[Rumour] AMD's Zen To Have Ten Pipelines Per Core

[Notional]

No offence but im calling it out :

AMD will be a goner by the time they make the ZEN CPU.

but on the other hand , the gpu part of AMD is getting shaky lately... ether the only thing that will keep AMD going gg or game over is the GPU market OR the GPUs will triple the rate of AMDs downfall.

 

I presume you have no real education nor insight into the business world? At least not with such a statement.

 

1. ZEN is already finished and taped out. Now we just wait for manufacturing to be done. Will take some time to fine tune that and start producing stock for launch.
2. The financial situation of AMD is extremely exaggerated. Even if AMD were to go belly up, it would take 4-10 years still. Too much can happen in IT in such a span of time, it's impossible to say either way.
3. Even IF AMD ran out of money, they would file a chapter 11 bankruptcy, which means they get safe haven from creditors, so they can restructure the company, and continue their business.

AMD and Nvidia has been going back and forth with KOTH and so on. But the FUD, propaganda and sheepiness of NVidia fanboys are over 9000. Like an admin in here once said, Whatever NVidia is best at is the most important thing ever. And he hit the nail on the head with that one.

AMD is the better choice on all pricebrackets, with the only exception of the 980ti and Titan X. Not buying those cards? Well get more for you money with AMD. And it is finally starting to show, as AMD are regaining market share. With DX12 and async compute (thanks to the consoles), we should see this paradigm shift continue.

 

As for their CPU business, yeah it's been bad. But even if ZEN cannot compete 100% with Kaby Lake or Sky Lake, it can still be very competitive, and with things like HSA and crossfire with ZEN on any GPU (at least in DX12), we might see AMD becoming the go to gaming CPU next year too.

 

 

"The reports of my death have been greatly exaggerated." - Mark Twain AMD

[Ridska]

Whow , the responses I got by making one post is amazing . . ok y'all are right , maybe I did jump the gun on AMD .

And BTW about those long responses I got : never knew AMD was a rocky ( the boxing boxing movie ) story .

[FratStar]

-snip-

 

My God how does @LukaP stand you people? At least when Opcode and I argued it was over business practices and projections, and the only reason he left was because of a blood feud with one of the mods. 

 

I'll answer that question for you...Ignore them. If someone can't see the difference between this and CMT they haven't the slightest clue about how hardware is designed on the engineering level.

[MageTank]

Intel stopped playing dirty when Paul Otellini stepped down. Kirzanich really is a saint by comparison.

 

No, just no... Larrabee died because Nvidia got scared and bought itself out of its full graphics licensing agreement with Intel, scuttling its ability to release Larrabee as a graphics card, but now we have the Xeon Phi and Nvidia's eating crow.

 

The problem for Nvidia and trying to keep developers in their pockets is Intel has

1) Far more developers on its platforms

2) Support of open standards that actually work well (OpenMP, OpenACC)

3) At least double the programmers to aid developers in transitioning to its own platform

 

Intel only has to win the following to beat Nvidia in the long run

1) HPC

2) Professionals

3) High end enthusiasts

 

Take those three and the rest falls in line on its own, not to mention Nvidia's finances would be a smoking pile of rubble at that point. There's very little need for Intel to be the king of marketing in consumer GPUs.

 

I still say Microsoft has no real interest and that Samsung would be legally barred from purchasing AMD due to national security concerns in the U.S., but that's more speculative than based solely in the facts.

http://www.vrworld.com/2009/10/12/an-inconvenient-truth-intel-larrabee-story-revealed/

 

 

The strained relationship between the two got into a state of war when Intel started talking to OEMs and claiming that nVidia does not have the right to create chipsets for Nehalem [QPI – Quick Path Interface] and Lynnfield [DMI – Digital Multimedia Interface]. Upon request, we were shown a cross-license deal between Intel and nVidia. I am not going to disclose which side showed it to me, since technically – the source did something it wasn’t supposed to do.

The wording in the original document, as far as my laic understanding, does not bar nVidia from making chipsets for Intel even after Front Side Bus is dead, because both QPI and DMI qualify as a "processor interconnect", regardless of what either party is saying.

Intel filed a suit against nVidia in Delaware court [naturally, since both companies are incorporated in the "Venture Capital of the World" state], claiming that nVidia doesn’t hold the license for CPUs that have integrated memory controller. nVidia didn’t stand back, but pulled a counter-suit, but this time around, nVidia wanted the cross-license deal annulled and to stop Intel from shipping products that use nVidia patents.

If you wonder why this cross-license agreement is of key importance for Larrabee, the reason is simple: without nVidia patents, there is no Larrabee. There are no integrated chipsets either, since they would infringe nVidia’s patents as well. Yes, you’ve read that correctly. The Larrabee architecture uses some patents from both ATI and nVidia, just like every graphics chip in the industry. You cannot invent a chip without infringing on patents set by other companies, thus everything is handled in a civil matter – with agreements. We heard a figure of around several dozen patents, touching Larrabee from the way how frame buffer is created to the "deep dive" called memory controller. If you end up in court, that means you pulled a very wrong move, or the pursuing company is out to get you. If a judge would side with nVidia, Larrabee could not come to market and well can you say – Houston, we have a problem?

http://www.pcworld.com/article/183810/Larrabee_Project_a_Casualty_of_Intels_Legal_Battles.html

 

 

 

The FTC, European Commission, and the state of New York all have ongoing antitrust cases against Intel. The FTC recently discussed Intel's predatory practices with Nvidia, and is exploring whether Intel's lawsuit against its former ally is merely an attempt to stifle competition disguised as a contract dispute.

