Colsta
New Member
Sorry about that although I'm unsure if that's down to the subject matter or my cold tempering my ability to explain it
Your actual choice of RAM of course is going to be tempered by price/availabilty, performance and possibly power consumption (every stick removed is power saved). Technically the first choice (2 x 16GB) would likely have a shorter data path between the pooled RAM & CPU than 4 x 8GB would expose but this advantage is vastly outweighed in practice by timings/frequency choice elsewhere - ignore this point. More on timings in a moment.Overall there is a small *performance* advantage (~5% in computationally intensive scenarios) in running your memory in dual channel mode rather than single channel mode. So assuming 4 slots are available, 2 x 16GB or 4 x 8GB are valid choices. Of these, the first choice would use less power (cheaper to run) and leave a future upgrade path. However the second choice may be cheaper to buy (smaller matched modules "bundled" can often work out cheaper) but with no upgrade path you'd have to choose the minimum size wisely.
Returning to frequency/timings, it's worthwhile getting these factors right which can offset performance deficiencies with running single channel or enhance your performance of dual channel further. The Ryzen 1600 supports up to 2667MHz DDR4 RAM out of the box so there is little point buying faster RAM unless you intend to overclock. However at this frequency, Wikipedia shows four options available that differ in CAS latency (a grossly oversimplified but accessible explanation of this feature is the number of cycles it takes to read/write to the RAM, so the fewer cycles, the better). They are 17-17-17 (12.75 nanoseconds CAS), 18-18-18 (13.5ns), 19-19-19 (14.25ns) and 20-20-20 (15ns). Invariably, the lower the latency the more costly the RAM. I don't have any hard data to hand to quantify this but I'm sure you can see that when you multiply up the sheer number of read/writes that occur to RAM in any given scenario, those nanoseconds quickly add up to another measurable improvement. The main takeaway point is not all RAM is equal even at a given frequency.
To summarise then: in your chosen scenario, with 4 slots to play with and providing you can afford 32GB as your first step, then the optimal configuration *for performance* would be 2 x 16GB with the lowest CAS latency you can find/afford. This has the additional bonuses of being the most power efficient whilst leaving 2 slots for future upgrades. (FYI - my own use case estimate doesn't require more than 32GB for the lifetime of the system, so I would follow the same logic but choosing 2 x 8GB to start with and hoping that RAM prices come down to more sensible levels before I need the upgrade).
Personally if I had Aston Martin money then I'd buy a Vincent Black Shadow! But I get your point. I'm sure we all know people that blindly follow the hype and consequently have a core i7/Nvidia 980 beast to browse the Internet. I believe in matching the tools to the job and avoiding unnecessary expense
If overclocking isn't entirely off the table then you may want to get faster (frequency) RAM upfront after all.
I know Ryzens contain a pair of four-core complexes (CCX) from which lesser models are created by disabling cores e.g. Ryzen 1600 has one core nobbled per CCX. I *think* (not fact checked) that each core gets it's own dedicated L1 and L2 cache leaving the L3 cache to be shared equally between each CCX. But whilst 8MB shared between 3 cores means each may have access to a larger portion of L3 before hitting RAM, I would think it far more likely that the fewer physical cores left can be pushed harder within the same 65W power envelope yielding a better per core performance (though not as good as all cores active of course).
The MoBo's only expose the iGPU incorporated in the CPU package. Intel's have them, Ryzen's don't. I guess the logic is sound insofar as Ryzen's were aimed squarely at gamers/power users who will want their own dedicated GPU anyway and the power that the iGPU would use can be channelled into powering the extra cores.
The Ryzen APU's are already overdue. The CTO of AMD promised “beautiful thin and light” laptops along with “AAA gaming performance”. If true to form, I think they and their desktop counterparts will be a compelling proposition for people like myself that don't want or need the power draw that separate dedicated CPU & GPU's take.
, but I didn't think you'd be aiming that high.