Ryzen 3

I'd have fallen for this if not for the Windows flag. 8/8

I'm not really defending mudslimes here, but Islam and the Arabs actually helped mathematics, astronomy and a few other divisions of science a lot.

I swear AMD is gonna BTFO Jewtel this year.

What happened to VIA and S3 Graphics anyway? Dead?

yes
no

They stole that from the indians, no muslim did anything more than rape and steal

VIA bought S3 and then they disappeared from the mainstream market because their offerings were insultingly bad on both the CPU and GPU side of the business. Weak chink CPUs, inferior graphics chips, and their drivers were disgusting in isolation but multiplied by being the entire chipset.
Throw them in a shallow grave and shit on it.

>yfw no ThinkPad T23 with that glorious 16MB S3 SuperSavage IX so I can run Deus Ex with 30+ FPS

Nope.

MILL ARCHITECTURE

I'm still skeptical on how a FIFO approach is better than moderm parallel designs. In addition it sounds like the relative-address nature of the architecture will make systems programming a nightmare

You mean ethnic greeks, right?

bump

The IBM PC Standard is the ONLY computing platform standard we currently have, yet it's dominated by just two companies. We should have countless companies making x86_64 chips but instead we mainly have 2 while ARM has no platform standard yet we see hundreds of ARM chip makers. Could this purely be the result of Intels aggressive marketing strategies driving everyone else out?

It's weird how every AMD processor is the "Intel killer" until it releases and we get benchmarks then those posters just vanish until the next AMD processor. I wonder if they're even real people.

The cost of entry in the high-end processor market is in the 100 billion dollar range. There won't be competition until some breakthrough that commoditizes chip fab.

Yeah, there are much better algorithms for thread queues and there isn't a whole lot of reason not to be using them. Imo processor development has grown stale. Instead of trying to employ and research new techniques the only thing they care about is 'muh nanometers'.

Faggot that is from liscenses/permission stamps you need to enter the field.
Actually building your own one layered pcb proccessor is entirely possible from a DIY lab costing around 1,200 USD and the knowledge. I don't know about multi layer fabs though. Some other user mind chiming in on the proccess involved for a multilayer board's creation en mass? Or atleast the idea behind its creation?

Look at the delusions on this faggot.

...

...

Not talking about PCBs here, but rather CPUs like was talking about. I too, have scoured the internet, and as far as I can tell the smallest feature size you can achieve as a hobbyist at home is about the micrometer level. So taking design into account, if you're really really persistent you could probably make something equivalent in speed to a 386 from scratch. Which is inferior to an FPGA, which is massively inferior to the current high end processor market. The facilities to make the high end stuff costs billions.

It is an interesting idea to completely separate fabbing from chip design, however. Most of the massive cost is in the former. It could play out much like places where internet infrastructure is government owned and ISPs all rent use of it at a consistent price, resulting in lots of competition.

youtube.com/watch?v=67WhV0EDqCA

Whats this then?

Faggot webm it.

Haha, no. I love triggering freetards by linking YT.

No, that's just the "we wuz kangz" version of islam (with every invention literally stolen)

They're not though, AMD is winning with raw core count and they are close enough in single thread, as far as the high end goes. Epyc caps out at 32 cores and IIRC Xeon is only 24 cores. Not to mention Ryzen is a bit more efficient, specially given Intel's flat out lying about TDP and power consumption.

x86 has more or less reached peak single thread performance, all that's left is micro-optimizations and adding instructions to speed things up. Which either means 5% performance per clock increases or waiting for developers and compilers to add support for new instructions.

Ryzen basically brought a doubling of cores at Intel's same price points, while single thread is still pretty good.

You should care, because the only way x86 is going to get faster is by adding more cores and more instructions and waiting for software to catch up.

It is a webm.
248 webm 1920x1080 1080p 2627k , vp9, 30fps, video only, 100.85MiB

The $13000 chips have 28 cores, they're arranged in a 6x5 grid with one area reserved for GPU and one for backdoors.

