Delay in graphic cards

Is there a delay in Processor to Graphic card technology release and why?

I had the impression that Graphic cards dont have the same micro architecture as the Processor they publish previously.
Intel & AMD appear to use 1-2 year old micro-architecture in their "new" graphic cards.

AMD released the new Radeon rx 5XX graphic cards but no one talks about Zen or a giant performance leap, similar to the Ryzen hype.

When can we expect a graphic card with Zen micro architecture?

Other urls found in this thread:

viatech.com/en/silicon/processors/
twitter.com/NSFWRedditVideo

Fabrication

BTW

GPU is a parallel processor

CPU is a serial processor

They are entirely different animals

???
Graphics cards use completely different processors than CPUs. A "Zen" GPU will never exist, because Zen is a CPU architecture.

Since you seem to have asked this question in genuine ignorance, here's why:
>CPUs need to do a wide variety of things including managing hardware and allowing security features
>In contrast, GPUs generally only need to crunch numbers
>Since GPUs only handle simple operations, they need much less silicon per "core" (think a CUDA core, not "more cores lmao" ayymd)
>This allows GPUs to pack a couple thousand simplified "cores" onto a single card, but the cores are only used for things requiring heavy number crunching (like calculating colors for pixels)

Basically, CPUs and GPUs are both processors, but fulfill entirely different roles and are optimized for completely different uses. It wouldn't make any sense to try to build a Zen GPU.

But they give them the same name?
Intel and Nvidia use the same Product name (Pascal,..) So it would appear that they are of a similar origin/design.
But while we have Skylake in Processors they still use Pascal in graphic cards

additional to AMD introduced the 14nm structure, why not use the same manufacturing precision in the GPU, or would they need to completely reinvent their GPU processors?

I don't even know about Oascal and Intel.

Pascal Nvidia is the next iteration after Maxwell, Kepler before, Fermi before that. Some generations are complete architectural changes (like G80 and GF200 to Fermi/Kepler) whereas others are smaller like Evergreen to Bart's (Ati).

>Intel and Nvidia use the same Product name (Pascal,..)
Gotta admit, I'm not going to say you're wrong, but I've never heard of an intel "pascal" product at all. And a quick google search isn't turning up anything. Care to link to what intel product you're referring to? As far as I know, anybody talking about a "pascal" chip these days would only ever be talking about NVIDIA GPUs. Skylake is most definitely not the same thing as pascal--they're not even close to the same chips (for the reasons in ).

I'm pretty sure that was TSMC. AMD and everyone get their contracts from Samsung, GloFo, etc to produce their designs.

Which leads to your answer. I was saying Fabrication, and that is what holds us back. Fabrication has gotten eerily close to the atomic level where we'll just break everything. It will keep going, for a little while. I think Qualcomm is already producing 10nm with Samsung or TSMC. I forget.

Thing is, new fabrications generally lead to smaller die packages, which is severely amounting to thermal consequences.

Hence why all the newest fabrications begin with mobile SoC. Nobody will ever consider a 250W GPU if we can't produce 5W SoC.

>intel product (skylake, pascal,...)
thats how they label the architecture of their processors and GPU. Skylake was introduced with the 6xxx row of their processors, before that they used pascal architecture.

impressive

But before Skylake, intel used the Broadwell architecture, not Pascal?

Indeed

What's really impressive is how well ARM has optimized. Performance of mobile SoC is at Sandy Bridge levels.

What the tech world is learning is that low power multi node is the only way to move forward, with or without ARM.

seeing that 5nm is in the future, is it reasonable to assume they will build pc cpu's with that architecture size (costs?) or will they only implement them in mobile devices?

literally most of the "its a new architecture" is just a meme marketing talk. Newest nvidia cards architecture is exactly the same as 5y old. Ofcourse its lower nm number, improved thermals and some optimizations means its much more efficiant and can reach higher clocks. Same thing with AMD.

Intel threatened to stop selling socketed CPU below their HEDT line. I don't know, computing is destined to change further in the next ten years. I think the age of high watt, high clock CPU is dying soon.

ofcourse they will, but first will come low power mobile solutions. also MS trying to implement ARM cpu's on windows, so we might actually see full on beefy ARM cpu's competing with AMD and INTEL.

