Intel Broadwell Thread

[witeken]

Doing a simple calculation, I came to the conclusion that Iris Pro (including eDRAM) shrunk to 14nm would be only 156mm², which is less than Ivy Bridge GT2. I'm excited for Broadwell Iris Pro with on-die L4 cache.

 

[Homeles]

Interesting, source?

Well, I'd first cite common sense. when have you seen a GPU with 128 (or whatever) discrete "cores?" Never.

There's a thread on Beyond3D that talks a bit more about the subject. I'll link it once I find it.

 

[witeken]

Well, I'd first cite common sense. when have you seen a GPU with 128 (or whatever) discrete "cores?" Never.

There's a thread on Beyond3D that talks a bit more about the subject. I'll link it once I find it.

Sure, Nvidia made the die look a bit nicer, but that doesn't automatically mean the proportions (ignoring the .little core) are wrong.

In any case, my point was that in mobile, it isn't uncommon to find much larger proportions between the CPU and GPU die areas. I guess I could also have taken a Cortex A7.

 

[Phynaz]

Pay no attention to Nvidia's Tegra die shots. They are complete fabrications.

But they are pretty.

 

[sxu]

Few benchmarks here: http://www.tabtech.de/windows/intel-core-m-5y10a-broadwell-unseren-benchmark-tests

Edit:
http://browser.primatelabs.com/geekbench3/753141
http://browser.primatelabs.com/geekbench3/753148

 

[liahos1]

From the first article

Peacekeeper 3322 (chrome)
Peacekeeper >2000 (IE)
Sunspider 194ms
Geekbench 3@~1ghz 2064 (sc) 4249 (mc)

vs IPAD Air

Peacekeeper 1902 (safari assuming)
Sunspider 384ms
Geekbench 3@1.4ghz 1358(sc) 2455 (mc)

seems really good esp since they are using the 5Y10 and not the faster y70. i wonder how dramatic of a step-up apple can get going to 20nm.

 

[witeken]

Lol, I remember people bragging Cyclone had a higher IPC than Core, but Broadwell, it turns out, destroys Cyclone, being 2X faster per clock (multithreaded is even worse due to MT).


"Desktop class"

Yes another myth busted.

__
Edit: according to a preview, 2 Broadwell-Y cores can turbo to 1.3GHz.

 

[liahos1]

Lol, I remember people bragging Cyclone had a higher IPC than Core, but Broadwell, it turns out, destroys Cyclone, being 2X faster per clock (multithreaded is even worse due to MT).


"Desktop class"

Yes another myth busted.

can't wait to see gpu performance. I just hope intel is realistic about the prices they sell these chips at.

Also need Cherrytrail to get some sort of superlative. if its not going to be a performance leader (as Core M seem to be placed there) would be good to be a battery life leader / cost leader.

Intel really should try to get Core M into android tablets too just so they can have leadership positions across the performance spectrum.

 

[Arachnotronic]

Lol, I remember people bragging Cyclone had a higher IPC than Core, but Broadwell, it turns out, destroys Cyclone, being 2X faster per clock (multithreaded is even worse due to MT).


"Desktop class"

Yes another myth busted.

__
Edit: according to a preview, 2 Broadwell-Y cores can turbo to 1.3GHz.

Broadwell 5Y10 turbos to 2GHz, AFAIK.

 

[witeken]

Broadwell 5Y10 turbos to 2GHz, AFAIK.

That's with 1 core.

 

[NTMBK]

That's with 1 core.

How many cores do you think a couple of browser benchmarks really use...?

 

[Enigmoid]

That's with 1 core.

Still it is safe to say that its not running at base clocks.

 

[witeken]

How many cores do you think a couple of browser benchmarks really use...?

What point to do you try to make?

 

[liahos1]

Intel Core i3-4020Y (SP3) on geekbench does
1585 3249

Intel Core i5-4300U (SP3) on geekbench does
2642 5277

vs 5y10 2064 (sc) 4249 (mc)

so sort of between an i3 and an i5 with lower specced core M. seems really good considering tdp cut from 11.5-15w to 4.5w.

would really like uncontested benchmark leadership from intel this year. not another baytrail type situation where perf/watt was very good, cpu was really good and gpu was ok. people instantly latched onto the gpu performance and disregarded what was an overall solid chip.

 

[witeken]

Still it is safe to say that its not running at base clocks.

Yes, Geekbench runs at 1GHz.

 

[NTMBK]

What point to do you try to make?

That browser benchmarks rely heavily on single core performance, so that 2GHz turbo will probably come into play.

 

[witeken]

Intel Core i3-4020Y (SP3) on geekbench does
1585 3249

so sort of between an i3 and an i5 with lower specced core M. seems really good considering tdp cut from 11.5-15w to 4.5w.

So performance/TDP went up by a factor 3.5.

would really like uncontested benchmark leadership from intel this year. not another baytrail type situation where perf/watt was very good, cpu was really good and gpu was ok. people instantly latched onto the gpu performance and disregarded what was an overall solid chip.

With 24 Gen8 EUs, GPU performance should be the best you can find in a tablet .

 

[North01]

That's with 1 core.

