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

[SPBHM]

the previous Atom was very disappointing and Bay trail didn't age to well, 50% boost sounds about what you need to make it ok.

 

[Abwx]

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

I guess that it s the same kind of "TDP" as cherry trail s, 6W officialy but measured at more than 10W by hardware.fr who stated that the TDP is underestimated...

As for the new iterations the "improvement" in geekbench comes mainly from AES, the other figures published by Intel are related to Spec_int rate, wich is a measure of bandwith more than anything else, the integer performance in SPEC is related to Spec-int 2006 CPU, not the first time that Intel use this bench misleadingly.

• The SPECspeed® metrics (e.g., the SPECint® 2006 benchmark) are used for comparing the ability of a computer to complete single tasks.
• The SPECrate® metrics (e.g., the SPECint®_rate 2006 benchmark) measure the throughput or rate of a machine carrying out a number of tasks.

https://www.spec.org/cpu2006/

 

[hemedans]

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

not all a72 deliver that kind of perfomance, check sd650

https://browser.primatelabs.com/geekbench3/7878018

 

[ShintaiDK]

Isn't that almost Core m performance?

Not even close.

 

[Sweepr]

I guess that it s the same kind of "TDP" as cherry trail s, 6W officialy but measured at more than 10W by hardware.fr who stated that the TDP is underestimated...

Funny because according to Hardware.fr itself they are more efficient than Jaguar/Puma parts, must be one of the reasons why they killed cat cores on desktops (no contrarevenue here). And now Goldmont, which has no competition at all on x86 (Stoney Ridge is 15W TDP), should bring another leap in terms of performance and performance-per-watt to Intel's small cores.

 

[Nothingness]

not all a72 deliver that kind of perfomance, check sd650

https://browser.primatelabs.com/geekbench3/7878018

And most likely not all Goldmont will deliver the same performance that was quoted

 

[Lodix]

not all a72 deliver that kind of perfomance, check sd650

https://browser.primatelabs.com/geekbench3/7878018

That is a mid-ranged SOC manufactured in 28nm and clocked at 1'8GHz, so it is not a fair comparison.

 

[mikk]

I guess that it s the same kind of "TDP" as cherry trail s, 6W officialy but measured at more than 10W by hardware.fr who stated that the TDP is underestimated...

I

You are confusing the notebook or desktop version with the tablet version. Notebooks SKUs have had a TDP of 6W, but this is offtopic. You are confusing notebook with tablet.

 

[poofyhairguy]

Not even close.

Ug then.

 

[KTE]

Throughput shenanigans. You can't have 1.56x ST increase AND 1.56x MT increase, and also base 2000 too? Who uses that? It's 2006 for a decade now.

Intels Atom needs a huge overhaul for its naming scheme. Very convulted and confusing now.

Sent from HTC 10
(Opinions are own)

 

[Raduque]

Not even close.

You sure? The Atom in my Surface3 didn't feel that much slower (in CPU tasks) than the 6Y30 in my Spectre x2.

 

[Roland00Address]

You sure? The Atom in my Surface3 didn't feel that much slower (in CPU tasks) than the 6Y30 in my Spectre x2.

Human subjective perceptions is not the best way to objectively measure things.

 

[SAAA]

Core-M is faster because it's a dual core with HT vs quad core for Broxton, so single thread is much higher for core but some multithreaded loads might be better on the Atom.
If we go by geekbench alone there's still a 40-50% advantage per clock for core overall (some individual scores go to 100% or better) but other benches might show a bigger difference.

Point is that Atom isn't laughable anymore and it's sad to see it disappear from mobile, maybe it will return with Surface phone? I'm not sure if 10nm will bring Cannonlake power down enough to get into phones but we'll see, maybe at 2-3W TDP it's low enough to start thinking about it.

 

[Lepton87]

If the core M is 40-50% faster per core then the 2C/4T core-m should still feel noticeably faster in most use cases because for a CPU to subjectively "feel" fast ST performance is the most important. Most applications even if they make use of more than 2 threads are still bound by the performance of one heavy thread. However in some cases the quad core atom is going to outperform the 4 threaded core-m.

 

[Sweepr]

Apollo Lake prototype (ES) 35% faster than Braswell @ Cinebench 11.5 ST:

However, we tested the sample with a single-core benchmark, Cinebench R11.5 (0.54 vs 0.4 in favor of the N4200, of course), and compared it to the N3700. The Apollo Lake chip scored 35% better than its Braswell counterpart even though it has considerably lower clock speeds. This seems promising and we suspect that the final unit will deliver even higher performance, especially in the multi-core tests./quote]

http://laptopmedia.com/news/intels-...entium-n3700-by-35-in-single-core-performancel

I suspect Turbo is still a mess with the sample they tested, so scores might improve on final devices. Expect design wins at IFA in September:

www.digitimes.com/news/a20160824PD203.htm

 

[majord]

So still looking at Sub Mullins IPC in this bench at least.. maybe a match if it wasn't hitting full Turbo.

