Discussion in 'Article Discussion' started by bit-tech, 19 Apr 2018.
Single thread is still weak sauce compared to my 4yr old 4690K...admittedly at 4.5Ghz, but still...
Maybe Zen 2 in 2019 will fix that, hope so because other than that I'd have no qualms about switching over to AMD.
There is a 2700X on my agenda very very soon, thank you once again AMD for turning the market on it's head - Now to just wait on the board reviews
Not enough to convince anyone but the fanboys to upgrade over the equivalent gen 1 chip (1600X to 2600X etc), but those that sat out AMD's first swing of the Zen hammer and still on ageing Sandy/Ivy rigs and want to join the rest of the world with nVME and USB 3.1 have a solid option.
Hoping AMD are saving up the very best chips for an all-core capable 4.5Ghz 2800X.
In how many real world applications where overall performance is actually limited by the CPU is single core performance the main factor?
Has this review taken into account the updates to patch spectre etc? Edit Doh.... browsed over that section
..in the world of emulation, single core speed & IPC are king.
The thing is it isn't really about the business side of things for many. What I mean is, head over heart etc. I can see a lot of people buying these who have existing Ryzen CPUs mostly because they can, without having to take their mobo out and etc. £300 is not a massive sum for the extra clock speed. I just wish Intel knew that not every one buys their CPUs because they need them. Maybe then they could mop up more sales to people just wanting to buy something new, without having to buy a new board etc.
I mean yeah, some people are still on Sandy. But, there is quite the difference between need and want
My 1700 can’t do past 3.8 reliably, so a boost to 4.2 or 4.3 would be great. As above, it’s not that much cash... and the 1700 can always be repurposed or sold to make the difference even smaller.
<looks sadly at A10-5800K>
Time to start saving so that when Zen 2 APUs are mature I'm ready for a build, methinks!
£300 for an extra 400-500Mhz and maybe a 10% jump in performance? Your cash to burn, lads.
I know people who went Sandy - Ivy - Haswell ETC. They probably got a lot less for their money.
on a 4770k so not for me
I do not see AMD ever catching up to Intel on single threaded performance, it's been well over a decade since they had a processor that was top in that regard.
But their inability to catch up on that front is what lead them to focus more on multi threaded performance, which is turning out pretty nicely for everyone who wants a PC for something else than gaming. Unless Intel's new CPUs are pretty special on mult threaded and extremely aggressively priced AMD is getting the creative market for the next couple of years.
I went Sandy, Ivy, Haswell, back to Sandy. Not sure what possessed me but I did.
Ryzen next, multi threaded photoshop performance is too much to ignore.
I went through every Intel gen from Sandy and up through Skylake too. Part of being a reviewer and enthusiast...
Besides, yeah, I do have cash to burn. I haven’t properly treated myself to any expensive gear in months. I work 40 to 55 hours a week. About time someone other than the wife gets to spend!
I don’t consider the chip to be costly because the ROI will be fantastic once I get a few contacts sorted.
Plus, one further point... $329 US for me, not £300 ($420+)
Until you turn up the resolution and or crank up the AA / AF, then the CPU barely matters at all.
Of course that is a double edged sword for Ryzen, since it also applies to to ancient CPUs and removes reasons to upgrade.
For example at 1440p in Shadow of Mordor the 2700x matches the i7 8700K, but then you see the medieval 2600k only 2fps behind...
Good review guys. Shame the improvement has come at a cost of power usage.
Few things I spotted in the review, idle power chart is labelled as Prime95 small FFTs test and load is labelled as windows desktop. Conclusion second paragraph "for a purely= gaming system"
I'm more interested in seeing how good the 7nm versions will be next year.
I know there's that endless cycle of waiting to buy the next version, but when I buy new hardware it has to last a long time. That's a similar reason why I wouldn't buy an new GPU at the moment, pascal is two years old and is due to be replaced. Hopefully the replacement will have quite a jump in performance and be top dog for a couple of years too.
Any application that is not explicitly designed to take advantage of threading (and to take good advantage. And to do so with enough parallelism that Moar Cores comes out on top of a smaller number of faster clocked cores).
Which for the most part means every application developed up until around a decade ago, and only almost every application since then. Tasks like video encoding and rendering 3D scenes (that haven't had their rendered moved over to GPGPU yet) are the exception rather than the rule.
For non-HPC workloads, scaling remains in Amdah's Law territory rather than Gustafson's Law, because unless you have a dedicated process for feeding in additional tasks (i.e. are task-right and compute-poor 24/7, rather than encountering fixed-size tasks) then you do not have sufficient independent workloads available to avoid the parallelism diminishing-returns trap.
Separate names with a comma.