Thoughts?
http://seekingalpha.com/news/3168017-bloomberg-intel-talks-amd-license-gpu-patents
"Bloomberg reports Intel (NASDAQ:INTC) is in talks with AMD to license its smaller rival's GPU patents.
The report comes with Intel's cross-licensing deal with Nvidia (NASDAQ:NVDA) due to expire in Q1 2017. Nvidia has been recording $66M/quarter in licensing revenue related to the Intel deal, which allows Intel to use Nvidia's IP in the GPUs integrated with its CPUs.
AMD is up 6.1% after hours to $2.79. Nvidia is down 0.7% to $32.87."
>>53767260
Interesting. Nothing else to say about it until either company confirms it
>>53767260
What does it mean?
>>53767260
Intel finally giving up and admitting it's integrated GPUs are dogshit. Plus, AMD is more useful to them alive than dead.
more cores + hyperthreading?
>AMD needs cash money, now
>AMD sells them their small GPU patents
>Intel produces better IGPs and fuck over AMD in the near future
Whatever AMD does, AMD loses.
>>53767260
So what does AMD get out of this, a little bit of Intel's petty cash? AMD and Nvidia should both refuse and force Intel to drop their iGPUs.
for intel this is a cash injection to amd as they don't want to be a monopoly when it comes to x86 cpus.
>>53767360
doesnt work like that. Intel now uses nvidias patents. It wont make much difference to use amds
sell nvidia stock now sell sell sell
>>53767360
This.
Once dealt, it's easy for Intel to exceed AMD in any possible advantage, given the Intel capitals.
>>53767345
Intel is already becoming like AMD.
>>53767511
so, what's the REAL power consumption on this 22/44 CPU at FULL LOAD?
>>53767542
not that much higher because the base clock is low as fuck on the core 22/44 and the other 4/8 core is 5,1ghz doofus
145wtdp for 22cores is efficent as fuck
>>53767604
I don't believe in it, not before i can see the tests with full system watt-meter with idle and full load states.
>>53767260
The iGPU IP lease could really benefit AMD. That $66M/quarter would certainly offset much of AMD's operating costs. Making their financial situation much easier to deal with.
Hell, AMD's iGPUs being far superior to nVidia's patents in that area might mean they can get more than the $66M out of Intel.
>>53767675
They should not crumble.
This is just wrong.
>>53767511
This is not designed for your GAYMAN RIG, retard.
>>53767712
Oh hardly.
All major manufacturers lease IP to eachother.
ie;
AMD leases Intel AMD64
Intel leases AMD x86
AMD and nVidia no doubt lease some obscure low level architectural shit to eachother.
The only thing that comes from AMD saying "no" is Intel go back to nvidia. No progress gets made on iGPU tech, and AMD is denied valuable income. There is no possible ethical or financial argument that could possibly be made against AMD leasing Intel some IP.
Hell, even with that IP, Intel would be no threat, they have no GPU talent, hence why they have failed to develop their own IP to suit their needs.
Looks like I'm gonna let my amd stock ride
Good for AMD and I guess Intel. I doubt AMD is going to sell their GCN arch, they're probably just licensing parts of it out like they did with Matrox. If they sell anything it will probably older stuff.
Just an article about Intel considering this is enough to bump AMD's stock 6%????
Fuck I wish I had a bunch of funds laying around to invest in AMD.
With Zen on the horizon, now this, there is only one way for AMD stock and its up. Unless of course Zen turns out to be a catastrophicly horrid shit pile cpu of the century, but not likely.
>>53767260
>Intel CPU with AMD GPU
AMD APU btfo? Or does Intel got an insight (or leak) about Zen?
>>53767283
Nothing, this is not for iGPU or APU this is just for some patents that both Nvidia and AMD hold in the manufacturing of GPUs and Intel needs if doesn't wants to be sued.
>>53767260
>intel enters the GPU market
>makes 11nm GPUs
>Nvidia is now the cheap shit
>AMD confirmed dead
Just watch it happen
>>53768086
>sell
GCN is trash literally nobody will buy it.
Intel is licensing their tech meaning they probably already have a better design and don't want to infringe on existing patents.
>>53770441
Intel entered the GPU market a long time ago anon.
https://jonpeddie.com/publications/market_watch
>>53770441
Or worse, Intel is planning to get into the HSA thing so they need some stuff from AMD. Leaving nVidia to fend for themselves and getting effectively squeezed out.
