Discussion in 'Article Discussion' started by bit-tech, 22 Nov 2018.
I'd say that currently, if anything, we have more cores than we really need, not too few.
IPC boost or Ghz boosts are what is needed, I have a Ryzen 1600 and I know I don't use all 6 cores to the max (no content creation to speak of) but wanted to support team red (plus cheaper motherboards). However a Zen 2 with a healthy speed bump for the same core count would be tempting (even if just to scratch the upgrade itch).
No. AMD needs to match Intel in IPC and clocks. Or better yet, beat Intel in IPC so we don't have to have chips being released already clocked to 95% of what the silicon is capable of.
If they keep the separate I/O design I'd much prefer to see them pair a chiplet with some form of Vega, although you can get Ryzen with Vega in the form of embedded and mobile chips it would be nice to see entry and mid-range desktop parts come with some sort of graphics capability.
You mean, like the ones you can already buy?
From AMDs pov I think they should. Hammer home their current tangible advantage/selling point and hope that the software (ie. games) follows asap. Increasingly multithreaded is the future.
Fix memory issues
Moar PCI-e lanes
The problem is, it's been "the future" for years now. It's only now we have mainstream 8c/16t parts that 4 cores have realistically become the minimum... and everything is catered for the minimum. It will be another half a decade, when 20-30 cores/threads are the mainstream "high end" that 6/8c then become the new performance "floor".
I think the 2 and 4 core mainstream CPUs we've had up until now were designed around a single core mentality. If increasing core counts becomes a primary focus for increasing performance as CPUs start to bump up against physical constraints for increasing single core performance (at least at the rate people have come to expect), then I think core counts could increase at a much more rapid rate. Perhaps doubling every two years?
AMD need to do one thing, keep it cheap.
No. Improve IPC, keep them good value and i'm a buying (so long as I don't cave into black friday...!!!)
higher platform costs come with threadripper because it is an actual better platform, well stone me.
Do we need more cores, not for me no, do we need more performance, depends on use case but for me nope, I have had the same IPC for years, we do need a platform that can hang with todays hardware...yes, that only comes from one place.
9900x is the same price as a 1920x and an x399 mobo, I’d take the old slow chip thank you very much, I know there's more power and more cores if I need it, but as it stands I have a platform that can handle actual the requirements for bandwidth in a modern system and you know what, Gen1 is not a shabby processor. Plenty good enough to push pixels or encoding.
Blimey, imacs are adopting NVMe raid for performance....Apple is never cutting edge hardware, just shows you how normal NVMe is as a bit of hardware.
Go on! Cave in with me!
I couldn't be happier with the IPC. It's more than good enough IMO. Now if you are asking me what I would like? well, how about 4.5ghz for a start. That would even things out a bit. But seriously other than that? no one needs more than it already offers. We just need it to remain cheap, as let's face is that is what 99% of buyers care about. I think the single most amazing feature of Ryzen is how well designed it was to cost AMD little to make, with a big success rate. Even the failed dies end up in TR chips to spread the load. That's quite tremendous efficiency IMO. Like in manufacturing.
As for more cores? I'm a complete core ****, but for right now I would rather see 8 cores being pushed to the limit. Then maybe we could have a discussion about whether I would like to see more cores (of course I would, but at the same time I like the prices...).
The thing is IPC improvements are hard, adding more cores is much easier - hence all super computers going from 1 very fast core to tens of thousands of slower cores. However from a programming point of view using more IPC in a program is very easy and tends to be very efficient. Using more cores is much harder and less efficient.
How many people do you need to change a light bulb quickly? - 1 very fast one or 20 slow ones all crowding around shouting at each other and getting in the way as they try to each do a different bit of the task?
Anyway obviously for AMD IPC is more important then anything as they are behind their main competitor, but it's so hard to improve.
Ah, but how many people do you need to change 20 lightbulbs quickly?
Forgot about that one.
Having said that it's based on the original Zen instead of Zen+ and afaik (obviously i don't know much so that means diddly) there doesn't seem to be updated versions, it's almost like integrated graphics has become an after thought.
Now be fair, there's actually two of 'em (and they're identical, except one's downclocked to drop the TDP from 65W to 35W.)
It took quite some time after Zen came out to see Zen-based APUs, and Zen+ ain't *that* old - give it time. I'mma probably be holding out for a Zen 2-based APU for my next upgrade, and if one don't come along then AMD's lost a sale.
That will depend on how closely spaced those 20 bulbs are.
Separate names with a comma.