241–250 of 359 posts
Out of curiosity, why aren't CPUs with asymmetrical core designs a thing? For example: a single, large primary core with a high clock speed supported by many lower-clocked smaller cores. Theoretically that would allow for the best of both worlds. I imagine if such a design existed, it would disincentivize proper multi-core utilization in software to some degree. It would likely further complicate parallel programming…
The major difference is that the PPE and SPE's have the same clock speed, but it is definitely a heterogenous architecture.
Earlier quoted context omitted.
> Can the Broadwell use the 128MB EDRRAM from the iGPU as a L4 cache if the iGPU is not used? Would that actually make such a difference? That's exactly what it does, and it gives you a "free"* ~20% increase in memory operations performance (which tends to be a ~20% increase in general performance). * not actually free because you paid for those 128MB EDRRAM...
Nice, now I'm wondering why Intel doesn't slap 128MB EDRAM on an Kaby Lake i7 4 core 8 thread and then release it for the enthusiast platform x99. Maybe even select chips to make it possible to release them at 5ghz. Of course with a large premium (expensive EDRAM etc.).
I'm still trying to work out whether only dual channel RAM means 64GB RAM is the limit, or do different mobo chipsets support actually support quad-channel? Kind of a deal breaker (for bioinformatics work) if I can't go over 100GB RAM, which is really annoying given those cheap threads.
Barely half of the PCI-E lanes than my FX-9350. Nooooooooooooope. Was looking forward to a nice SLI rig but looks like Ryzen just died for my choice on upgrade. Guess I'll stick with the power-hungry beast.
Earlier quoted context omitted.
GCC is part of spec2006, which they optimize for. But the actual compilers people use (IE not that version of GCC) are rarely well-optimized for CPU. In fact, most compiler profiles are not flat yet, and where your compilation takes a long time, it's because some optimization or algorithm has gone nuts and needs to be fixed. Until that kind of state of the world changes, measuring compile time is IMHO, a bit pointles…
GCC is a useful benchmark not because it is unoptimized, but because it represents a real-world work profile that is otherwise poorly represented in SPEC. The working set sizes are large relative to the cache size, and its composed of a bunch of pointer-chasing branch-heavy logic as it winds through the various trees and intermediate languages. That profile is quite a bit closer to the profile of common business and…
Sure, in that sense, but you also don't want benchmarks that are trivial to game by performing optimizations to them. :)
Earlier quoted context omitted.
GCC is a useful benchmark not because it is unoptimized, but because it represents a real-world work profile that is otherwise poorly represented in SPEC. The working set sizes are large relative to the cache size, and its composed of a bunch of pointer-chasing branch-heavy logic as it winds through the various trees and intermediate languages. That profile is quite a bit closer to the profile of common business and…
"GCC is a useful benchmark not because it is unoptimized, but because it represents a real-world work profile that is otherwise poorly represented in SPEC." Sure, in that sense, but you also don't want benchmarks that are trivial to game by performing optimizations to them. :)
If someone managed to make an optimization that could suddenly make GCC 2x faster without humans editing GCC's source, that would be a major CS breakthrough :)
Looks great, coup successful by AMD. Takeaways after a few of the reviews on this list[0][1]: - If you're buying a desktop quadcore processor in 2017, you're doing it wrong. May as well buy a laptop. - Kabylake sucks as a workstation chip but is ok for gaming if your PC is solely a glorified Xbox. - Vast majority (90%+) of gaming rigs are GPU-bound. Think very carefully before giving up 8 cores and 16 threads at the…
Not true, IDEs, web browsers, office tools need single core performance and not 4+ cores. A laptop is more expensive, will make more noise when loaded, is harder to upgrade, is less customizable, has less ports, is more expensive to expand (ThunderBolt devices are not cheap at all). What are you talking about?
>> Intel's margins and lineup are utterly destroyed especially once you consider how expensive Intel HEDT boards are on top of it all.
There are few people who build workstations for themselves(even among software developers). HP and Dell haven't announced any Ryzen systems as far as I'm aware. Right now AMD's new chip will have minimal impact on Intel profits. Only if AMD can remain competitive on stability, performance and price for a few years will Intel start loosing major chunks of income and market share.
>> Kabylake sucks as a workstation chip but is ok for gaming if your PC is solely a glorified Xbox.
PCs are universal machines, you can play games AND do work on the the SAME PC ;-)
>> Vast majority (90%+) of gaming rigs are GPU-bound. Think very carefully before giving up 8 cores and 16 threads at the same price.
GPU-bound does not mean all CPUs produce the same framerates on a given GPU.
I personally do hate Intel and nVidia monopoly and hope AMD can increase their market share, but your points are very weak.
Impressive! While it doesn’t utterly destroy Intel, AMD does offer a MUCH better price/performance ratio. Things have gotten much more interesting indeed. Intel’s dominance is being called into question. As a result, we all profit.
It seems that in common use (web browsing, office, gaming), fewer but stronger cores still shine and even the fastest Ryzen is slower and more expensive than Intel's offering.
So it really depends. A lot.