Intel Atom Z600

Cedarview is ww15-r3

68.4 fps in Egypt 720p and 91.7 in Pro 720p. Performs like Tegra 3. Looks like its the 640MHz version from its performance.

 

Given the timeframe, I think it's much more likely that this is a sighting of clovertrail-w or whatever intel is targeting for win tablets at the end of this year. We know the 2580 is using 544mp2, but we don't know what clovertrail-w is using. Assuming its powerVR of some sort, it could be a single core 544 @400-500mhz, which would make more sense of the fill rate.

The tested platform also begins with "CT" which might also lend a clue to it being clovertrail.

 

Good catch. Clovertrail-W with one 544 seems to fit the rumors and those benchmark results (which were hosted only briefly) better.

 

I see chipworks have done their thing on the Z2460.

http://www.chipworks.com/ko/technic.../inside-the-lava-xolo-–-intel-penwell-inside/

die size is approx 63.5mm2.

Apple 32nm A5 is around 69nm2 according to Anandtech, and thats with dual A9 and perhaps x3 the graphics performance of the intel chip, likely similar video encode/decode (supplied by IMG in both cases).

 

The size is in line with early Anand's estimates (62 sq.mm).
I'm not sure your comparison is relevant though, the A5 offer way more GPU power than what most manufacturers seems to deem as relevant.
I don't think that the number of cores is a metric of performances, Medfield beats the A5 dual cores in CPU benchs. They may be sucky benches but that's all we've got and double standards are a bad thing.
CPU power is what is relevant, either way AMD vs Intel fight would not turn out as it does.

Let wait for more Krait and A15 based designs. Let see how much they consume and perform and then speak.
Against those opponent the bigger brother will be a more proper opponent. I don't expect it to be outgunned significantly, a few percents are nothing, manufacturers are to look for price, production capacity, yields, let see if Intel is OK with lower margins than in its traditional markets).

 

If we know that the benchmarks are 100% single threaded (like Sunspider, which prints its CPU usage afterwards if you don't believe it) then it's correct to not use them to compare a single Medfield core with two Cortex-A9 cores.

Furthermore, we don't really know how the 32nm A5 CPU cores would perform while clocked liberally; it's a fair assumption that they can achieve both better performance and perf/W than its 45nm predecessor. Any perf advantage is lost when Apple clocks them the same, but the perf/W advantage is at least evident in an impressive reduction in power consumption for a device that has a lot of other major power consuming components.

It'd be interesting to see some die specs for the Exynos 4412. That's on the same Samsung 32nm HKMG process as Apple's new A5, which could be considered a fair competitor for Intel's 32nm HKMG SoC process Medfield is fabbed on. I'm curious is Exynos 4412 and 32nm A5 are using the same Cortex-A9 circuit layout; previous SoCs (A4 and S5PC110) both used the identical Intrinsity designed Cortex-A8. Since Apple acquired Intrinsity they would perhaps have to be licensing the design to Samsung's fabs. Chances are Apple/Intrinsity would still have a close relationship with Samsung's fabs and this sort of continued cross licensing would make sense, at least until Apple rolls out their own ARM cores.

If they really do end up beign the same core then it'd be somewhat fair game to compare Exynos's 1.4-1.8GHz Cortex-A8 performance with Medfield (although there will be models that can boost up to 2GHz, and some of this will depend on binning on both sides)

 

Yea we have seen the inflated javascript scores, but how does it fair in other tests?? well for starters its about on par to a Exynos 4210 @ 1.4ghz, gpu is obviously much worse on the Medfield.

It could be that Intel has used a crap GPU to pull down the power consumption numbers? if it had a comparable gpu o Tegra 3 or worse Exynos 4412 then that would change things i think.

One plus point i noticed with that Medfield reference platform (XOLO LAVA) is that although battery life is average, the battery in the thing is only 1450 ma/h, which is quite small now for a smartphone, on the other hand, that is the same size as the Iphone 4s, which is near the top of the charts, although screen size would obviously play a part in that, if you bolted Medfield into a HTC ONE X how would the batterylife compare then?? thats with only 1 core HT, and a poor GPU, double the resources of those 2 parameters to bring it in line with current Android flagships and battery life would be much worse i think its fair to say.

The fact you can't play many high end games no matter what the gpu is due to the poor emulation then maybe it isn't cut out for high end Android at the moment in anycase, unless specialised apps are wrote, which with zero market share i doubt very much.

