Discussion in 'Article Discussion' started by bit-tech, 19 Sep 2019.
No Zen2, no buy.
Still think these are absolutely cracking chips for their target audience - 4Ghz+ and 4c/8t is more than enough general CPU grunt for a family/kids PC to power Windows/Office/Youtube etc, and the iGPU is genuinely powerful enough for light gaming. The low power consumption and heat output are just the cherry on the cake.
Zen 2 APU's, if they come with Navi/RDNA graphics, will be a game changer though.
Sorry to be that guy, but when it comes to the benchmarks wouldn't it be best to benchmark this against lower end CPUs and/or older higher end chips? It's a given that they'll be slower than the latest Core and Ryzen chips, but nobody who is looking at one of these is realistically looking at one of those. Seeing how it does against something like an i5 4570 bolted to a GTX 960 or 1050 is surely a more useful comparison for the budget, since they're in the same sort of ballpark for total cost.
I can't see that this level of CPU would be aimed at anyone wanting to do more than post on social media &/or watch the odd YouTube video; running 'real work' productivity or AAA gaming benchmarks would never give any stellar results, as those with this kinda budget won't be after that kind of use case.
I'm looking to get an APU for a portable gaming rig for Overwatch and CS:GO, something small and light
in other words its a decent choice for replacing my dad's phenom II x3 with a geforce 520 ... which is kind of what the lower sku ryzens always have been for ... mean seriously what do you guys want a threadripper with geforce 1060 igpu's for less than most ram kits?
Wouldn't say no.
Exactly this - CPU's should have a ceiling on what your results/comparison lists include for any give part, to just show 'target market' use cases.
Having said that, though; are Zen 2, non-G parts better matched to Intel's -F parts? If my deductions from this premise are correct, are non-F Intel chips effectively APU's, too...?
And how would you go about retesting the age-old hardware? You can't simply use the old numbers, because they might not be accurate anymore. And where do you draw the line? 2 years old? 5? 10?
Not saying I don't want it, but I don't think it's realistic.
A universally agreed, multi-version benchmarking system would need to exist that could categorically state any given version of it is either too weak for meaningful results or out-of-scope for the part you intend to test with it. Take Cinebench; at what point is 15 too easy - or 20 too hard - for any given CPU? If that was or even could be universally agreed, then maybe a similar metric could be applied to other similar software.
That said, is there any (benchmarkable) software that is still largely the same as it was 10-20 years ago that has not been updated to exclude newer versions of parts it's intended to test? That works equally well for i5-2xxx as it does for i5-9xxx...?
Xeon & Threadripper/Epyc et al would have one set of benchmarking to test their intended use cases, but sub versions for chip generations... the earliest Xeons won't fair favourably against the latest Threadrippers, so a timeline cut off would make sense.
Same would go for GP CPU's - there would need to come a point with these use case parts where generations of parts would need some kind of divider, to infer that a new part B is definitely an upgrade for 5-10yr old part A.
Separate names with a comma.