 

[sm625]

Even with a 50% increase in performance, you're still only looking at an Octane v2 score of 12000. A core m 6Y30 scores 50-70% more than this so you would need yet another 50% boost on top of this just to match the very lowest end Core m. Before that happens, Core m will receive a die shrink making it potentially lower power than this atom if they focus on power consumption and choose not to boost the Core m's performance.

 

[cbn]

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

A 57% higher passmark would put a N3700 Pentium into the ballpark of a Q6600:

http://www.cpubenchmark.net/cpu.php?cpu=Intel+Core2+Quad+Q6600+@+2.40GHz

http://www.cpubenchmark.net/cpu.php?cpu=Intel+Pentium+N3700+@+1.60GHz

Likewise, a 81% higher passmark would put a N3050 Celeron into the ballpark of a P8700 Core 2 mobile processor:

http://www.cpubenchmark.net/cpu.php?cpu=Intel+Core2+Duo+P8700+@+2.53GHz

http://www.cpubenchmark.net/cpu.php?cpu=Intel+Celeron+N3050+@+1.60GHz&id=2541

Will be interesting to see how other benchmarks turn out.

P.S. Will be also interesting how Apollo Lake's eMMC 5.0 (interface speed of 400 MB/s) compares to a SSD on a typical Core 2 era SATA 3 Gbps port with IDE (rather than AHCI).

 

[cbn]

Apollo Lake on Intel Ark:

http://ark.intel.com/products/codename/80644/Apollo-Lake#@All

 

[Sweepr]

Some new numbers from Intel marketing material:

In comparing the mobile Pentium N4200 versus the most powerful mobile Braswell SOC, the Pentium N3710, Intel is promoting that the N4200 is 30% faster in SYSmark 2014. Meanwhile they’re reporting that the same setup is 45% faster with 3DMark Skydiver.

On the CPU side, keeping in mind that the two chips have a maximum turbo clock within 60MHz of each other, the performance gains are significant. While it’s unclear if Intel keep memory bandwidth the same (Apollo Lake supports LPDDR4, Braswell does not), it either way hints that the Goldmont CPU core offers significantly higher IPC than the older Airmont CPU core. This is only one test – and from the manufacturer at that – but it will be interesting to see how well other benchmarks respond to Goldmont once we get our hands on Apollo Lake devices. At the same time, for performance to increase by 30% despite the much lower CPU base clock, this indicates that Intel is likely still coming out well ahead even if they had to pull back on frequencies a bit due to power consumption.

As for the iGPU performance, a 45% increase is a combination of factors. Just by the raw numbers, the Apollo Lake Pentium has a 20% GPU throughput advantage due to its higher GPU turbo frequency and overall wider GPU configuration. At the same time this accounts for less half of the reported performance increase Intel is reporting. So we are also seeing the overall performance and efficiency improvements of the Gen9 GPU architecture in action.

Couldn't find Pentium N3710 results, but a slower Pentium N3700 scores 439 (SYSmark 2014 Overall).
Pentium N4200 (Apollo Lake) might perform around ~600 in the same benchmark. In comparison, Core m3-6Y30 scores 720. Most interesting part is, any performance increase probable comes from architecture improvements, given that base clock is significantly lower than Braswell.

http://www.anandtech.com/show/10635/intel-quietly-launches-apollo-lake-soc

SYSmark 2014 Overall
- Pentium N3700 (Braswell): 439
- A6-5200 (Kabini): 440
- A8-7410 (Carrizo-L): 469
- Pentium N4200 (Apollo Lake): ~600
- Core m3-6Y30 (Skylake-Y): 720

 

[Sweepr]

Geekbench 4 results for Pentium N4200 (Apollo Lake):

Single-Core Score
1575
Multi-Core Score
4263

https://browser.primatelabs.com/v4/cpu/547636

Pentium N3710 (Braswell):

Single-Core Score
1158
Multi-Core Score
3223

https://browser.primatelabs.com/v4/cpu/500154


New bits about Goldmont!

 

[dark zero]

Core 2 Quad levels it seems and trade blows with Stoney Ridge on ST...

 

[Nothingness]

It's still not at the level of higher-end ARM chips such as Cortex-A72, but definitely much better than the previous poor Atom designs.

 

[hojnikb]

Does it have support for VP9 and H265 10bit ?
If it does, it would be a great replacement for my ol' z3735f

 

[majord]

3 issue - funny how Atom has gone full circle and has ended up looking more like an optimised Pentium M , albeit with 128b datapaths now, like Jaguar?