There are even rumors that Intel delayed the much more important Xeons to 2020, which could give AMD a 1 year head start with Zen2 on 7nm in the server space (comparable to Intels 10 nm).
> If the rumor [0] that Intel's first 10nm server chip will only release mid 2020 is true, then AMD's shares will probably skyrocket again if they truly can release their Zen 2 server CPU mid 2019 (on 7nm which is comparable to Intel's 10nm).
Do you mean the single threaded part or that OP claims that it's what most users desire? If it's about the single threaded performance then OPs claim is true.
In single threaded games this can be even more apparent. But Intel is mainly faster because of a significant clock advantage. Clock for clock the advantage is small.
I'm looking forward to Zen 2 which could allow AMD to close the IPC gap or even get a bit ahead of Intel. Combined with the 7nm process which should allow better clocks it's looking good for AMD.
Question is if AMD uses the high performance process which is touted to allow up to 5 ghz [0] at the cost of increased power usage, or if they use the low power process (which is also more affordable). Since Ryzen, Threadripper and Epyc use the same die probably the latter since power consumption is so important in the data center.
I (and many others) would take a 33% increase in core count over a 10-15% advantage in single core workloads. Even the gaming comparison is most of the time overemphasized. Does it really matter if you are getting 88FPS or 80FPS if your monitor refresh rate is capped at 60FPS?
I feel that most of these attacks at AMD and their CPU-s tend to bend the truth a lot, and that isn't surprising, given the shady tactics that Intel has employed previously.
> I (and many others) would take a 33% increase in core count over a 10-15% advantage in single core workloads. Even the gaming comparison is most of the time overemphasized. Does it really matter if you are getting 88FPS or 80FPS if your monitor refresh rate is capped at 60FPS?
Luckily it seems AMD can probably deliver both with Zen 2. For Epyc the roadmap suggest a 50% increase in core count [0] over Zen 1, and since Ryzen and Threadripper uses the same die as Epyc this will most likely trickle down to Ryzen 3xxx and Threadripper 3xxx.
Depends on your game and monitor I guess. Quite a few people seem to buy ugly 144hz gaming monitors and some games like Warhammer and Far Cry 5 can have ~20 fps difference at the same clocks at 1080p.
> I feel that most of these attacks at AMD and their CPU-s tend to bend the truth a lot, and that isn't surprising, given the shady tactics that Intel has employed previously.
Most of the time the limiting factor is not the CPU but the GPU anyway. The good thing about AMD closing the gap or potentially even overtaking Intel in single threaded performance is that the fanboys, fangirls and corporate shills have one argument less for spreading FUD and suggesting Intel over AMD.
[0] there are even (unlikely) rumors floating around about a 64 core Epyc, but I don't see that happening before 7nm+ on EUV.
> Does it really matter if you are getting 88FPS or 80FPS if your monitor refresh rate is capped at 60FPS?
It does when you want "the best of the best" for gaming on an expensive 144Hz display, and also spend a ton of money on a GPU to handle that. You don't want your single-threaded CPU performance be the bottleneck in your games then - so for a top-of-the-line gaming PC, Intel is still the best choice.
I don't really follow the ML space, but when you have a good GPU in your PC (which I assume quite a few people here have) you can learn/use Tensorflow by running things locally instead of paying for Google Cloud etc. Why pay extra money if you already have a good GPU in your computer (unless you really need to decrease the time it takes to run the computation etc.)
If the rumor [0] that Intel's first 10nm server chip will only release mid 2020 is true, then AMD's shares will probably skyrocket again if they truly can release their Zen 2 server CPU mid 2019 (on 7nm which is comparable to Intel's 10nm).
AMDs Zen architecture uses 32KB for data and 64KB for instructions (was curious if there are differences between AMD and Intel designs regarding L1 cache).
Since I read about this a few years ago I really want to try it, but I don't want to buy an extra GPU for it. I hope AMD [0] brings their SR-IOV implementation called MxGPU down to their mainstream GPUs, which allows to split a single GPU between host and guests. Apparently this would also be more secure than passthrough?
In order to not affect their pro GPU sales they could maybe limit the number of virtual GPUs from 16 to 2, which would be enough for the host and a single guest.
[0] or Nvidia, I don't care, but since Nvidia is the market leader they have less incentive than AMD.
Another difference between e-bikes and scooters is that e-bikes do not require a license to drive while scooters do. One of the reasons for the speed limit on e bikes is to avoid this license requirement.
Copy paste from my posting yesterday:
EDIT: it seems the twitter post got deleted.
EDIT2: anandtech still has the pictures: https://www.anandtech.com/show/13119/intels-xeon-scalable-ro...
> If the rumor [0] that Intel's first 10nm server chip will only release mid 2020 is true, then AMD's shares will probably skyrocket again if they truly can release their Zen 2 server CPU mid 2019 (on 7nm which is comparable to Intel's 10nm).
[0] https://twitter.com/david_schor/status/1022142835989118977