I agree with you, I think you will get another CPU upgrade on AM4. It already has the tricks which drove intel chipset/socket evolution and AMD settled on PGA, so I think that's it for a bit. I hope it goes well for you Fascinate, I think it was a bold but not a silly move.
I reckon once foibles are ironed out Ryzen will be in a good spot going forwards, even for gaming. Doom is currently the only game which has done Vulkan properly (even id are learners though), the others have hacked in a temporary abstraction layer to cover what isn't now handled by the API/drivers. The way most engines are handling/distributing render and physics pipelines is basically as they were over dx11 (dx12 also suffers from this). It's only in the new generation of engines, when we have stuff written expressly to properly utilitse threads,m_alloc, command buffers, fences, semaphore etc will we see the new API's shine. Logical thread count will overhaul a slight ipc + clockspeed advantage at some point on this curve, as it does now in straight computation.
The downside is that all that takes a lot of work; some truly horrible engines can be created using either low-level g-api and there isn't a gpu driver dev in the world that can save them. The upside is that engine dev's have far more visibility of the hardware, it's now up to them to determine how to address it. We will see fewer engines initially. Addressing a low-level api is beyond the scope/ability of a typical game dev, and until there is additional open resources available (dx12/vk higher level abstraction layer templates) it's the big engine houses which will do the work.
A lot of vulkan development is coming from mobiles; opengl es became neglected with khronos' attention on vk, but the mobile platform offers itself to parallelism anyway, and vk will much improve things on newer devices there. A good chunk of a vk implementation in a mobile engine can be directly re-sourced for desktop, and it's platform agnosticism seem to make it a better option to dev for than dx12 (though no striped multi-gpu for vk yet). Feck x-box and it's hacked dx implementation, it''l soon be the only platform steadfastly not supporting vulkan (Apple will have MoltenVK).
It's built on high density libraries and in an incredibly power efficient process, efficiency and heat output are really great because of that =p
Downclocking Zen makes it even more ridiculous, the perf/watt around ~3GHz is basically insane.
And yeah, it didn't really make any sense when someone said here the 1700 was magically cooler. It had no reason to, the 1700X and 1800X can even reach the same clocks are lower voltages which invalidates that even more.
I hope they dont continue to prioritize low power/eff at 3 Ghz over 4.5-4.6+ clocks for Zen2
Im not a server, Im a home user/gamer and those clocks would be greatly appreciated
elsewise Intel will still be ahead in games no matter IPC improvements, especially when they roll 6c into mainstream
Noticed this on Reddit:
It looks like the article on Guru3D that contained it has been removed but reddit still has the image.
Edit
And some more:
http://imgur.com/a/FGkBF
And this:
Last edited by Gray_Matter; 2017-03-15 at 09:23 AM.
Hmmm leaks were wrong then, pretty sure leaks had the 1500 being a 6 core part.
Here's to hoping the 1400/1500X are actual 1 CCX dies and not a dual CCX part with 4 cores shut down.
This would decrease CPU complexity and very likely offer a more potent OC-able chip.
If it's the same die you'll get identical results as the Ryzen 7s.
Edit:
Though I honestly believe that Ryzen 5 is the same part and Ryzen 3 is the single CCX die.
Nvm i am a retread, i was looking at vid not vcore.
I really dislike HWinfo, i hope HWmonitor gets their stuff together soon so i can go back using that lol.
Last edited by Fascinate; 2017-03-15 at 01:29 PM.
The AMD subreddit is going crazy. Someone needs to remind them the wait for benchmarks. Intel is still king.
As I said a couple of weeks ago, the R5 series is where things get interesting. The 1600x looks to be an 7600k killer. Ten bucks more for 2c / 8t more? Even if the IPC is 5-10% lower I would take that deal any day of the week. The 1400 just roflstomps the 7300 / 7350k. Who in their right mind would choose a 2c / 4t chip, even with higher IPC, over a 4c / 8T chip?
Last edited by ovm33; 2017-03-15 at 04:53 PM.
I sat alone in the dark one night, tuning in by remote.
I found a preacher who spoke of the light, but there was Brimstone in his throat.
He'd show me the way, according to him, in return for my personal check.
I flipped my channel back to CNN and lit another cigarette.
There is more to the story than who is "king". Very few people can afford or even want to buy the fastest processor. For the most part, it's about price/performance. A strong price/perf option from AMD is good for everyone, even those people who wouldn't touch AMD with a barge pole because at the very least, Intel will drop prices to remain competitive.
Yea, I don't see a reason to buy Intel at this point. You pay $170 for what is essentially an i7. Good luck convincing people to buy Intel.
I actually don't think arrogant Intel will drop prices. The R5s are due in 3.5 weeks, the R7 has been out 3 weeks, they haven't budged on pricing. They must think if they do that means defeat and Intel cant handle that lol.
I agree, no more Intel, at least until pricing and matched core/thread options become available to the mainstream market. Dual cores are so outdated but they stayed due to no competition. Those days are over.
Yea the i5's and i3's need some serious repricing, and the i5's need hyperthreading. I don't even know what they'll do with Pentiums. Their quad core i7's need to go down tremendously in price. Then and only then can anyone here debate me about buying an Intel CPU.
This is Intel right now.
They used this slide some time ago:
And it is completely pointless if they're not releasing the 4/8 SKUs as single a CCX. One of the reasons they even went with CCXs was that they could split Zeppelin in two and have working quad-cores too. The other reason is being able to stack more of them and make larger SKUs like Naples.
But I mean, who knows. I'm saying "supposed" because that's what logic implies, but who knows really. I'm lazy to go look if they made any other statement of any kind.
Zen 2 will have to convince me to buy it over 6c Coffee, not the other way around
Having a good power efficiency will allow them to to push further, though. Hopefully, the new node-size does ineed lend itself for better clocks.
They also set up a target for IPC with Zen - they told their engineers that you needed 40% more IPC than Bulldozer.
The engineers said this was too high of a goal, but I think they landed somewhere around 52%~ right?
This is a huge jump and a great first step.
They'll only grow better once they don't have to pull something magic out of their hat and had to discard their entire CPU lineup and architecture and are able to reiterate and improve.
I haven't had a desktop AMD CPU since the AMD Athlon XP+ 1600+, but when I need to upgrade my CPU in a few years, Zen 3 or Zen 4 is looking like a good option unless something magical happens on the CPU market in terms of performance, my 4770k will have to chug along with its poor clocking performance.