any much better than the mid 2015 one!
[doublepost=1479211823][/doublepost]

There is obviously either an issue with the driver or an issue with geekbench. There is no way that hd 530 is faster than 5200

We'll see if more benchmarks come in, but for now, could it be the two generations later HD Graphics 530 almost catches up with the Iris Pro just for this OpenCL test?

 

[leman]

Nope I don't, still trying to figure out if the 2016 model is any much better than the mid 2015 one!
[doublepost=1479211823][/doublepost]
We'll see if more benchmarks come in, but for now, could it be the two generations later HD Graphics 530 almost catches up with the Iris Pro just for this OpenCL test?

Any test they shows 530 faster than 5200 is a crappy test and should be discarded from making any kind of judgement. Of course ther might be isolated things where 530 is faster because of its newer architecture, but Geekbench strives to be representative.