>>53770506
discrete GPU not shitty iGPU
>>53770489
You're a complete retard, AMD and Nvidia are the highest end GPU designers there's nothing else even remotely close to what they're capable of. Any business that is mildly interested in designing GPUs would be interested in acquiring some or all of their assets.
>>53770585
The majority of the money in the GPU business is at the low end. Intel has already cut Nvidia off from selling motherboards with Nvidia integrated graphics, if they were to start releasing low to mid end GPUs that were halfway decent they could severely damage Nvidia in the same way that they damaged AMD when they got a lead in the CPU market.
The future of the PC industry could easily just be Intel vs their one blessed competitor AMD who is kept afloat simply to keep regulators at bay.
>>53767360
it's not selling.
it's licensing. Just like x86 (intel->amd) and x64 (amd_>intel). Plus they can negotiate for an expiration date, renegotiation of terms, etc.
>>53767360
get royalties on each cpu sold by intel.
intel cpu's sell like hot cakes.
I don't see how you think that's a bad scenario. Especially when this revenue would be going to your another rival in any case.
nVidia is kill
RIP in piece, never 5get
>>53770441
this
intel is going to steamroll all over the market.
AMD has no choice but to stand and watch
>>53771250
Intel stockholders pls go
>>53767260
AMD will almost definitely get more than $66M/quarter if they're going to license a fair portion of their GPU IP.
A nice little uplift in their earnings but nothing to write home about.
>>53771292
Go where? Their 200 Million dollar mansions?
>>53770506
just because something has an iGPU doesnt mean it's getting leveraged.
>>53767283
someday soon intel is gonna have no drivers.
intel could easily scale up their igpu tech into expansion cards, i don't think they'd enter the gaymur market or want to cannabalize their own HPC offering (xeon phi)'s market share though.
>>53767604
>>53767542
>>53767511
TDP =/= power consumption
Intel has been forced to pay Nvidia a licensing fee for years despite not technically using any of their IP. There was an issue with intel's integrated graphics maybe infringing on a vague patent held by Nvidia.
Intel wants out of the arrangement, and licensing some IP from AMD would solve the issue of needing a basis to continually advance their IGP without stepping on any toes from a legal standpoint. That doesn't mean intel CPUs will start using a GCN based IGP, what intel would end up licensing is far more vague than any specific architecture.
This is just a move to protect themselves from future litigation and nothing more.
>>53773690
everyone in this thread seems to think they're going to make a GPU when in reality it's just what >>53767447 said. nothing will change except who gets intel's money.
makes perfect sense. intel iGPU's are shit and AMD CPU's are shit
the only reason they are doing this is because once zen is release and AMD can build APU's that arent choked by bulldozer intel's integrated market will disapear
the deal means massive revenue for AMD so if you havent bought AMD stock start buying RIGHT NOW cause its about to go up
>>53770070
>Unless of course Zen turns out to be a catastrophicly horrid shit pile cpu of the century, but not likely.
i truly believed in bulldozer, i want to believe in zen but...
>>53774417
have faith
>>53767360
Intel already makes the fastest iGPUs with broadwell, they're just expensive as fuck due to how they achieve that speed
>>53774612
its hard anon
>>53774417
Bulldozer was pretty great for multi-threaded workloads though
>>53774704
how do they differ from the other ones?
>>53774725
if you were on a budget, yes. but intel high-end still beat it by quite a margin.
>>53774759
bulldover was winning in certain workloads over ivy bridge or sandy bridge though
as for broadwell, the chips have a 128MB cache specially for the GPU. it's like how the PS4 uses GDDR5 for it's APU, making it way faster.
>>53774797
it was the crushing blow anon
we waited for bulldozer for ages and it was a turd
Patents aren't GPU tech, they're usually software, but can be firmware, microcode related power management and the like.
It's not selling out the GPU like IT does with PowerVR
Baseless speculation
regardless, it could be a smart business move for intel, they need someone to play 2nd place
but likewise they can probably get a better deal from amd than nvidia.
