K1 Denver In-order execution?

[Nintonito]

(09-25-2014, 10:30 AM)Fiora Wrote: The A7 Cyclone (from what I heard) supposedly has a ~192 uop reorder buffer, which is about as far as you can get from in-order.

Alright well my point is still half there. Focus on silicon per core rather than cores per silicon.

[KHg8m3r]

Here's something I found that could shed some light on this issue, unless you've already read it.
http://www.extremetech.com/computing/187862-nvidia-details-64-bit-denver-tegra-k1-claims-haswell-class-performance-for-first-64-bit-android-chip

The Cyclone apparently has the same amount of single core processing power as the Celeron 2955U, and the Denver K1 is still faster than that.

[EverythingPortable]

Just read this in one of the article comments, dosnt sound good for dolphin:

"Errm, it's a dual-core chip, and there's no third core for running the optimizations. They run on the same CPU cores that everything else does.
It's a dual core chip with one core dedicated to doing the optimizations and the other for running the code."

Is that why nvidias own benchmark didnt show any geekbench 3 dual-core result? Will only one core run dolphin(or any other app) on k1 denver devices?

[Nintonito]

(09-25-2014, 07:04 PM)EverythingPortable Wrote: Just read this in one of the article comments, dosnt sound good for dolphin:

"Errm, it's a dual-core chip, and there's no third core for running the optimizations. They run on the same CPU cores that everything else does.
It's a dual core chip with one core dedicated to doing the optimizations and the other for running the code."

Is that why nvidias own benchmark didnt show any geekbench 3 dual-core result? Will only one core run dolphin(or any other app) on k1 denver devices?

I sincerely doubt they need an entire core dedicated to cache optimized code blocks. It's most likely just distributed throughout cycles, or offloaded to a low power controller.

[NZtechfreak]

(09-25-2014, 08:41 PM)Nintonito Wrote:

(09-25-2014, 07:04 PM)EverythingPortable Wrote: Just read this in one of the article comments, dosnt sound good for dolphin:

"Errm, it's a dual-core chip, and there's no third core for running the optimizations. They run on the same CPU cores that everything else does.
It's a dual core chip with one core dedicated to doing the optimizations and the other for running the code."

Is that why nvidias own benchmark didnt show any geekbench 3 dual-core result? Will only one core run dolphin(or any other app) on k1 denver devices?

I sincerely doubt they need an entire core dedicated to cache optimized code blocks. It's most likely just distributed throughout cycles, or offloaded to a low power controller.

I would think it still has a 'companion core' for that, like how the 'quadcore' has a low-clocked 5th core for battery savings for lower-level computing duties. Must check on that, usually Nvidia makes a big deal about the battery saving core, so if present would have expected them to have mentioned it.

[Nintonito]

(09-26-2014, 06:49 AM)NZtechfreak Wrote:

(09-25-2014, 08:41 PM)Nintonito Wrote:

(09-25-2014, 07:04 PM)EverythingPortable Wrote: Just read this in one of the article comments, dosnt sound good for dolphin:

"Errm, it's a dual-core chip, and there's no third core for running the optimizations. They run on the same CPU cores that everything else does.
It's a dual core chip with one core dedicated to doing the optimizations and the other for running the code."

Is that why nvidias own benchmark didnt show any geekbench 3 dual-core result? Will only one core run dolphin(or any other app) on k1 denver devices?

I sincerely doubt they need an entire core dedicated to cache optimized code blocks. It's most likely just distributed throughout cycles, or offloaded to a low power controller.

I would think it still has a 'companion core' for that, like how the 'quadcore' has a low-clocked 5th core for battery savings for lower-level computing duties. Must check on that, usually Nvidia makes a big deal about the battery saving core, so if present would have expected them to have mentioned it.

Well normally you don't advertise things that are supposed to be unnoticeable. Nvidia claims the caching is entirely driven hardware side, so it would make sense for them to kept it opaque. Nobody will need to develop with that in mind (unlike the 5th companion core, which encouraged the OEMs and rom debs to make governor modifications to use said core under certain situations at all times.) Since this isn't meant to require any additional effort, they don't need to complicate things. At any rate, a teardown will answer all questions.

[NZtechfreak]

(09-26-2014, 07:10 AM)Nintonito Wrote:

(09-26-2014, 06:49 AM)NZtechfreak Wrote:

(09-25-2014, 08:41 PM)Nintonito Wrote:

(09-25-2014, 07:04 PM)EverythingPortable Wrote: Just read this in one of the article comments, dosnt sound good for dolphin:

"Errm, it's a dual-core chip, and there's no third core for running the optimizations. They run on the same CPU cores that everything else does.
It's a dual core chip with one core dedicated to doing the optimizations and the other for running the code."

Is that why nvidias own benchmark didnt show any geekbench 3 dual-core result? Will only one core run dolphin(or any other app) on k1 denver devices?

I sincerely doubt they need an entire core dedicated to cache optimized code blocks. It's most likely just distributed throughout cycles, or offloaded to a low power controller.

I would think it still has a 'companion core' for that, like how the 'quadcore' has a low-clocked 5th core for battery savings for lower-level computing duties. Must check on that, usually Nvidia makes a big deal about the battery saving core, so if present would have expected them to have mentioned it.

Well normally you don't advertise things that are supposed to be unnoticeable. Nvidia claims the caching is entirely driven hardware side, so it would make sense for them to kept it opaque. Nobody will need to develop with that in mind (unlike the 5th companion core, which encouraged the OEMs and rom debs to make governor modifications to use said core under certain situations at all times.) Since this isn't meant to require any additional effort, they don't need to complicate things. At any rate, a teardown will answer all questions.

Agreed, this is all speculation at this point and it really won't be until we see the actual devices in use that we know whether Nvidia has successfully implemented things here.