Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Interesting: The 5,1 doesn't run a 3.46GHz at full speed!

F

fendersrule

macrumors 6502
Original poster
Oct 9, 2008
423
324
I discovered this in Windows 10, but I'm assuming the same will happen in MacOS.

Just as the title says, our 5,1s will never achieve x28 multiplier (3.73GHz) during stress testing. Stress testing the CPU with CPU-Z shows that the processor will not go past x27 (3.60GHz), even though x28 is included in the possible range.

And yes, the power options is set to High Performance and the minimum and maximum processor speed is set to 100%.

So this is interesting. h98 and myself both think that there's something in Apple's firmware that treats the 3.46GHz CPU like a 3.33GHz CPU, since the max turbo speed on a 3.33GHz is 3.60GHz. This is probably because Apple never officially supported the 3.46GHz CPU.

Just just a general FYI that your 5,1 Mac Pro will never fully utilize a X5690 or a W3690.

I can't be the first that stumbled onto this, though!
 
Quite certain I did see the maximum W3690 boost clock in CPU-Z, back when I had Windows installed on my 5,1 and was putting it through its paces.

Do you have the little palette window open that shows the live clockspeed of each individual core? By design, all of the cores can't boost to max at the same time. The single clockspeed number that's shown in the main CPU-Z window is probably some kind of average.
 
  • Like
Reactions: dabotsonline
Yep, I went off of the clocks window. See attached, that's the screen I took during the CPU-Z CPU stress test.
 

Attachments

  • 21039773_10154936007949021_467515710_n.png
    21039773_10154936007949021_467515710_n.png
    75.7 KB · Views: 1,930
Looking at that jogs my memory and I'm 98% sure I did often see 3.7 GHz during my testing. Can't prove it now, though: since building a separate Ryzen box, my cMP has no Windows.

Maybe try a much older version of CPU-Z and see if it shows any different.

p.s. I wouldn't take CPU-Z as gospel either way. It's just a 3rd party program trying the best it can.
 
My PC "Mac Pro" system which uses a W3680 with the clock set to 30x (4GHz) works as expected. It immediately moves to 30x when the CPU is pegged. I think all cores move to 30x if I remember correctly.

It's just weird that the Mac Pro with the 3.46GHz doesn't move to the 3.73GHz turbo speed according to CPU-Z...and I don't think I'm doing anything wrong! I'm using the same method with my PC which is very similar specs and it works as expected.

Any people want to take a look at this on their systems?

Frou, thanks! Maybe you can recheck this if you ever get windows reinstalled!
 
  • Like
Reactions: dabotsonline
You said you were doing the Stress Test in CPU-Z. Try looking at the clocks during the *single core* Bench.
 
Sure, I'll try "Bench CPU" instead of "Stress CPU", but I would be confused if the data changes. "Stress CPU" has no problem bringing up the max clocks in my PC immediately, so if the Mac differs that would be interesting. I'll try this later today.
 
IIRC manual BIOS overclocking changes the way Turbo works.

Without that, Intel have/had some kind of notation that specifies how Turbo varies depending on how many cores are loaded:

Screen Shot 2017-08-23 at 21.46.10.png
 
Good stuff guys. This makes sense: Sounds like "Bench CPU" may be worth trying out then! I'll certainly do that later today.
[doublepost=1503524512][/doublepost]I did notice something interesting yesterday though. You guys can maybe see why I got to this conclusion.

PC (W3680 @ 4GHz w/ 12GB 1600MHz ram): 2910 Geekbench 4 score w/ "Balanced Power Savings". I noticed that CPU-Z showed my 4GHz clock scores being hit only once in awhile.

PC (W3680 @ 4GHz w/ 12GB 1600MHz ram): 3450 Geekbench 4 score w/ "Performance Power Savings". I noticed that CPU-Z showed my 4GHz clock scores being hit nearly all the time.

Mac Pro 5,1 (X5690 w/ 24GB 1333MHz ram): 3000 Geekbench 4 score w/ "Balanced Power Savings".

Mac Pro 5,1 (X5690 w/ 24GB 1333MHz ram): 3000 Geekbench 4 score w/ "Optimal Power Savings".
 
Did "Bench CPU" instead. Same exact result.

Fired up Heaven 4.0. Never left 27x, even when only 1-2 cores were being pushed.
 
I confirmed that on my cMP. I can only reach 3.73GHz in MacOS, but not in Windows 7/8/10.

I am very awear that's single core turbo boost, and I can only stress one core during the test to achieve that. But CPU-Z (or any other software) never able to show me 3.73 in Windows, but always maxed / stopped at 3.6.

However, OSX / MacOS is fine. Not sure if X5690 or W3690 will make any difference on this matter. Or purely indication problem in Windows.
 
  • Like
Reactions: Squuiid and fendersrule
During the CPU Bench, I opened up Performance Monitor (or Activity Monitor), what ever it's called.

That was showing it pegging up to 3.70GHz.
 
  • Like
Reactions: dabotsonline
I'm almost positive I've checked my single core boost above 3.7 at times in win 10. I will confirm today

Edit: Verified
 

Attachments

  • Screenshot (21).png
    Screenshot (21).png
    218.7 KB · Views: 486
Last edited:
  • Like
Reactions: dabotsonline
This is my cMP running Windows 10 Pro. 3.7GHz
fendersrule might be having some other issue going on I think.
 