Post the relavent parts or downscale the video quality faggot. You have a brain to use youtube-dl you can have a brain to use ffmpeg and downscale.

intel.com/content/www/us/en/products/processors/xeon-phi/xeon-phi-processors/7230f.html

64 cores

No, I didn't type that wrong, it's not 64 threads, 64 PHYSICAL CORES. I honestly don't know where the fuck people here get their information and then spout conjecture on shit they know nothing about

Where's yours?

From lists of actual CPUs, not the specs of PCIe cards.

Please elaborate because that spec sheet even lists the number of PCIe lanes the processor itself supports

wtf? I love diversity now!

Copyright protection of media should be reduced to 20 years or how long a patent lasts. No way a movie should have more protection than the cure for cancer will ever get.

Any movie,show,or song that's currently 20 years old or older would become public domain.

You misspelled AMDs long awaited death so Intel can finally get fucked by Antitrusts.

Microsoft is working on making ARM x86 compatible with software, and it's rustling Intel's jimmies.

We could have backdoor proof "x86" CPUs thanks to Microsoft.

extremetech.com/computing/249292-microsoft-declares-windows-10-arm-devices-will-run-x86-code-near-native-speed

Linux Devs will implement this somehow down the road so we won't need windows 10.

It's running x86 through a hypervisor that routes requests to equivalent ARM instructions and vice-versa with a minimal piece of emulation software

This is not a new concept either and we've seen similar projects in the past with both hardware and software implementation

I am skeptical we will see anything of this materialize and if we do it will most likely be limited to Microsofts own implementation in Windows. If it was practical we would have a practical solution by now. We even have a limited WINE implementation for Android that uses QEMU libraries to emulate hardware-specific bits. It's not that great

...

That's how it used to be, google mickey mouse rule. It's a lost cause now, just ignore it and stay safe.

The Phi isn't a general purpose CPU, more like a CPU/GPU mix.

You're a retard, OP. IBM POWER competes with high end Xeons directly and does it very well.

...

I want AMD fanboys to leave

The most powerful PPC chips come nowhere close to Xeons. PPC barely compete anymore, see the image here;

That's what Charles Holland Duell said.

Keyword is general purpose. The Zion Phi is a manycore arch.

Xeon Phi is an accelerator card that runs it's own OS. It isn't completely x86 compatible either, it's missing a lot of features.

The fuck are you smoking? IIRC Xeon Phi cores are literally Intel Atom cores. It is absolutely x86 this board man smh

Seems that AMD is kicking Intel ass so Hasbarats are temporarily diverted from fighting ebil nazis to shill for the chosen processor.

The shilling for it is real. Like don't get me wrong ryzen > intel for high end. Other than that AMD's proccessors are shat on by intel. I have seen the amd shills on other sites too. Even some of their engineers help shill too, it's sad and pathetic.

Earlier versions were based off P5 Pentium. It's a die shrunk P5 with AVX and other vector instructions added on, all on a ring bus.

Knight's Landing, the newer one, is based on a highly modified Silvermont core. They basically removed a lot of features/instructions to reduce die size, added vector instructions and SMT, and then put it all on a ring bus. It's not capable of running generic x86 binaries without recompiling.

Knight's Landing, the newer Phi, only supports x86, x87, SSE, AVX1, AVX2, and EMU. It's missing a lot.

Xeon Phi came out of the failed Larrabee project, where Intel tried for a long time to create a GPU that could compete with Nvidia and ATI/AMD. Eventually they just gave up and created Xeon Phi as an accelerator card for vector math.

It is based on an x86 CPU, but it's not actually an x86 CPU since it can't function as a central processing unit for a computer, it needs a host system to run on. So it doesn't really count as a CPU.

We have a word for that, it's called a co-processor.
Either way, solid information, thanks man.