Overall i suprised Intel has done as well as it has, although its got a way to go yet, Exynos 5250 & S4 PRO are hitting by the end of the year....

Oh yea can someone explain why the Medfield XOLO is so damn thick?? despite such a small battery? its even got a reasonably largish form factor to spread the components out? i know its olny a reference design but this chunkyness is something of a trademark for intel phones....

 

Just a thought....

With Google buy out of Motorola, Moto's deal with Intel for Medfield this year, and Google looking to use different manufacturers for Nexus designs and chipsets...could we be seeing a Higher end Medfield derivative used in a Moto Nexus phone in Q4?? thought id throw that one out there....

 

I was positively surprised as well. This thing isn't even running on Intel's highest end process technology (22 nm tri-gate), but is instead using a standard 32 nm process. And on top of that it's "emulating" part of the Android software (likely JIT compiled to x86), and Atom isn't the perfect architecture to run poorly optimized JIT code (simple in-order cores).

Lava International Ltd isn't Samsung or Apple. It took these big companies several generations to get their designs as thin and lightweight as we expect current phones to be. I think XOLO is pretty good considering it is their first phone, and they don't have countless of dollars to spend on chassis/material development as Samsung and Apple do. Intel surely would benefit if they had Samsung, HTC, Motorola or Apple as their partner instead of some new startup.

Intel was likely wanting to introduce their mobile Atom in a Nokia Meego phone (they co-developed the OS after all). But things didn't unfortunately go that way...

 

Intel getting their excuses in early regarding multicores in the mobile space;
http://www.theregister.co.uk/2012/06/07/intel_bell_mobile_launch/

Looks like they have been smoking the same shizz that Stephen Elop has, ''single cores rule!!'' er no no they don't, its just that you can't compete on that level more like it. Can you imagine a 4 thread Atom phone in the next 12 months??.. no neither can i.

 

I'm not sure about what so funny about it.
Phones still have a lot to gain wrt to single thread performances.
The point is more "can Intel deliver something say a A15 can't?"

 

Current atoms are dual threaded. Since they are in order, they can gain a lot of performance if the OS supports multithreading. Very different from what nokia's saying, (as I understood their situation).

 

The point im trying to make is neither Nokia or Intel can supply multicores in the mobile space ATM, which is needed on an ADVANCED media device, my Galaxy S3 proves that.

..So what they do is construct a media campaign to discredit the use for multicores in an Android phone, not because they are not needed, but because they can't/wont supply them.

Yes of course we need more single thread performance, but thats not what they are saying, they are saying multicores are a waste, there not.

Whats the difference between HT in order and HT OoO?? your not telling me an in order design is better than the former are you?
-ARM is already on 4 threads since end of last year, 2 cortex A9s are slightly faster than a higher clocked Atom with HT, by all accounts Medfield seems to struggle with heavy UI overskins, according to early reports on Orange Santa clara, (it could be early software) Qualcomm already has dual Kraits that crush Medfield in performance and power despite being lower clocked,and despite having LTE baseband and a more powerfull GPU.
(Intel is already on the more advanced mature 32 HKMG process, krait uses a early problematic non HKMG 28nm.)

Quad core Kraits, quad A15s and High clocked dual A15's are nearly apon us, Intel is no where near that performance metric.

If they had multicores and the competitor didn't they would be dissing single cores instead, in the future both Intel and Nokia will be using multicores and that has got nothing to with when consumers want or need them, its because they can't provide them right now.

-The funny part is Intel was originally going to be supplying Nokia with some single core goodness for Meego, Elop is the CEO of Nokia and is now peddling the same tune despite rejecting Intel, you would have thought they would be a perfect match.

 

The difference is that an in order micro architecture stalls more often than an OoO one, which means there's more opportunity to switch threads, resulting in a higher speed up. Of course that's a rough explanation but I hope you get the idea.

 

http://www.glbenchmark.com/phonedetails.jsp?benchmark=glpro21&D=Intel+ctp_pr1&testgroup=overall

The ctp_pr1 has re-appeared in the database; however judging from the normal results compared to the offscreen results, it smells like severe driver immaturity. Not usable even for an indication at this point.

 

That depends. So far I've still not really seen any benchmark which proves that those 4 cores in Tegra 3 provide any benefit in a smartphone (linpack doesn't count). I'm sure though it will be more useful in the future.

