Resolved Mac Pro Geekbench Drop

[mgb3558]

Just purchased a Mac Pro 5,1 and decided to make some upgrades to it. After changing out the CPUs for the dual six core 3.46 I got a score of 18,163. Pretty awesome. Next step, upgrade the RAM and try and get that score over 20,000! Changed out the 8GB of 1066 RAM (4 modules) for 16GB of 1333 RAM (1 module OWC).

Performed SMC/PRM reset, system recognized all the changes, yet upon doing the geekbench test again, my score fell off the cliff down to 15,122. . . Nothing is standing out as to what the issue could be. And absolutely nothing really explains the massive drop especially since geekbench is seeing the RAM.

Any ideas?

 

[Dr. Stealth]

Not sure what your issue could be but for reference I'm running the same two processors and here is my GB 4 score.

So I guess the bad news is you need to figure out what the issue is.

But the good news is once you do you will be very happy with your x5690's.

 

[Fooze]

The CMP will achieve the highest Geekbench scores when slots 1-3 and 5-7 are populated with RAM sticks. If you only have 1 of the 8 slots populated then you will notice a huge performance hit in these synthetic benchmarks. Real world performance is much less noticeable.

 

[leon771]

Seems to me that it is because you only have a single RAM slot populated.
Buy more RAM.

 

[mgb3558]

The CMP will achieve the highest Geekbench scores when slots 1-3 and 5-7 are populated with RAM sticks. If you only have 1 of the 8 slots populated then you will notice a huge performance hit in these synthetic benchmarks. Real world performance is much less noticeable.

Thanks, that must be my problem. I was stressing out that maybe I applied the thermal paste incorrectly or a hundred other different issues. I installed the old 4 RAM modules back in and the Geekbench Score shot back up. That had me worrying that maybe the new RAM was defective! Good to know it's just a superficial hit and everything is working how it should. I guess I will have to wait to flaunt the Geekbench score until I grab a couple more modules.

 

[DPUser]

Check out used server RAM on eBay... thoroughly tested (!!!) and a lot less expensive.

 

[SoyCapitanSoyCapitan]

1 module doesn’t have enough bandwidth to feed 12 cores (or even less)

 

[carpsafari]

I installed the old 4 RAM modules back in and the Geekbench Score shot back up. That had me worrying that maybe the new RAM was defective! Good to know it's just a superficial hit and everything is working how it should.

And the cause for it is the crappy OWC ram.
Needs to be told.

 

[frou]

When you see talk of CPUs/Motherboards being "dual-channel", "triple-channel" etc, this is the phenomenon that's referring to.

 

[LightBulbFun]

ok i see a couple things amiss

the reason your score is bad is because your running just 1 memory stick, for maximum performance your meant to run 3 sticks per CPU, I dont recommend OWC ram as well, you can get suitable RAM for silly cheap on ebay thanks to all the used servers being parted out

secondly your boot ROM is VERY out of date (as in even before High sierra your boot rom is out of date) https://support.apple.com/kb/DL1321?locale=en_US download this then once thats done when you go to install High sierra install its Boot ROM update

 

[mp2017]

This is why Geek Bench scores have little value outside of benchmarking.

 

[flowrider]

^^^^What is why?

Lou

 

[mp2017]

^^^^What is why?

Lou

That a more capable system scores lower on the benchmark than a more capable one.

 

[LightBulbFun]

That a more capable system scores lower on the benchmark than a more capable one.

your ignoring the fact geekbench also benchmarks the memory, and in this case the system is severely limited by having only 1 RAM stick between 2 high bandwidth CPUs...

 

[mp2017]

your ignoring the fact geekbench also benchmarks the memory, and in this case the system is severely limited by having only 1 RAM stick between 2 high bandwidth CPUs...

I am doing no such thing. It's the reason I said "little value outside of benchmarking". If the goal is to obtain the highest Geekbench score than the four 2GB modules should be used. If the goal is to run applications then the single 16GB module is the better choice.

 

[LightBulbFun]

I am doing no such thing. It's the reason I said "little value outside of benchmarking". If the goal is to obtain the highest Geekbench score than the four 2GB modules should be used. If the goal is to run applications then the single 16GB module is the better choice.

if 1 16GB module is so much better then why did anyone bother with multi channel memory and such? if that application is very memory intensive (which if it uses 16GB of ram it probably is) its going to be severally hurt by the reduced bandwidth that running with 1 16GB RAM stick is going to give. (which is what geekbench is showing)

 

[mp2017]

if 1 16GB module is so much better then why did anyone bother with multi channel memory and such? if that application is very memory intensive (which if it uses 16GB of ram it probably is) its going to be severally hurt by the reduced bandwidth that running with 1 16GB RAM stick is going to give. (which is what geekbench is showing)

All else being equal configuring for multiple channels is preferred. However, outside of benchmarks and some very specific cases (of which I can't think of off the top of my head but I'm sure there are some), such configurations don't offer much in the way of real world performance increases.

For situations where 8GB of memory is insufficient a single 16GB module is preferable to four 2GB modules. The lack of memory and the corresponding thrashing to disk will erase any advantage gained by configuring insufficient memory capacity in order to obtain a multiple channel configuration. In an ideal configuration the OP would use four 4GB modules. I suspect he did not as he wants the ability to increase his memory in the future without having to discard memory he already purchased.