So theoretical max bandwidth is around 4GB/sec. Actual bandwidth is a bit less.
A single 960 Evo for example at 3.2GB/sec would come awully close. You then have to remember that interface is shared by SATA and ethernet connectivity as well. further splitting it.
So yes a single high end SSD could saturate it easily. So even transferring lage files between SSDs or from a network to an SSD would be enough to saturate it. Especially when you consider most X299 boards have multiple M.2 slots⌠Bandwidth gets tight.
So even a normal power user could fill it. Not even super high end pro workloads.
Actuality the 32Gbps you get with a x4 is includes any PCIe transaction packet overhead, and itâs a 4KiB max transfer size, I wouldnât be surprised if the 3.2GB/s transfer speeds on the 960 PCIe were bottlenecked by PCIe
I am not sure what Intel is doing here. Looking at Anandtechs tabel with the TR-1950X and i9-7960X, you get a bit more turbo (not on all cores?), a bit less cache and 15W less TDP (at least Intel claims that) and less PCI lanes, for 700 bucks more!?
The i9-7980XE however⌠for 2000 dollars, you might aswell go Xeon 6130 ($1894) or with Epyc 7401 ($1850) to enjoy features your âprice classâ should be bundled with.
Yeah they should really stop with these obfuscatory terms â just list the possible all-core turbo, max single core turbo; the âbase clockâ is meaningless given that everything now downclocks when idle.
Okay. So youâre saying that 2.0 boost clock thing is a lie? Or is there a bureaucratic reason why avx(2) perf isnât relevant for those turbo 2.0 minimums?
Not quite â it informs you of the minimum power draw under full load. Except it doesnât, because when itâs at base clock, itâs because itâs being heavily heated.
AVX loads use quite a bit more power and therefore canât boost as high. Youâll likely still see some boost over base but not as much. Bar clock is just absolute minimum under worse case scenarios.
If the motherboard has all core enhancement which nearly all do now the boost becomes less relevant since it pushes all cores to max turbo. Basically an automatic overclock
TDP isnât a fantastic measure of anything because everyone does it differently and there are a million other factors.
Generally TDP and actual power draw and heat were very close on Intel CPUs. If it says 91W it generally was. However Skylake-X is not that. It pulls way more than the TDP
Thereâs a little chatter around that itâs not actually a proper boost, but an overclock since the all-core boost leaked so far is nowhere near 4.2GHz.
Canât imagine there will be too many reviewers getting samples of these either.
@DerKrieger was saying there is some bios setting that enables the boost on all cores so could tech not be OC but those benchmarks with 18/36 dont take very long so probably wont be sustainable in a real stable OC without a very solid cooling solution
