960 Pro specs released

Still reading that 950 Pro thread, on and off.

The main thing is, so many users don't research at all what they need to do to use a 950 Pro. I  assume everyone is so used to plug and play, that they don't think they need to do anything different than a SATA SSD as the OS drive. While what needs to be done is not hard, it is different.

Plus add the requirements of the mother board, which is one of the main confusions, and it becomes a mess for some users. Then there are posts about it working as an OS drive on old boards with BIOS firmware, which is strange.

The Samsung Magician benchmark test is Samsung's own worst enemy. It won't match the performance specs for some aspects because it is designed or configured wrong, and users think their 950 Pro is bad. Of course, we all use 300,000 random read IOPs for our usual PC work all the time, right?   No, not at all, but I understand why users would wonder about it. I never use Samsung's benchmark anyway, it's too simple and why should I trust it? Unless they finally fixed it.

Samsung actually has a good guide about using the 950 Pro, but it's hidden by the document name, SSD 950 PRO White paper. Also, 47 pages long. It confirms everything I've preached about UEFI booting, although I've learned/confirmed recently that is not the only way to get an OS installed on an NVMe drive. The document does leave out a few basic details that might cause problems for some users, but is better than nothing. I think I see a mistake too, in the Boot Order selection.

In the section about installing an OS on a 950 Pro, guess which mother board's UEFI they use as an example?

The ASRock Z97 Extreme6. See page 20. Gives you an idea how old this document is:

http://www.samsung.com/semiconductor/minisite/ssd/downloads/document/Samsung_SSD_950_PRO_White_paper.pdf

At one time, the Z97 Extreme6 was the only board that could provide the PCIe 3.0 x4 M.2 interface the 950 Pro requires for full performance. This document only lists 23 mother boards as compatible with a 950 Pro, which is not up to date at all. That's a big help for people that actually read it.

I highly suggest at least reading parts of this document if you use or plan to use a 950 Pro. It may not be perfect, but it can be very helpful. At least parts of it should be included as instructions with the 950 Pro... not that we need to read any instructions, right?

DooRules wrote: Haswell E can generally o/c better than Broadwell E 6950x but you do not need as high an o/c on 6950x to beat out Haswell. Raja is one seriously knowledgeable dude. Anyone who chooses not to listen to him on an Asus mobo should be shot.  On my 4.571 o/c I have baseclock moved to 102.3, uncore is at 3850 and I have the V for uncore at 1.35, max safe voltage. Hey parsec I have said many times in terms of Vc it is not the voltage but the cooling that is most important. People that still choose to run prime95 and hammer their cpu for hours on end deserve what they get. Seriously, it proves nothing other than you could run prime at that point in time. You could just as easily fail the run the next day. Everyone has a different opinion of what "stable" is. I personally could care less about any of these synthetic tests that hammer the cpu to try and prove you are 'Stable' .

I agree with everything you said. You also must know that your results are better than most if not all of the others in the OCN Broadwell-E thread. Which is also strictly the territory of the manufacture of the X99 board you are using.

I some ways, that thread reminds me of the OCN 950 Pro SSD owners thread. Except the 950 Pro thread is ridiculous, I'm about half way through it and I cannot believe much of what I see. I guess it's just me, but really, asking which IRST driver to use with a 950 Pro? I'm not talking about RAID either. The problems some of those people have, I thought I've seen it all until the next post I read in that thread. Such as, thinking every PC platform will "boot" at the same speed? Ever hear of POST guys?

Plus Samsung has been ZERO help. Not to mention the mistakes in the Magician software's display of the 950 connection interface.

 

Edited by DooRules - 22 Sep 2016 at 5:59pm

Wow, you tried some wildly high VCore settings, given what I have read. The dead CPUs were killed at a bit over 1.3V, but there is more to it than that. That was 6950X CPUs.

The apparent killer was running Prime95 with AVX instructions, that supposedly causes super high current draw/flow through the CPU. I believe that was using Adaptive voltage. This is what the Asus rep, whose name is Raja, is saying in the OCN Broadwell-E thread.

Otherwise, that is one of the highest Broadwell-E processor OCs I've heard of. Nice.

I wanted to see how far you got the Broadwell-E Cache to OC. Your result is among the highest I've seen others achieve. An ASRock X99 board user was complaining in this forum about why he was able to get a better OC using a Haswell-E CPU, rather than a Broadwell-E. I tried to explain that Broadwell-E is usually limited to an OC of 4.3GHz on the cores, and below 4GHz for the cache.

He expected the new CPU generation to do better than the previous one, which is usually the case for an Intel "tock" update, in their "tick" - "tock" progression. Unfortunately, Broadwell-E is a "tick" update, from 22nm to 14nm. Plus it retained the Haswell integrated CPU VRM.

Skylake it a "tock" update, from mainstream Broadwell processors, that we know are not designed for over clocking. The only thing Skylake shares with Broadwell is the lithography size, 14nm.

Broadwell-E is better than I thought it might be, given Broadwell mainstream processors. A Broadwell-E core has better IPC than Haswell-E, but simply looking at the OC numbers, Haswell-E can do "better".

An excellent indepth discussion of the announced 960 Pro and 960 Evo

https://www.pcper.com/news/Storage/Samsung-960-PRO-and-960-EVO-Announced-Details-and-Specifications-Inside