Max dual core turbo for the 5Y10/5Y10a is 2.00 GHz.

 

[TreVader]

Lol, I remember people bragging Cyclone had a higher IPC than Core, but Broadwell, it turns out, destroys Cyclone, being 2X faster per clock (multithreaded is even worse due to MT).


"Desktop class"

Yes another myth busted.

__
Edit: according to a preview, 2 Broadwell-Y cores can turbo to 1.3GHz.

First of all, the supposed "myth" you are not even close to busting was the FACT that haswell had about the same IPC as Cyclone.

Second, It is and will always be a FACT

Third, you are using intel marketing slides and their word on clock speeds and benching cyclone from 2013 against broadwell, a mid to late 2015 delivery product. That's not a good comparison, for many obvious reasons.


Furthermore, you ignore turbo frequencies and you rarely post anything that isn't an intel marketing slides. These are the facts.


Bust whatever myths you find after you stop taking intel marketing slides and doing mystery-math to get your weird numbers. The word that Cyclone IPC was as good as Haswell (or Ivy?) came from Anand himself. It's in his review, or do I need to link you? My guess is you have read it because I and many others have linked it to you specifically before when you went on these rants.

 

[liahos1]

i think y10 turbos to 2ghz and y70 turbos to 2.6 and ipad is 1.4 if remember correctly.

 

[Sweepr]

Thats SAD and basically these benches say Core M is iPad class. Which is not what I want in a macbook. Not even kidding I was hoping (and expecting) intel to trash the iPad with 2 die shrinks and vs a 2013 product but i guess not.

Max turbo for that model is 2GHz. There is a specific Broadwell SKU for the Macbook Air/Ultrabook PCs, Broadwell-U. Based on these benches the fastest 4.5W Core M SKU (2.6GHz Turbo) could offer similar performance to a current 15W Haswell-U Core i5, which is very impressive. I don't think Apple (A8) or NVIDIA (Denver) will come anywhere close this year but time will tell. Apple could very well release a 12'' Retina Macbook Air based on a Core M, it would be even thinner or accomodate a larger battery @ same size and still offer solid performance (24 EUS Gen 8 iGPU should help push tose extra pixels).

 

[TreVader]

Max turbo for that model is 2GHz. There is a specific Broadwell SKU for the Macbook Air/Ultrabook PCs, Broadwell-U. Based on these benches the fastest 4.5W Core M SKU (2.6GHz Turbo) could offer similar performance to a current 15W Haswell-U Core i5, which is very impressive. I don't think Apple (A8) or NVIDIA (Denver) will come anywhere close this year but time will tell. Apple could very well release a 12'' Retina Macbook Air based on a Core M, it would be even thinner or accomodate a larger battery @ same size and still offer solid performance (24 EUS Gen 8 iGPU should help push tose extra pixels).

Yeah i edited it.



It appears, if the benches are true, it could be a decent product. However, it's plainly misleading to say you are benching a 1Ghz core product against a 1.4 A7 when the core then runs at 2.6 for benches. Or 2.0, or really anything but 1Ghz like the A7 would, if the multiplier were set to 10x



Edit: I doubt it will beat K1 or Denver and probably not A8 in GPU. Intel is strong in single core and multicore CPU but actually very weak in GPU generally but especially in mobile. I think Core is better than their mobile junk tho.

 

[TreVader]

Taking the numbers from the article

SC Geekbench 3
1Ghz Broadwell Core M ~2000
1.4Ghz Apple A7 ~1400


Which looks impressive until you consider the broadwell sku is actually running at 2Ghz (assuming it's not thermally limited otherwise, and I doubt intel would gimp it).

So really it's

2Ghz Broadwell 2000
1.4Ghz A7 1400



So... according to the numbers, Broadwell Single Core IPC is about equal to A7.


If I fudged the numbers somehow, please explain. You can't say I'm cherry picking because I'm picking my cherry off intel's very own marketing slide.




edit: Multicore and GPU broadwell should outclass A7 but I don't know how anybody can expect Core M to beat Apple A8 when A7 was literally 2x as fast as A6 across the board (and 4x faster in GPU).

 

[jdubs03]

Third, you are using intel marketing slides and their word on clock speeds and benching cyclone from 2013 against broadwell, a mid to late 2015 delivery product. That's not a good comparison, for many obvious reasons.

It's a late 2014 product with a new stepping in early 2015, idk where you got mid-2015 from.

The GPU should be quite a performer, I expect it to beat both Denver and the A8 by a decent amount.

 

[TreVader]

It's a late 2014 product with a new stepping in early 2015, idk where you got mid-2015 from.

The GPU should be quite a performer, I expect it to beat both Denver and the A8 by a decent amount.

No, it's been pushed back to early 2015 for ultrabooks and even further for all the rest.



The only people getting broadwell in 2014 will be Apple Consumers. I guess it counts as 2014 but I'd say it is more a 2015 product vs cyclone which sold a hundred million in 2013.





edit: I suspect the multicore performance and chiefly power consumption are pushing Apple to use Broadwell in the Macbook Air. In these two areas intel has a clear lead over Apple and all the rest.