Atom T5500/T5700 (Broxton): ~50% improvement over Cherry Trail

[Sweepr]

New hope for budget desktops/notebooks? Looks like the Goldmont core brings a substantial improvement to CPU performance.

56% faster than Airmont at similar clocks @ SPECint_2000 according to Intel:

An interesting talk regarding the IoT aspects of Intel’s Next Generation Atom Core, Goldmont, and the Broxton SoCs for the market offered a good chunk of information regarding the implementations of the Broxton-M platform. Users may remember the Broxton name from the cancelled family of smartphone SoCsseveral months ago, but the core design and SoC integration is still mightily relevant for IoT and mini-PCs, as well as being at the heart of Intel’s new IoT platform based on Joule which was announced at IDF this week as well.

Broxton in the form that was described to us will follow the previous Braswell platform in the sense that it runs in a quad-core configuration (for most SKUs) consisting of two sets of dual cores sharing a common L3 cache, but will be paired with Intel’s Gen9 graphics. The GPU configuration will be either in a 12 execution unit (EU) or 18 EU format, suggesting that the core will have a single graphics slice and will implement Intel’s binning strategy to determine which EUs are used on each sub-slice.

www.anandtech.com/show/10576/more-details-on-broxton-quad-core-ecc-up-to-18-eus-of-gen9
http://liliputing.com/2016/08/intel...cherry-trail-designed-different-products.html

 

[SAAA]

Uh... that's coming out of nowhere, I tought 30% at best from Geekbench leaks yet now they claim over 50% and similar clocks.
If the 14nm+ process advantages carry on here we might also see some 3GHz quads: Core-level in tablets here we are? Well like core-M didn't do that already.

 

[Sweepr]

Bodes well for Apollo Lake.

On the graphics side they are claiming up to 50% better 3D performance and 60-80% higher bandwidth, but I think the CPU improvement is what most were looking forward to.

http://cdn.liliputing.com/wp-content/uploads/2016/08/atom-2016_03.jpg
http://cdn.liliputing.com/wp-content/uploads/2016/08/atom-2016_04.jpg

 

[frozentundra123456]

Too little too late. They have abandoned phones and the tablet market is saturated. This is what cherry trail should have been.
It is a nice technical improvement (dependent on more testing in a wide range of applications), but what do they put it in? Maybe 300 or 400 dollar tablets, but still not sure I would pay more than half that for anything with atom in it.

 

[Azuma Hazuki]

This will probably end up in those God-awful sub-US$300 "value" desktops at Worst Buy. ...well, whatever, something with near the performance of a 65nm Core 2 Quad on the power budget of a night light isn't a bad thing.

 

[VirtualLarry]

something with near the performance of a 65nm Core 2 Quad on the power budget of a night light isn't a bad thing.

True, true.

Here's hoping that if they do do that, that they will at least make them USFF / Mini-PCs. No need for a big tower if it's going to end up empty inside.

Then again, some people still buy big ole buicks ("boats").

 

[imported_bman]

Look good for Compute Sticks, Netbooks, and Mini-PCs. I wonder if or when Intel is going to support UFS 2.0, it would significantly improve the products that these processors are going to be used in over eMMC.

 

[Sweepr]

I'd love to see a Surface 4 based on Atom T5700 (Broxton) or Pentium N4200 (Apollo Lake). Considering some of the best selling budget laptops are still on Bay Trail-M, we can expect a significant boost in terms of CPU/iGPU performance if they update those with dual-core/quad-core Apollo Lake.

 

[jihe]

Yeah but it's going to cost 300% more

 

[psolord]

Should I wait for this, or build my mini PC on AMD's 5350?

 

[dark zero]

The hell Intel! You have something comparable to Mongoose and you quit the mobile competition... Really?

What Intel killed mobile wasn't Moorefield... Was SOFIA. That was the worst move for them.... Broxton was even better than A72 levels... Near Kryo levels.