Quite sure all 28nm foundries use HKMG (pretty much required at that size). I don't know how it compares to 32nm intel process, it might be of similar quality.

Quad core Kraits are next year, dual A15 possibly too. I don't know how much earlier they will be here compared to dual core Atoms (Z2580) my guess is probably not too much.
The only quad-core A15 SoC I'm aware of is Tegra 4 (Wayne), and from what I know it won't be seen in smartphones, only tablets.

So really I don't think intel is that much behind. Dual-core Krait is probably the only SoC where I'd say it has some real advantage on the cpu side right now. Interestingly, dual-core A15 should have an advantage in single-thread performance against a Z2580 (but possibly not in multithreaded performance), but unless we've got some perf/power figures (on both sides) from these future cpus I'm not yet going to declare it a winner.

 

Yes thanks for that, i know what you mean now, however you sacrifice IPC with an in order micro-architecture, and besides HT isn't as efficient as 2 independant cores at multitasking, put it this way, 2 Cortex A9's at 1.66ghz properly optimised and on Intels 32nm HKMG process would destroy Saltwell in every metric including battery life and even die size wouldn't it? (?)

Well i can confirm with my own hands on and tests with system tuner pro that 4 threads in Galaxy S3 do get used very often, sometimes with all 4 cores running at 200mhz only to keep things smooth and efficient, sometimes with just 1 core with the rest power gated, all 4 cores @ 1.4ghz in 'race to sleep' mode and every variable inbetween.
...I can't log on this site with my smartphone for some reason, else i would upload some screen shots...trust me it works and makes the usually stuttery Android buttery smooth.

Well im pretty sure that early manufacturing over at TSMC is not on HKMG, and they certainly have had some serious problems with that process as Snapdragons have been delayed, Intels 32nm is very mature now and their HKMG technology is several generations along so i can't really see a scenario where it wouldn't be better, with the exception on die size.

I think, although i cant be sure that Snapdragon S4 pro will arrive in a couple of months, and that it will use Global founderies 28nm LP HKMG process, the S4 Pro is a quad core Krait and has already been benchmarked on an LG Eclipse phone, benchmarks can be found on one of these forum links.

Other quad cores that might be out this year, is Exynos 5450- quad A15. Tegra 4 is scheduled for Q4? STE is going to be bringing a very high clocked dual A15 @2.5ghz, Omap 5 series and Samsung also have high clocked dual A15's out year end ish, not to mention Exynos 4412, Tegra 3, Huawei in house quad out NOW...there is LOADS of competition already here with many more on the horizon.

'Wayne' is slated for smartphones as it says on their roadmap that it is using another 4+1 setup, basically an upgraded Tegra 3 with the same format..again sans baseband.

Intel has suprised many with the Medfield introduction, but it isn't competitive with high end ARM SOC's, the only real benchmark that makes Medfield shine is Sunspider benchies, - something people have said favours x86 (im not sure), is single threaded, not forgetting Medfield is clocked higher. Even then we are talking a very slight advantage, every other benchmark Medfield gets trashed.

Medfield is comparable to Exynos 4210 @ 1.4ghz (Galaxy note) with inferior graphics.

 

Yes, I think so.

 

I don't see how OoO execution sacrifice IPC.
Anyway A9 on Intel won't happen, it's irrelevant plenty of RISC architecture died to Intel process.
You can't take Intel CPU on insulation.
Something Intel has is really FPU / SIMD for example.
It's nice how irrelevant benchmark are relevant depending on who win.
Medfield is tiny barely bigger than a tegra2 on 32nm process.

This whole Intel sucks, X86 sucks, bores me I leave the discussion it's pointless.

 

I made a mistake in my statement, i had corrected it but i think you read the copy and paste in Laurent06's post.

I meant in order execution sacrifices IPC...
Besides when talking about a low power SOC thats going into smartphone..aka Medfield you have to compare it against it competitors...that im afraid is ARM designs,if you don't like the comparisons then don't discuss smartphone SOC's.

In isolation Medfield is a very decent chip, but its comparable to last year designs on 45nm with the notable exception of Sunspider.

The point i was trying to make was that Intel is already on the most advanced process node with Medfield, it can't compete with multicore A9's in any metric besides Sunspider, it is already getting its excuses in regarding multicores thats all im saying..