I know it's not nearly related but imagine with their tech combined
full high end cpu and gpu in one chip?
would make water cooling a lot easier
Dam, I knew intel HD graphics sucked but this bad? Can't intel just throw more executions on the die and call it a day? Do they really have no choice but to suck AMD's cock and use their iGPUs?
>>53777870
*execution units
>>53777870
That's what they've been doing since literally forever. Turns out it just makes the chip extra hot and causes more agressive throttling. This is why broadwell flopped. Good news is if they can strike out a deal AMD then putting 128/192 GCN 1.2 (1.3?) cores on die would improve temperatures and sustained iGPU + CPU frequencies.
I don't see it happening though, AMD is BTFO intel on the laptop/desktop market since people want to be able to play vydias in the sub $600 range.
>>53775324
AMD ain't giving intel jack shit, they are fucking dominating the laptop market with the newer carizzo APUs. Giving intel a leg up in integrated graphics performance would be pure stupidity.
>>53778184
>they are fucking dominating the laptop market with the newer carizzo APUs
In what country is that even remotely true? Carrizo is having power delivery issues on some laptops, and the performance is still overall crap for the CPU given the lack of per core performance AMD chips have
>>53778226
>In what country is that even remotely true? Carrizo is having power delivery issues on some laptops, and the performance is still overall crap for the CPU given the lack of per core performance AMD chips have
per core performance means jack shit now. Things like decoding VP9 streams, browsing the web with chrome, multi-tasking, playing DX12/mantle API based vydias rely on multi-core performance.
The power delivery issue exist for those using their laptop on battery. Of course it's gonna throttle if it's gonna risk the laptop battery becoming overheated. This is pretty much true for intel laptops as well.
>>53778338
no like carrizo chips have a 35 Watt TDP, but some laptops are only set to delivery for a 15W TDP
>>53778593
Not really the laptop's fault, those specific carrizo APUs were designed around a 15w tdp specification from the start. It's not that the laptop can't give the chip enough juice but when running both the cpu and gpu at once (eg vydias) the entire APU starts to heat up and over time throttle to prevent itself from reaching Tjunction.
Anyway because of the low specified TDP, laptop manufacturers cheap out on cooling and only aim to cool the chip at the specified 15W of TDP, nothing more. Carrizo was designed with energy efficiency in mind not raw performance thus only carrizo with a TDP of 35W will see constant ~3GHz CPU frequencies and ~800 Mhz iGPU frequencies.
>>53778593
Thats no a problem with power delivery or anything of the sort. That is the manufacturer setting the cTDP so they can put it in a cheaper package.
35w Carrizo can be set to 15w or up to something like 42w for max turbo.
>>53778853
That's not how this works, it all has to do with the cooling. If the laptop has a good cooling solution then keeping ~3ghz on the cpu and 800mhz on the igpu shouldn't be a problem even if the chip was rated for 15w tdp.
How do I know this? Because I've left my A10-7300 lanovo laptop in a room with the AC set to 60°F for a day encoding video while I went to the mall. I also left open source hardware monitor writing down temps and cpu frequencies. In the end the laptop stayed at 3.2GHz most of the time and temperatures under 150°F.
tl;dr laptop cooling sucks ass on most laptops
>>53779061
>I don't understand things: the post
Manufacturers use whatever thermal solution is necessary for the corresponding cTDP.
AMD APUs throttle even under normal circumstances because of preprogrammed thermal margin and GeAPM.
>>53767260
Nvidia has burnt lots of bridges and once the contract is up, companies flock to AMD because of how horrible it was to work with Nvidia. Intel isn't the first nor the last. It's a ongoing trend for companies to go to AMD after the contract is over with Nvidia.
>>53767283
Intel's iGPU implementation probably uses some techniques that are covered by Nvidia's patents.
Intel has previously licensed them from Nvidia.
Nvidia being Nvidia, they're probably planning up the patent licensing costs when the current agreement expires and sue if Intel doesn't pay up.
AMD probably has similar patents on techniques similar to ones used by Nvidia, so Intel could license them to prevent legal assault by Nvidia for more IP licensing bucks.
I don't know, I'm kind of at two minds with this. I have a A10-8700P and the thermal throttling issues kicks in depending on the driver installed. The original catalyst driver had the GPU throttle kick in at 65c and kept it steady at 59c. While the latest Crimson drivers throttle at 60c and keep it steady at 54c.