Attachments

  • cpu.PNG
    cpu.PNG
    35.1 KB · Views: 604
  • Like
Reactions: dabotsonline
ActionableMango said:
How can the maximum speed be lower than the speed? That's confusing.


Click to expand...

Exactly, and both W3690 / X5690 CANNOT run at 3.70 when ALL cores under stress (100% utilisation rate). That number is wrong. And I never see the recorded max speed in any software can go above 3.6GHz. In fact, 3.70GHz itself doesn't make any sense, these CPU should either go to 3.73 GHz (133 x 28), or 3.6GHz (133 x 27), but not stay at somewhere in between. Obviously, the CPU is not in transiting through 3.70, but stay there. It doesn't make any sense, which I believe the indication is wrong.

However, if indication is wrong, the next question will be can we expect the CPU is actually running at 3.73GHz?
 
  • Like
Reactions: dabotsonline, Synchro3 and fendersrule
Squuiid said:
This is my cMP running Windows 10 Pro. 3.7GHz
fendersrule might be having some other issue going on I think.
Click to expand...

That's exactly what mine shows, and it doesn't make any sense.

As Mango said, why is "Maximum Speed" not matching up with the higher current speed?

3.70GHz is not possible with the simple math of 133 x 27 and 133 x 28. You cannot have somewhere in between.

Looks like everyone has this issue. h98 is on the mark. Either there's something fluky with CPU-z' "clocks" indicator AND Windows's indicator, or our clocks are simply not turbo-ing to x28 in ANY circumstance.

It's less likely that two independent programs would be incorrec

This still ain't solved!

Is there another program out there that will show the multiplier besides CPU-Z? Windows right off the bat cannot be trusted.
 
  • Like
Reactions: dabotsonline
I feel confident the processor is performing as it should.
[doublepost=1503604740][/doublepost]
fendersrule said:
That's exactly what mine shows, and it doesn't make any sense.

As Mango said, why is "Maximum Speed" not matching up with the higher current speed?
Click to expand...

Maximum speed refers to the clock that ALL cores can turbo to at the same time. As stated several times already, max clocks for all cores is 3.591, however if power and thermals allow, up to two cores can reach a MAXIMUM of 3.73. I see these numbers reported slightly differently due to rounding at times depending on the program, however I am fairly certain that the windows task manager is most accurate.

h9826790 said:
Exactly, and both W3690 / X5690 CANNOT run at 3.70 when ALL cores under stress (100% utilisation rate). That number is wrong. And I never see the recorded max speed in any software can go above 3.6GHz.
Click to expand...

100% utilization doesn't forgo reaching max clock so long as the other cores are within their turbo limits, yes? ie 6 cores at 3.47 and 2 cores turboing higher.
 
thornslack said:
100% utilization doesn't forgo reaching max clock so long as the other cores are within their turbo limits, yes? ie 6 cores at 3.47 and 2 cores turboing higher.
Click to expand...

Are you sure? The X5600 series only has Turbo Boost 1.0. From the Intel documents The X5690's turbo table is 1/1/1/1/2/2.
Screen Shot 2017-08-25 at 05.13.03.jpg

And that means, when

1 core ACTIVE, max speed is 3.73 GHz
2 cores active, max speed is 3.73 GHz
but 3,4,5,6 cores active, the max speed will be 3.6GHz (on any core)

NOT all cores works, and 2 of them can go to 3.73 with remaining at 3.6 or 3.46. That's ACTIVE core. If you want 3.73GHz, you CANNOT use all cores but only 2, and the rest should be at around idle (step down to around 1.6GHz and almost no loading at all). That means, when the CPU is utilised to 100%, it's IMPOSSIBLE to get 3.73GHz.

In fact, the newer Xeon's table is easier to understanding. That's Turbo Boost 2.0. So, more powerful than 1.0, but still depends on the ACTIVE core count (and clearly stated that).
Screen Shot 2017-08-25 at 05.04.33.jpg


Also, I don't know where explain that max speed is all core active's speed. In general, that should be designed to describe the max potential of the CPU speed, that means 3.73 in this case, not 3.6.

And since Turbo boost 1.0 CANNOT allow any core to run at max speed when all cores are stressed, that means that 3.7GHz is WRONG. It's impossible according to Intel's documents.

I personally also believe the CPU is working as expected, because the Turbo Boost technology is pretty much driver / OS independent. Also, Intel also says software may reporting erroneous number.
Screen Shot 2017-08-25 at 05.16.58.jpg

However, I really can't accept that 3.7GHz is a prove that the CPU's turbo boost is working properly. It just doesn't match the Intel's official info.
 
  • Like
Reactions: dabotsonline, Synchro3, m4v3r1ck and 1 other person
iv got a w3680, in windows i see all cores clock at 3.46ghz most the time (think it's 3.46 or 3.40 something like that) iv seen it spike faster once or twice but it will only do it when only one core is being used (turbo 1.0).

to be simple the CPU's are sold as a lower speed with the potential of hinting X speed, there not sold as being at that speed.

but as a plus with all cores maxed in windows i see it hold the 3.46ghz speed and not drop to the 3.33ghz speed so i assume it works out ok
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back