Is this the one that has a GPU integrated or is that still coming. I already have a r5 1600 for linux but the idea of dropping a chip and a board into a small box and using it as a toy to play around with old games on win7 is vaguely appealing.

go outside and pick something up off the side of the road, it's how I've gotten like half my computers for the last 17 years

Here's what's going to happen:

>Microsoft gets tired of their shit and blindsides them by making devices with both CPUs inside, just like consoles with backward compatibility have been doing successfully for 25 years

Do the shills of this value brand do it for free?

AMD just anounced a "secure processor" built into their GPU's. It prevents tampering/using different unbackdoored firmware like Nvidia's recent GPU's. Every single last one of them is kiked. They all need to gas themselves. Linky archive.fo/yo2rt
Nvidia is kiked post maxwell gpu's. AMD is kiked post vbios mandatory in kernel firmware gpu's. Is there anything better than pre maxwell nvidia GPU's for non kiked GPU's?

Sure is diverse. Unless youre talking about video and audio chips and not just cpu

8-but had the older Motorola chips, Intel, MOS, the Z80, and so on, you're just cherrypicking

Might as well consider an R9 Fury or RX Vega and it's 4096 streaming processors since we're bringing up computing cards that aren't strictly x86_64.

That would still miss the point since AMDshit has jack shit market share in the massive parallel computing market

They better be considering the low effort put into it

Even AMD shills are underpowered.

Even ignoring numerous PC, workstation, and supercomputer ISAs that have today been snuffed out root and branch, the level of monoculture in embedded hardware today is unprecedented:
landley.net/history/mirror/tech/processors/cpu.html

It's true they preserved a great deal of discoveries from both the west and far east, in addition to collating them to synthesize new insights, but they also did a lot of entirely new work themselves. Really, Europe's post-Roman descent into backward fanatical infighting barbarism, colonized by an enlightened and advanced cosmopolitan civilization, was in many ways a perfect inversion of today.

Scattered among countless dreams. 3Dfx, Matrox, Intergraph, 3DLabs, Rendition, SiS. About the only thing that's truly independent of AMD/nVidia is PowerVR, which hasn't made a desktop card in ages. Funny thing is, nearly all of their original staff are SGI diaspora.

If SGI had seen the writing on the wall for how badly consumers wanted 3D hardware in PC platforms at the crucial moment (~1992-1995) they easily could've had a monopoly.

Sounds similar to the old PPC/68k approach old Macs used, where references to OS toolbox code were rerouted to native PPC equivalents, while app-specific 68k code was emulated.

What is/was the method Windows NT/CE used for cross-architecture stuff back when they supported more platforms?

This is missing "dark ages" somewhere

That's marketing bullshit. It has 64 "compute units" (think "Bulldozer module") with four hardware threads ("cores") each that share an instruction decoder and scalar unit, but each have their own 16-wide vector unit. It has either 64 or 256 cores, depending on whether you believe the threads in a Bulldozer module are "cores".

If we were to go with the marketing bullshit definition, we would have to call a modern desktop Core i5 a 32-core processor, since each real core has 8 vector lanes wide AVX units. Wait, stop, it has two AVX units working in parallel per core, so it's a 64-core processor!

GPU driver discussion is offtopic to this thread (therefore sage), but to answer your question:
AMDGPU is just the kernel part of the open driver. If you want the fully Free stack, you run Mesa (with RadeonSI driver) on top of AMDGPU,
AMDGPU-PRO is basically their old and creaky Catalyst driver rebased to run on top of AMDGPU instead of their own proprietary kernel module. It's slower and quirkier than modern Mesa, but it works just as the old Catalyst, with all it's quirks that Serious Proprietary Software relies on. It also supports some shit that the open stack doesn't (yet) - OpenCL, OpenGL compatibility profiles and Vulkan.