 

[FIVR]

This isn't anywhere near mongoose performance and it is likely being tested at 2-3x the power budget and TDP of a mongoose-powered device. 30-50% is relatively impressive for intel, sadly for them they've improved so little since 2013 that their competition is now hovering around 100-200% faster in IPC terms.


It is good that they are improving though. It could be worse, like when they released atom on 14nm and it somehow had worse CPU performance than it did on 22nm.

 

[ninaholic37]

Bodes well for Apollo Lake.

On the graphics side they are claiming up to 50% better 3D performance and 60-80% higher bandwidth, but I think the CPU improvement is what most were looking forward to.

http://cdn.liliputing.com/wp-content/uploads/2016/08/atom-2016_03.jpg
http://cdn.liliputing.com/wp-content/uploads/2016/08/atom-2016_04.jpg

Oh, just noticed they went from 4-6W to 6-12W TDP, so 50% improvement makes sense. Intel you comedian!

 

[mikk]

Oh, just noticed they went from 4-6W to 6-12W TDP, so 50% improvement makes sense. Intel you comedian!

This is a wrong assumption, so you are a comedian.

 

[Sweepr]

Oh, just noticed they went from 4-6W to 6-12W TDP, so 50% improvement makes sense. Intel you comedian!

Look at the second slide. Performance improves at equal power as well.

 

[SAAA]

Looks like someone's testing Broxton on Android:
http://browser.primatelabs.com/geekbench3/7352874

Also it's interesting to compare Geekbench itself when running the same cpu on different os:
http://browser.primatelabs.com/geekbench3/compare/6034406?baseline=7352874

The Android run is 64bit vs 32bit for the Windows test, it has better ram too, but there are some weird differences among scores... sometimes 50% more others it's half. Guess it has something to do with the particular instructions used.
It also makes me wonder what will happen with the incoming 4th version of the benchmark and if it will be a fairer comparison between desktop and mobile platforms.

 

[mikk]

Look at the second slide. Performance improves at equal power as well.

These T-SKUs in the slides have a TDP of 4W according to Intel.

 

[Nothingness]

It also makes me wonder what will happen with the incoming 4th version of the benchmark and if it will be a fairer comparison between desktop and mobile platforms.

The desktop and mobile platforms will use the same datasets so comparisons should be more fair.

 

[ehume]

Well, for those of us who use our laptops infrequently -- we can't justify anything but a cheap device. So this is welcome news, especially since I'll use this for perusing pdf's. Nice update, and just in time for the after-Christmas sales.

 

[dark zero]

Looks like someone's testing Broxton on Android:
http://browser.primatelabs.com/geekbench3/7352874

Also it's interesting to compare Geekbench itself when running the same cpu on different os:
http://browser.primatelabs.com/geekbench3/compare/6034406?baseline=7352874

The Android run is 64bit vs 32bit for the Windows test, it has better ram too, but there are some weird differences among scores... sometimes 50% more others it's half. Guess it has something to do with the particular instructions used.
It also makes me wonder what will happen with the incoming 4th version of the benchmark and if it will be a fairer comparison between desktop and mobile platforms.

Is near A72 levels it seems... Is not bad... Intel should try to continue to enter..

 

[Nothingness]

Is near A72 levels it seems... Is not bad... Intel should try to continue to enter..

Closer but not at A72 level yet: https://browser.primatelabs.com/geekbench3/compare/6064783?baseline=7352874

 

[dark zero]

Closer but not at A72 level yet: https://browser.primatelabs.com/geekbench3/compare/6064783?baseline=7352874

So.. ARM A72 are on Intel Core 2 Duo levels?

 

[Nothingness]

So.. ARM A72 are on Intel Core 2 Duo levels?

Yes, but C2D has a big advantage for many: it can run Windows

 

[dark zero]

Yes, but C2D has a big advantage for many: it can run Windows

Windows 10 Mobile wants to talk to you with their Continuum... that was the worst mistake Microsoft did... if they wanted to succeed, they should put a x86 processor only into a initially Old Blackberry like devices (with keyboard and trackball)

 

[Raduque]

Isn't that almost Core m performance?