How much faster will 8-core iMac Pro than iMac i7

[FredT2]

I've never really understood the relationship between the nominal speed rating and turbo-boost of an Intel processor. The 8-core iMac Pro is rated at 3.2/4.2 and the iMac i7 at 4.2/4.5. From those specs I would think that the i7 will be much faster than the iMac Pro for tasks that don't use all of the processors. But from experience I know that it's not that simple. My example:

When I bought a 2014 iMac i7 (4.0/4.2) I expected that it would be much faster the my Mac Mini (2.6/3.4), but it isn't. When I run Handbrake, the Mini runs all four cores full time at 3.4 GHz, it's boost speed. The iMac, however, after a very short time slows to to 3.8 GHz, slower even than it's nominal speed. It's faster than the Mini, but not nearly as much as I had expected.

So I wonder: when do the processors run at their nominal speed versus boost speed? I realize that at this point there's no way to know what the iMac Pro will do, but based upon how these processors normally behave, what can be expected?

(Oops, I didn't edit the title very carefully, sorry)

 

[Trebuin]

At little slower for single core which would affect games, but the graphics card would reverse that significantly. About 60-80% faster than the 4 core for encoding, encrypting, rendering, photoshop & other CPU related tasks.

 

[FredT2]

At little slower for single core which would affect games, but the graphics card would reverse that significantly. About 60-80% faster than the 4 core for encoding, encrypting, rendering, photoshop & other CPU related tasks.

So when would the 8-core run at 3.2 GHz versus the i7 running at 4.2?

 

[Trebuin]

The way it works is it will run at 4.2 to do short fast tasks, but longer jobs would case the cpu to drop the clock with the temperature to 3.2.

With multicores, you have to drop the speed with more cores to keep in the thermal design limits. Just check the iMac pro design options.

You can compare simple math by multiplying the clock by cores, but it is usually about 15-30% slower than that On multi threaded jobs. Anything that runs single threaded will actually suffer more significantly, which includes many games.

For example
3.2GHz 8 core vs 3.0 10 core
They are about the same at single threaded jobs, but the 10 core is 17% faster at multi threaded jobs.

 

[jerwin]

I've never really understood the relationship between the nominal speed rating and turbo-boost of an Intel processor. The 8-core iMac Pro is rated at 3.2/4.2 and the iMac i7 at 4.2/4.5. From those specs I would think that the i7 will be much faster than the iMac Pro for tasks that don't use all of the processors. But from experience I know that it's not that simple. My example:

If you have both machines at hand, and have Intel Power Gadget installed on both, maybe you could run a few handbrake benchmarks on both.. Post times and graphs here.
[doublepost=1513281570][/doublepost]

The way it works is it will run at 4.2 to do short fast tasks, but longer jobs would case the cpu to drop the clock with the temperature to 3.2.

the Core M machines are an extreme example of this... I don't imagine that the Xeon-W's power management is anywhere near as sophisticated.

 

[thefriendshipmachine]

hard to say since we havent seen how much it throttles under heavy sustained load.

 

[T'hain Esh Kelch]

I've never really understood the relationship between the nominal speed rating and turbo-boost of an Intel processor. The 8-core iMac Pro is rated at 3.2/4.2 and the iMac i7 at 4.2/4.5. From those specs I would think that the i7 will be much faster than the iMac Pro for tasks that don't use all of the processors. But from experience I know that it's not that simple. My example:

4.5 Ghz is not that much faster than 4.2 Ghz. And there are not many application which could need a full core over an extended time, which is also not optimized for multiple cores, so in real life you'll never feel the difference.

When I bought a 2014 iMac i7 (4.0/4.2) I expected that it would be much faster the my Mac Mini (2.6/3.4), but it isn't. When I run Handbrake, the Mini runs all four cores full time at 3.4 GHz, it's boost speed. The iMac, however, after a very short time slows to to 3.8 GHz, slower even than it's nominal speed. It's faster than the Mini, but not nearly as much as I had expected.

It may down throttle to 3.8 Ghz due to thermal constraints in the iMac body. You can try to lower the ambient temperature and see if it can't hold the full speed longer.

And you are then looking at 3.8 Ghz vs. 3.4 Ghz, which again, is not much of a difference. So you are seeing expected results. While the processor in the iMac is capable of hitting 4.2 Ghz, Apples tight enclosure likely does not allow it to run at that temperature for extended periods. And Handbrake sure gives it a run for its money, since it cranks everything up to 11 during encodes.

 

[ddublu]

I'm also trying to determine the gains with an upgrade. I have a Late 2015 5k with 4ghz, 64gb RAM, 1TB ssd, AMD Radeo R9 4Gb.

 

[joema2]

hard to say since we havent seen how much it throttles under heavy sustained load.

https://arstechnica.com/gadgets/201...ilable-heres-how-people-are-already-using-it/

"The company claims...when running in a room at normal temperatures, little to no performance throttling should occur....Since the iMac Pro now uses only flash storage, the space previously occupied by the standard HDD has been used for heat management...."

"...We observed one machine running numerous, CPU-intensive virtualization tasks...Even with all 10 cores in heavy use over several hours, the aluminum body felt no warmer to touch than room temperature....We could hear the fans running in a quiet room, but it seemed comparable to the existing iMac and would usually not be audible if music or conversations were going on nearby..."

 

[FredT2]

I'm beginning to think that the 8-core may not be a very good choice compared with the 10-core. Intel shows the 8-core at 3.7/4.5 while Apple has it at 3.2/4.2. The 10-core is 3.3/4.5 Intel and 3.0/4.5 Apple. Or maybe I'm just overthinking it.
[doublepost=1513300319][/doublepost]

https://arstechnica.com/gadgets/201...ilable-heres-how-people-are-already-using-it/

"The company claims...when running in a room at normal temperatures, little to no performance throttling should occur....Since the iMac Pro now uses only flash storage, the space previously occupied by the standard HDD has been used for heat management...."

"...We observed one machine running numerous, CPU-intensive virtualization tasks...Even with all 10 cores in heavy use over several hours, the aluminum body felt no warmer to touch than room temperature....We could hear the fans running in a quiet room, but it seemed comparable to the existing iMac and would usually not be audible if music or conversations were going on nearby..."

Great article!

 

[jerwin]

avx-512 may be one of those things that separates an ordinary i5 or i7 based imac from the Xeon based imac pro. We'll have to see.