So TL;DR:
If you want just gaymes and swishy wobbly windows, you're best off with the fully open stack (AMDGPU+Mesa/RadeonSI). If you need OpenCL and/or Serious Proprietary Software, you go with AMDGPU-PRO.
Vulkan is actively worked on in the open stack, so it will be the best choice for Vulkan gaymes too in the near future.
If you have an old card (pre-290), the open stack is the only supported option.

Wait, you forgot hyperthreading, now it's a 128 core processor ;^)

Not really, SMT doesn't increase the number of execution resources, it just lets them be shared by multiple execution threads. GCN also does up to 10 threads by each core (though they're not executed simultaenously, rather, they're only switched when the current thread hits a high-latency event)
Also, i5s don't have "hyperthreading" you dunce.

Sony's Emotion Engine™ is a 2560-bit CPU.

...

The EE was used for the PS2, did you not mean the Cell Broadband Engine?

Sony's Emotion Engine™ is a 2560-bit CPU.

So Threadripper vs Core i9 benchmarks are finally out, and it looks to be strong at multi-threaded workloads despite just barely beating out the i9, despite the i9 having less cores. Single-threaded performance the i9 beats out Threadripper, but just barely. in other words exactly what we all expected. Personally I am actually genuinely excited about Threadripper and AMDs comeback overall. It feels like the early 2000s again when AMD was actually competitive with Intel, makes me nostalgic.

Don't ATIAMD cards still need OS-loaded firmware blobs in all cases? They're still useless in that case.

Black Edition with unlockable CCXes when

I wonder how fast it will be once they patch all the microcode bugs.

Processors now need firmware blobs, too. If you're running a server without the AMD/Intel microcode packages installed and set to update you're asking for it.

Yes this is why nouveau/nvidia is shilled so hard against. If you use nouveau you don't need a single firmware blob. If you could somehow get a libreboot computer with a nvidia pre maxwell gpu you are golden for libre software. But this is impossible with AMD since the "(pre-290)" days. AMD shills the word "open" so hard because like kikes if you perpetuate a lie enough than it becomes true to idiots. You can't use nvidia hardware post maxwell though so it is a defunct point.

Or you just don't use post sandybridge CPU's or hell even the x86 arcitecture.

Is it really microcode optimization affecting performance? I genuinely hope AMD sorts that shit out and makes Zen the microarchitecture it was born to be. But not having that shit sorted out at launch sends a very bad message, hopefully the damage is minimal.

Microcode fixes are often for bugs and security issues and reduce performance.

Bigger problem is that these are NUMA-on-a-chip, and like multicore, amd64 and SMT before this, consumer/gamer Windows shits itself hard and lags years behind when faced with the steady march of technology.

I thought that most of the functionality (such as the 3D acceleration) still requires firmware blobs.

Only if you use post-maxwell chips. Otherwise blobs are only wanted(not needed) for better video decoding. Which you can always use EGL/open source for or software/mesa/your proccessor for.

amdgpu pro is needed for 1440p144hz on my 290

Hah, how things changed in just a week. I got a T23 in near new condition!

Now hope DX runs well with double the VRAM compared to the T22 and a better CPU architecture,

Nothing short of RISC-V can get me exited anymore.

Maybe POWER would?

That's a total crock of horseshit, quit talking out your ass you fucking fagot. Firstly you can get any LGA 3647 motherboard and slap a phi in that shit and boot to your choice of x86 OS. Secondly even the 'accelerator cards' are just little computers running linux off a flashable rom, you even get full root on them how the fuck is that not a cpu for a computer. You need to get fucking gassed.

That's as bullshit as saying POWER9 is $300 for 3 times the performance of a $1700 i9, you just need to buy a $2500 motherboard to put it in.

...

The mobo is only $2150. Stop exaggerating.

I see you talking but this is all I hear.

20 seconds on google gives me this

ebay.com/itm/322603950288

And some sweet engineering samples for $400 each

ebay.com/itm/132296643849

wew