>>52675453 we already know they're behind. their first 14nm is coming out all of the way in like october IF they're on time while skylake already came out last year. intel will have their second gen 14nm out by the time amd finally shits out their 14nm.
>>52674688 Zen based chips will be a major source of revenue for AMD. The enterprise market with Opterons and big package APUs, then the consumer market with the Raven Ridge APU. GPU sales are harder to gauge.
>>52675503 Broadwell Core M went into production sometime in early 2014, and were on the market by the end of year. Skylake is their 2nd arch produced on their 14nm process.
>>52675971 Stop before you dig an even deeper hole. Hyperthreading tags ops with registers and simulates another thread on an unoccupied ALU. AMD's Bulldozer family uses separate hardware integer cores. Hyperthreading is a marketing name for intel's specific implementation of SMT. AMD's current chips are CMT, not SMT. They are not even remotely the same.
>>52675972 Polaris will probably be fine unless there's some huge fuck up in the fab or a lack of rops which is standard radeon trouble. But it will just be GCN on a new node, gcn3 made some decent strides with power usage (not as much as maxwell), so expect around 2x performance for the same amount of silicon. It'll go toe to toe with pascal most likely.
Will amd's gpu drivers be good enough when polaris comes out for their gpu's to be somewhat close to nvidia's in performance, on linux? I heard they were currently terrible. I read a bit about AMDGPU making open sauce only drivers for their new cards, but I don't know if that will result in decent performance gains.
>>52674911 You do know that console drivers/apis and pc drivers/apis are completely unrelated right? Dx12 won't make porting from console any easier, maybe even harder in the start, because it'll be new developers. Seeing as so many games are first made for consoles is already a sign that game companies are massive cunts, so I don't expect any improvements in terms of optimisation.
Zen will reach somewhere between 70 and 90% of Kaby's IPC with slightly higher power drawings, their top chip will be sold at slightly lower price points to Intel's offering, with a slightly debunked runner-up being significantly cheaper and serving as a budget option for enthusiasts. Competitive mid-tier offerings. AMD will finally be an option again, but they're not going to trump Intel.
On the other hand, their Excavator APUs might suffer from a potential success of Zen.
I doubt that Polaris will end up worse than where AMD is now, respectively to GCN vs Maxwell. At a best case scenario, it's going to beat Pascal on slight margins. Nvidia's likely not gonna do any experiments, but neither will AMD.
>>52675332 Doesn't matter much, as Kaby is going to be 14nm, so likely your usual 5-10%.
>>52675453 Predicted 40% on Excavator, you fucks on /g/ still cannot comprehend it and measure on Piledriver.
>>52676039 Learn to read please, it's a definition. A core operates at index 1 regardless of brand, hyperthreading operates at index 0.23, AMDs additional cores operates at index 0.5. When you buy an Intel 8 core cpu you get an 8 core with hyperthreading. When you buy an AMD 8 core cpu you get a 4 core cpu with 4 semi cores.
>>52674688 Zen has potential to be interesting if they hit 'close' to Intel IPC but sell a CPU with more cores. An 8C/16T or even 6C/12T Zen CPU that OC's decently at the typical mainstream unlocked i7 price point could be a very compelling option.
As for Polaris, as a 290X CF owner I have essentially lost all trust and hope for AMD drivers. It could very well be very nice hardware, but I'm nearly certain that software support will be crap, both from AMD's side (especially in DX11 which is going to be quite relevant for a few years still) and from game devs' side (since almost all of them seem to deeply enjoy getting spitroasted by NVIDIA).
>>52676443 Summit Ridge isn't competing against mainstream i5s and i7s. They're competing against the i7E line, Summit Ridge is HEDT not mainstream. The chips aren't going to clock high without TDP making significant jumps unless you're dealing with a binned quad core part.
>>52674911 >Also what people forget is that consoles running amd system will be easier to optimize the ports for amd pc setup. >MGSV was designed for mainly the PS4 console >PC port uses NVIDIA GameWorks Yeah no, people will only switch their rendering engines to use bullshit over-tesselaters for money.
I really hope that AMD starts closing the IPC performance gap with Zen but I'm afraid they won't and they'll just fall further behind. We need some good competition in the CPU market. atm the only thing AMD has going for it is low prices.
>>52676463 It doesn't matter that the API differs. What matters is that similar optimizations have similar effects on both platforms (i.e. some optimizations can actually harm performance on one setup and not the other, another optimization can have virtually on effect on one platform and provide a significant performance boost on the other, and so on).
>>52677280 Yet that is not happening currently, precisely because of API differences (DX11) and some design differences as well. APIs may get more similar, but the design differences will remain, most notably the fact that the CPU and GPU do not share the same physical memory on PCs.
Currently AMD is often struggling to provide good performance in a lot of console ports, while NVIDIA is often ahead. Their drivers are frequently late and subpar, be it because of DX11 overhead, Gameworks, insufficient resources or even incompetence, AMD isn't reaping any benefits on the PC front because of the console hardware.
>>52677421 The post I initially quoted said that AMD having the consoles means easier/better optimization for AMD on PC. But that does not happen, it is not a real thing. It doesn't even matter why, it's not something that you as a person who would buy an AMD card can benefit from. It's not real.
>>52675453 Each module (there are 4 of them in the 8350 and 8370) has two cores. The two cores in each module share the same L1 and L2 cache with the entire module. Windows registers it as 8 core, yes, but it's really only 4 cores.
The same goes for every AMD FX chip. The quad cores has two modules with two threads per module. The two threads in each module share the same L1 and L2 cache pool.
The 6 core AMD FX chips have three modules, with two threads per module; again, each module sharing the same L1 and L2 cache pool between both threads in each module.
Also, you need to quit talking out of your ass. I have an AMD Phenom II X4 965 and I run Blade and Soul just fine with no stuttering at silky smooth frame rates. I didn't have my frame rate counter on, but I know it didn't drop anywhere below 45 FPS, and that's with a fairly old GeForce GTX 670.
The Core i5 only thrashes the AMD FX chips on certain games. Some games it's the other way around, with the AMD FX 8350 taking chips such as the Intel Corei-i5 to school, but you never hear any talk about that. Fanboys only want it known when Intel dominates. That being said, there is only about a 20% difference between an FX-8350 and a Core i7-4790k. To say AMD would still be behind if they get a 40% improvement is just talking out of your ass.
AMD went with a new microarchitecture with the AMD FX (and APUs) chips and it didn't succeed as well as they hoped for. Now they're going back to a traditional multi core architecture and I can guarantee you things are going to be a lot different with the AMD Zen. Unless you of course are forgetting that AMD is going back to their roots with the Zen processor.
Also, do not forget 2001 where Intel struggled to keep up with AMD. http://www.anandtech.com/show/835/11
Thread replies: 53 Thread images: 4
Thread DB ID: 476135
All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.
This is a 4chan archive - all of the shown content originated from that site. This means that 4Archive shows their content, archived. If you need information for a Poster - contact them.
If a post contains personal/copyrighted/illegal content, then use the post's [Report] link! If a post is not removed within 24h contact me at firstname.lastname@example.org with the post's information.