AMD doesn't actually believe that they can compete with

>>59534381
snapdragon 835 is already DOA, they announced it like months ago and still no phones

>>59534381
>mobile means phones
>couldn't possible mean Windows tablets, laptops, AIO etc

Fucking retard delete this thread.

They already have an ARM core in the works, and they have more expertise than Qualcomm.

They don't have to. ARM is still significantly behind in terms of IPC and energy efficiency compared to x86 zen.

>>59534403
no, no they do not.

when it comes to mobile processors
samsung>qualcomm>apple>mediatek>anything amd

>>59534381
probably. they'll most likely use cherry picked geekbench scores or something to prove they don't make shit processors.

>>59534426
>ARM is still significantly behind in terms of energy efficiency compared to x86 zen
[citation needed]

OP doesn't even know that qualcomm's processors are (were) AMD.

Kinda weird that are going back to mobile, since i'm pretty sure they sold the IP patents - instead of simply licensing.

>>59534479
Yeah, but that's part of the whole marketing game. Everyone shows their own in the most favorable comparison. Intel, Qualcomm, and even TI, also do it. That's why you have to read third party reviews and benchmarks, and if you can, get sample builds to test your workload (rare outside of companies, but awesome when you can do it).

>>59534426
It would a legal quagmire for them to go back to ARM, even if they developed new stuff.

>>59534390
wasn't this the one that samsung was acused of hoarding?

>>59534565
It's only OK when Intel does it.

>>59534560
my point still stands. amd can't hope to compete with Apple's a10 or Qualcomm's 835. stay delusional.

>>59534643
Stop pretending you know shit, faggot.

>>59534658
t. amd shill

>>59534560
Only the GPUs were AMD (technically ATI). The adreno ("radeon" but mixed up) GPUs were based on ATI's "Imageon" GPU cores.

>>59534665
t. delusional Samshill

>>59534643
I have no doubt that AMD could produce an extremely competitive SoC if they actually cared about the smartphone market. I think they're more interested in markets that aren't already crowded. x86 in general, servers, scientific/HPC seem to be their currently targets. K12 is still apparently in the works but it's not as clear when it will come

>>59534693
t. amd shill

>>59534601
If they are hoarding so much, where is the S8 then?

>>59534714
t. stubborn crApple shill

>>59534381
guys, I think they mean laptops, not phones

>>59534724
t. braindead amd shill

>>59534734
t. cumguzzling Crapcomm shill

File: Comb22032017110645.jpg (157KB, 720x1267px) Image search: [Google]

157KB, 720x1267px

>>59534503
Zen is now on par with delidlake and ARM hasn't gained any significant market share in servers like at all.

Pic related is a 2 physical core delidlake i3 vs 4 physical core snapoven. Both capable of 2.4 GHz.

Look at the multi-FP score, the snapoven barely beats the i3 by 15% despite having 2X the cores.

ignore the single-core FP benchmark, geekbench treats threads as physical cores but it's the best benchmark I could get my hands on.

>>59534728
They do but they also claim to have K12 still in the works which won't be going into regular laptops for sure. It could escape the microserver market and end up in phones for all we know

>>59534667
nope, Qualcomm took it all, everything AMD had on its ARM division -- for 65mil..
>The agreement means that Qualcomm will not only acquire AMD's handset division, but the company also inherits the graphics, multimedia technology and other intellectual property from the business. With the deal complete, Qualcomm plans to use that technology and intellectual property to develop new types of 2D (two-dimensional) and 3D (three-dimensional) graphics technologies for handset devices and cell phones as well as enhanced audio, video and display capabilities.

I mean, it's what you said and then some more.

>>59534756
>reading comprehension

>>59534829
shhh, let him live out his delusions.

>>59534829
Oh shit sorry, disregard me - I'm a dumbass.

>>59534426
>>59534503
Close, but not quite.

1CCX 4core Ryzen at 1.9Ghz with a 4 CU GPU at 500-650Mhz or so when you factor in a lower uncore as well would likely still be over 10w TDP. Probably about 12w TDP, more precisely.
HBM could offset that TDP a bit, as well as improve performance, though.

You're talking about the power to run most games games from 2015 and earlier at 30+fps at 10w, but that's still too much for a phone.

Snapdragon 835 is 5w TDP. I'm going to assume its IPC of it bigLittle 4+4 2.35+1.6 puts it at similar performance of a 4 core Ryzen at 1.9 (that's surely being EXTREMELY generous), and its Adreno 530 at 635Mhz is probably about half the power of 4 Vega CUs at the 500-650Mhz range. (I'm assuming the 4x perf per watt of Vega over Polaris isn't true and that it's really generally in the 2-2.5x range. Adreno 530 perf/watt is actually right on par with Polaris if not better.)

Though considering that Zen is 14nm and the Snapdragon 835 is 10nm, and Zen is x86-64, it should not be right within striking distance of ARM, that's for sure.

Also when you consider some of the fucked up "optimizations" in ARM that means you can't relie on two different ARM CPUs getting the same result when they add two floats and other shit, tons of developers would welcome it being replaced with x86-64.

>>59534932
>snapdragon 835 is 5w tdp
*3.5

>>59534932
Oh I may have miscalculated here, adding in the TDP of the graphics memory for AMD when the Snapdragon TDP was not including the shared memory.