 

 

ntel and Nvidia originally entered into a strategic alliance in 2004, agreeing to share patents and work together. The purchase of ATI by AMD was perhaps a response to the Intel-Nvidia partnership, pairing the CPU and GPU (graphics processing unit) underdogs to do battle with the CPU and GPU leaders.

The honeymoon between Intel and Nvidia is over, though. Intel filed a lawsuit against Nvidia claiming that the 2004 agreement does not allow Nvidia to develop or manufacture chipsets--the chips that provide the brains of the motherboard and facilitate communication between the CPU and the rest of the system.

Nvidia countersued claiming that it does, in fact, have the right to develop chipsets as a function of the 2004 licensing agreement. With the collapse of the arrangement between the two, Intel loses access to Nvidia's considerable intellectual property related to graphics processing.

That loss has a direct impact on development of the Larrabee Project. Without Nvidia, Larrabee is dead in the water. At the very least, Intel has to go back to the drawing board and start research and development from the ground up since it can't stand on Nvidia's shoulders.

 

Multiple sources point to Larrabee's demise stemming from Intel getting greedy with OEM's, claiming Nvidia lacked the right to QPI-DMI chips, when they were legally allowed to do so. Instead of NVidia taking the high ground like Adults, they launched a counter suit of their own, and things got ugly from there. 

 

If you can point me to a source (ha...i am funny) of your version of the Larrabee affair, i'd appreciate it. After all, the internet is still up in arms over what went down. Some even point to angry engineers, and ineffective rasterization techniques (http://www.drdobbs.com/parallel/rasterization-on-larrabee/217200602)

 

Though, i'll stick to my original claims of Intel getting greedy with OEMS, and trying to revoke rights of another company without any reason to do so. After all, there are more sources to go with that story than any other story regarding Larrabee, so until i see evidence stating otherwise, what else am i to believe?

[patrickjp93]

http://www.vrworld.com/2009/10/12/an-inconvenient-truth-intel-larrabee-story-revealed/

 

http://www.pcworld.com/article/183810/Larrabee_Project_a_Casualty_of_Intels_Legal_Battles.html

 

 

 

Multiple sources point to Larrabee's demise stemming from Intel getting greedy with OEM's, claiming Nvidia lacked the right to QPI-DMI chips, when they were legally allowed to do so. Instead of NVidia taking the high ground like Adults, they launched a counter suit of their own, and things got ugly from there. 

 

If you can point me to a source (ha...i am funny) of your version of the Larrabee affair, i'd appreciate it. After all, the internet is still up in arms over what went down. Some even point to angry engineers, and ineffective rasterization techniques (http://www.drdobbs.com/parallel/rasterization-on-larrabee/217200602)

 

Though, i'll stick to my original claims of Intel getting greedy with OEMS, and trying to revoke rights of another company without any reason to do so. After all, there are more sources to go with that story than any other story regarding Larrabee, so until i see evidence stating otherwise, what else am i to believe?

The sources used in this article are garbage and not backed up at all. Why are people still quoting this? It's well written, but the substance isn't there, 0 linkage back to any original documents.

[MageTank]

The sources used in this article are garbage and not backed up at all. Why are people still quoting this? It's well written, but the substance isn't there, 0 linkage back to any original documents.

This begs the question of... do you have any original documents to link yourself? Like i said. I don't know how it went down, and the internet is still up in arms. All i know is, this scenario is the more largely accepted one, and has more links in google regarding it. The lesser used claims are: excessive stress on engineers, and the shoddy rasterization i linked above in my previous post. We do know Intel and Nvidia had some legal troubles with each other:

 

http://allthingsd.com/20110110/intel-will-pay-nvidia-1-5-billion-to-maintain-patent-peace/

http://www.cnet.com/news/intel-vs-nvidia-the-tech-behind-the-legal-case/

http://www.intel.com/pressroom/legal/

 

It also is not hard to derive that NVidia wanted to damage Intel in return, and did so by pulling their IP agreements (even if they bought themselves out to do it, like you claim). This would not be out of fear of Intel bringing a world of hurt in that market, but instead, out of childish spite for the legal war Intel started. Remember, NVidia has a history of doing this exact same "counter-suit" strategy. Look at the Nvidia vs Samsung case. 

 

Seeing as you are a huge fan of Inductive Reasoning, my logic behind this should not be seen as impossible, right?

[Syntaxvgm]

OP, good post, however 'Hyperthreading' is Intel's word, it's only become synonymous as they are the only ones to provide it. The general term (until AMD brands their own) is SMT

[Tomsen]

This was a nice find by Mathias.

 

Something we can conclude by this:

• Zen have a 10-wide execution engine (EE). Haswell's EE is 8-wide.
• 4-wide decode. Haswell have 4-wide decode too. Will AMD still use their 'standard' decoders meanwhile Intel use a combination of complex and simple decoders?
• AMD continue with dedicated schedulers. This time for INT, L/S, and FPU. Meanwhile Intel use a unified scheduler for the entire core.
• More realistic cache scheme compared to bulldozer. However, we are still very limited in terms of information avaliable on this topic.
• Reasonable SIMD/FPU. Worse theoretical highest throughput than haswell. However, FPU/SIMD workload, by far is not determined by highest theoretical throughput.
  FPU/SIMD performance will be determined by workload. Could be better than haswell at some, meanwhile be worse at other.

There are still very much information not been disclosed, but so far, this news is pointing in the right direction.

My biggest worries been their memory subsystem and implementation of SMT. 

There is also a lot of misconceptions going around in this thread.