Production version of 6 core nMP Geekbenches much lower

clamnectar said:

Check this out:
http://browser.primatelabs.com/geekbench3/compare/179463?baseline=283619

Many of the single core benchmarks are 50% (!!!) slower on the official MacPro6,1 release than the earlier version (see different BIOS) codenamed AAPLJ90,1.

However, the multi-core benchmarks are slightly better.

Anyone want to take a stab at explaining this?

----------

It looks kinda like turbo isn't working

Click to expand...

Something was probably amiss when the single core process was running. Frankly the system could have been doing something at the same time, hitting the same core and driven the results down. Just doing some basic math, there is no way that a single process could only score 2700 and somehow 6 cores would do almost 8 times the work at 20200 right? It takes several runs of any benchmarking tool before you get consistent results.

Some of the results shown are definitely hacks. Small differences can be picked up in the data reported to different fields, such as the OS version, motherboard ID, memory speed, etc.

Except Macworld managed 3599 with their 8 core model (which should provide very similar single threaded performance). I'm assuming there's something wrong with the scores you're seeing.

http://www.macworld.com/article/2082022/the-new-mac-pro-first-impressions.html

clamnectar said:

There are several runs showing exactly the same thing. Just search for 1650 v2 and you'll find them.

There are also several other runs with a different BIOS that show a single score of around 3300, and a multi score of around 18000. Go figure.

Click to expand...

I know that Apple released an updated firmware already. It's possible that under one they throttle the single core output to maximize multi-core.... Pure conjecture on my part. It's still too early to take look at these numbers as anything more than a basic benchmark. The Mac Pro isn't even in end users hands (only reviewers) at this point. Being a completely new computer, there are still many software issues to work out between now and January 1.....

clamnectar said:

That's not a hack. The motherboard and BIOS both say Apple, and reflect what you'd expect.

Ohhhh I just noticed… it was run on 10.9.2. Test build of the OS. That makes sense now.

Click to expand...

And there is no build number for 10.9.2. It should be reported as: Mac OS X 10.9.2 (Build 13C32)

Crappy thermal paste application if it's not faked. Perhaps the staff in the states constructing it should visit artic silvers website and learn about it!

Gav Mack said:

Crappy thermal paste application if it's not faked. Perhaps the staff in the states constructing it should visit artic silvers website and learn about it!

Click to expand...

Geekbench has no thermal impact to turbo boost in any way. It's a sprint. You need a marathon.

theSeb said:

Geekbench has no thermal impact to turbo boost in any way. It's a sprint. You need a marathon.

Click to expand...

I've seen rubbish paste application do the same on MBP's far too often. There was a chap who had two almost identical MBP on that forum who couldn't explain why the late 2011 was slower than the early a week or two ago. Turns out one of his fans in the late 2011 had fluff in it. Imagine its rubbish over the one core that is running turbo goes high for a second or two then scales back the overclock.

Gav Mack said:

I've seen rubbish paste application do the same on MBP's far too often. Imagine its rubbish over the one core that is running turbo goes high for a second or two then scales back the overclock.

Click to expand...

I suppose when it comes to this sort of thing anything is possible. My mind tends to think in binary by default

theSeb said:

I suppose when it comes to this sort of thing anything is possible. My mind tends to think in binary by default

Click to expand...

When it comes to applying thermal paste both the manufacturers and the 'genius' frankly do not have a bloody clue applying it to Apple products.

Gav Mack said:

When it comes to applying thermal paste both the manufacturers and the 'genius' frankly do not have a bloody clue applying it to Apple products.

Click to expand...

Considering this is the 1st Apple product to get the Austin Texas treatment, I have my fingers crossed.

handheldgames said:

Considering this is the 1st Apple product to get the Austin Texas treatment, I have my fingers crossed.

Click to expand...

I have more optimism because of that, though I have seen no change in thermal policy in Apples service manuals for 4-5 years. Glop a load on - job done is what they say

Does the amount of memory make a difference?

clamnectar said:

Check this out:
http://browser.primatelabs.com/geekbench3/compare/179463?baseline=283619

Many of the single core benchmarks are 50% (!!!) slower on the official MacPro6,1 release than the earlier version (see different BIOS) codenamed AAPLJ90,1.

However, the multi-core benchmarks are slightly better.

Anyone want to take a stab at explaining this?

----------

It looks kinda like turbo isn't working[/

Click to expand...

Real world testing of the real thing will be most telling.

You will have to wait and see when the early adopters will start to post their benchmarks then we will be able to see it
For now people are still working with preproduction units