So it really may be more like 7w TDP for a 1.9ghz 4 core Ryzen APU with 4 CU Vega iGPU at 500-650mhz, as long as they saved a little bit on the uncore in adapting it for phones.

That would give it higher perf/watt than ARM when it comes to smart phones, but that's comparing older tech Adreno to the newer higher perf/watt Vega and not just the CPU side alone.
But I think that's fair since they're SoCs and it's up to AMD whether they will license their new graphics tech or not to Qualcomm.

>>59534966
No there are 2.5-5w TDP SKUs of the 835. There's not just one 835. I was comparing to the highest clocked one which is 5w TDP.

Ryzen Mobile refers to laptops you fucking idiots. Ryzen won't scale well down to <5W TDP range just as ARM won't scale well up to 95W.

>>59534581
stop spouting shit you don't understand.

File: LFaWnyI.png (207KB, 1006x354px) Image search: [Google]

207KB, 1006x354px

>>59535097
Raven Ridge mobile SKUs will range from 4w to 35w.

>>59535188
4 watt ryzen will be insane...

>>59534581
I don't think there's actually anything stopping AMD except the lack of motivation. If Nvidia can get into tablets and shit then AMD should be able to as well. Hell Intel tried to get into the phone market. I think AMD is deliberately avoiding markets that have a lot of competition already.

>>59535188
It's still strange seeing bulldozer become a low power core. It completely replaced bobcat

>>59535188
4W will be cTDP low in ultramobile/tablet space; you will see similar numbers for Core M and other such x86 products that do not directly compete with ARM.

>>59534458
um... you do know that qualcomm is using amd ip in their processors, and are the reason amd doesn't have a mobile chip right? Also considering amd's current cpu, and what they did with fucking bulldozer, they have more experience making cpus use less power than anyone else.

How long was the non compete for?

>>59534390
Xperia xz/xz premium using it, no?

>>59535234
Intel tried to go into mobile with x86, also they tried to make an x86 gpu, amd has no 'we must use x86' ties so they can use arm.

the issue is back with fagot ceo they fucked amd by selling all their mobile ip at the time and signing a non compete.

>>59534695
If they aren't trying to make something for the mobile market then lisa should kill herself, she failed as a ceo.

You got 2 markets

1) enterprise
2) consumer

making an enterprise solution will make a good enough consumer pc and laptop cpu so long as you mcm the enterprise, zen proved this.

however a massive amount of people went to mobile and never looked back, not to mention mobile is a market that has seen fairly steady growth and will only increase.

If you are not making a cpu for mobile, you are not a cpu company.
Oh and before intel gets brought up, intel has not made a new uarch in 20~ years that has been successful, they are stuck on the fucking p6 that was extended to what you have today.

Basically speaking, all the talent that made intel what is is fucking gone.
Meanwhile jim put together a team that put amd back on the map, and possibly made the second coming of jesus in gpu form is coming from that team too

Then you have via making a cpu worth a damn apparently... just face facts, intel is a shit company propped up by one arch they can't improve on further, the last time they tried, they made the pentium 4, and they are trying again...

>>59534774
This was actually the turning point for Qualcomm vs hisilicon/Mediatek/Marvell/other now minor ARM Manufacturers

>>59535340
At this point if there even is a non-compete clause I don't believe it would be enforceable. These kinds of contracts do not tend to last more than a few years. I'm actually not even sure if there is one or what the conditions for it even are but I can't imagine it would still be in effect 8-10 years later. At the very least the IP they gave Qualcomm will become public domain soon(like within 10 years) so they'll have access to it again anyway if they still care about it.

File: 8Rch6JF.png (55KB, 1509x905px) Image search: [Google]

55KB, 1509x905px

>>59534932

Where are you getting your numbers from for power consumption? You're just pulling them out of thin air as far as I am concerned and the numbers seem more like an estimate for Zen 2. If you look at the frequency vmin graph as attached, and look at some power consumption numbers where the 1800X consumes at peak around 110-120W and do some math to dervive power consumption at a given vmin, and cut in half to extrapolate to 1 CCX at your supposed 1.9 GHz, you still end up at around 14 W TDP not including GPU. GPU right now with TPU's estimates https://www.techpowerup.com/gpudb/2494/radeon-r3-graphics where even a low clocked R2 is rated with 15 W and maybe Polaris can lower consumption by 33% to 10 W at a most optimistic estimate, you're looking at a ~25 TDP or so unless clocking down more on Zen 1 with Ryzen, and that matches with the figures AMD has in >>59535188 which makes sense.

You would have to bin extremely well to scale to mobile or use a completely different die and tear up some of the performance wins that take too much power in the design to scale down to mobile , which I figure might be the strategy. They are only going to probably expand to laptops and tablets for now.

Keep in mind that Core M, Intel's most scaled down Core architecture parts that isn't Atom, clocks at around 800-1.2 GHz and consumes around 10 W TDP (Their stated TDP is 4.5 W which is kinda bullshit).

They absolutely need another generation to work things out. The GlobalFoundries 14 nm LPP may not be optimal yet so squeezing out more from Ryzen may be possible as yields get better. But that still won't yield the power consumption numbers you want at that frequency.

>>59534756
not to mention all the instruction sets intel has on chips vs the none arm does

>>59534458
You do realize that AMD sold its SoC division to Qualcomm years ago right?
Qualcomm got there because AMD's bad decissions but now AMD is getting back on their feet.