It feels to me that the problem is somewhat spread out between OEM and AMD. Perhaps better co-operation would resolve the issue.
>>53767511
>5.1 jiggahurtz
Wat
>>53768086
I would laugh so hard if they sold terrascale
>>53767511
>>53780180
I have doubts that the 5.1ghz thing is at all real
>>53767360
>ARM Buys Nvidia
>we finally get rid of those horrible amd based mobile based phone gpus
>???
Do it ARM plz
>>53780223
Why?
ARM already has Mali, which can actually run without burning a hole in the pocket unlike GeForce in Tegra.
>>53767260
>polish their turds until they're almost competitive with AMD's using Nvidia's IP
>just throw it all and buy AMD IP
Looks like they want to go into le trash too
>>53774704
The only reason why it's faster than AMD's APU's it's due to the massively expensive SRAM that they ship with those, pretty much every APU would beat it if it wasn't bandwidth starved
>>53767360
Won't happen, Intel couldn't really get their iGPU's to speed after almost two decades of R&D, even while buying IP from GPU designers
>>53773769
Power consumption is always going to be higher than TDP at full load
>>53780831
Right we need Zen APUs with HBM2 on die
>>53780856
HBM APU's are apparently planned, but won't come until 2018, maybe 2019
>>53780889
Q2 2017 most likely
>>53780944
Raven Ridge still isn't confirmed to have HBM on package.
We hope it does, but nothing is concrete.
>>53781057
you want the laptop market it needs it desperately.
>>53767260
>tfw the last shred of hope for a CPU with an integrated MIC coprocessor just died
Good night, sweet prince.
>>53773769
Then why do these faggots advertise it as so?
nvidiafags and intel
>>53783169
Why do you even need this?
Just buy a xeon phi extension card, they are cheap.
>>53770070
amd has been in a consistent slide for 5 years, they have had "saivors" before it's never been enough it would be a shit investment monetarily. buy as much of their shit as you want but you will lose your ass betting amd in the market.
>>53770070
Yes. That's how the stock market works. People who trade stocks will be trading AMD one day and dow chemicals the next.
If in the news there's a big stable company investing in a smaller one, you buy the shares of the smaller one because their share price will go up.
Yes, because other people are doing the same thing.
>>53780223
>>53780383
Sorry to say that but PowerVR produces the most efficient GPUs there are on the market.
>>53774889
>using windows to benchmark
What a joke you should compile the benchmark in native march. Otherwise you have no idea if that excutable has a instruction set bias.
>>53774725
so you were able to compile a kernel with it faster? I mean, if you use Gentoo, maybe it's useful?
>>53775348
So which high-end GPU and CPU should they combine?
And let's say they pick a 6700K and a GTX980.
Now they need to double the size, need a new socket, a new motherboard.
I mean, just for a few people, this entire shit.. it's just stupid
>>53778338
>per core performance means jack shit now
>>53778593
TDP means how much heat it generates at full load, it is not power consumption.
>>53787544
Do you not know the first law of thermodynamics?
>>53787544
the exact meaning of what "total" dissipated power is up for debate, but you're retarded if you think that power consumed and heat dissipated aren't identical.
do you think your CPU is telekinetically lifting rocks somewhere out of sight with excess energy?
>>53787520
He's right though, outside of poorly optimized vydias, per core performance isn't that important. Most programs today can efficiently use 4 threads so having an i3 or A10 APU gives you pretty much the same performance (except on hardware accelerated ones in which thr APU performs better).
>>53787544
even if it's not 100% accurate it's still a general representation of power draw
Why do companies make deals with novidya and shilltel? They are known to fuck over everyone they've worked with and it's common knowledge that they are deceitful and dishonest.
Both companies are extremely unethical and seek to dominate the market by sheer sabotage or gouging, because they can afford it.
I truly wish the next gen AMD processors will be at least an alternative to Intel's shit.
>>53788731
>they're known to fuck over everyone they've worked with
Cool story bro
>why do companies work with each other?
Companies can buy and sell things just like people can you fucking moron. They're not asking to work with AMD they're literally saying here's a bunch of cash if you'll license me some patents.
>deceitful and dishonest
>AMD is always honest and would never lie to anyone
>bulldozer
kek
![5[1].jpg](https://i.imgur.com/bEnVkja.jpg "5[1].jpg")