First they're gonna pull all K series off mainstream. Then they'll start selling motherboards with BGA i3's and i5's instead of just celerons like they do now.

they wont lol, amd would roast there ass so hard

I think they are trying to push all enthusiasts onto their HEDT platform. Kaby Lake X is the start of that.

It's already in the works

All that solder driving Intel bonkers

>Intel and Nvidia use the same Product name (Pascal
Nigga, what???

just doesnt make sense
you can boy b350 for 80$, cheapest x299 is like 250$.
Then 1600 is like 200$, while intel option is 400$.
650$ vs 280$.
Sure intel cpu might be faster somewhat, but games will always primarily be limited by GPU!.

One thing thats bothering me is x86 license, its like 50y old, shouldnt patents expire at some point in time? Intel is slowing down progression.

I never said it was a good idea.

Isnt AMD actually making ARM based CPU? Cant they just fucking buy Snapdragon 835 and produce it themselfs they fucking have the fucking license? They could literally make windows 10 on ARM a reality and Intel couldnt do shit!

Microsoft and Qualcomm are already partnering to do that. Intel even got peeved at their x86 emulation.

ye, but with AMD they wouldnt need to emulate it, AMD could just go ham and add/modify/make compatable x86 directly to the ARM cpu.
Obviously for MS it would be cheaper if everyone could make cpu's for there devices, but if x86 emulation fails. AYYMD doing native x86 ARM cpu would be logical

What do you need 32-bit for. Is the trend not going beyond 4gb ram?

>They could literally make windows 10 on ARM a reality
That already exists. The problem is that its shit because it has no software.

are you retarded? x86 includes x86_64

Qualcomm owns their design but ARM will liscense their Cortex design. Thing is, that is why Qualcomm exists in the first place, to build custom SoC.

yes, thats why they want to emulate x86, so all native shit works. They could go 1 leg up and just do native x86 arm cpu, unless AMD license specifically prohibites them doing arm x86 compatable cpu.

If they use ARM for anything, it's going to be server shit. And seeing as they have Epyc, there doesn't seem to be much point.

There is the caveat that both Intel and Sony experimented with specialized, high-core-count x86 and Power-derived "GPU"s running tile-based software rendering.

It seems to have been a dead end, though, Larrabee was cancelled and the planned dual-socket 2x Cell PS3 was apparently so dire that adding $100 to the cost and pushing the release back a year to integrate an already-obsolete Nvidia chip was preferable.

Base x86, and extensions through the 486 series, are out of patent. There are two problems:
a) No one wants or can really use upclocked, reverse-engineered 486es. You could probably compile Linux on them, but you can also compile it on ARM and the ARM licensing fees are less than what it costs to physically design a processor from specifications.
b) x86 is out of patent protection. MMX left last year. SSE and EMMX are still in. x86-64 is still in for nearly a decade. The only reason to use the old and busted x86 architecture is for compatibility, but as long as extensions keep getting added, those extensions will be patented and software compiled for use with them will not be compatible with your chip.
There is a business case for "well, most software probably won't use built-in AES instructions or large vectors", and perhaps in the future this will be less of an issue, but right now the sliding 20-year window is still firmly in the era when they were bolting extra general-purpose registers on and even gcc hello.c may decide to put something in one.

See ttp://opencores.org/project,zet86 , it's not that (((Intel))) and AyyyyyyMD are shutting down base x86 implementations, it's that even the FPGA guys think it's less effort to just roll down to the dump and find someone's old Packard Bell to desolder.

There is an obscure x86 russian processor, i dont remember the name, was launched some years ago. Performance was shit, i dont know if they have patents to use it, but since performance is so shitty no one cares, it was in 486 - pentium levels.

Also, some younger ppl will not remember, but there was other vendors for x86, like VIA.

It seen VIA still is in the market,
viatech.com/en/silicon/processors/

From time to time i see some vias running machine boxes like 3d printing and industrial appliances.

actually its the reverse, x86_64 contain x86 instruction set. Also, x86_64 is AMD.