• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

FEATURED AMD ZEN Discussion (Previous Rumor Thread)

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
It looks like M$ is sticking to their W10 guns with regard to hardware. Want Zen? You'll need Windows 10. They are extending support to 2020 for 7/8/8.1 on Skylake and FX but new processors will have no support for older OS' from M$. Zen is Ten or nothing. There is still Linux, maybe Steam OS will get some traction, but for the vast majority W10 is it. http://www.forbes.com/sites/antonyl...7-and-8-1-intel-skylake-support/#72f3be407781

So, no HWBot for Zen, lol!
 
I've been building PCs since 1997 when a 16-year old kid named Anand Lal Shimpi started reviewing motherboards on line. In the early years I built AMD, Intel and even Cyrix setups and enjoyed every minute of it. Never a fan boy, I tried everything in the single and dual socket market. Up to 2003, It was even Steven between AMD and Intel and then with sockets 754, 939 and 940 AMD shot ahead and became my favorite platform. I built mainly dual socket 940 setups since single socket meant single core in those days. That all changed in mid 2006 when core 2 socket 775 and dual socket 771 came out. For the next few years I built a few AMD rigs out of used parts for fun but nothing since 2009.

So, I really want Zen to be a success to make it interesting and know for sure I will build one, but when comparisons with Haswell came out I was and still am disappointed. It needs to be superior just like when socket 754 came out otherwise AMD remains a niche market player. Like Barcelona in late 2007, there is a lot of hype but hype isn't truth and Zen can only be even with Haswell if it clocks much higher than the current 3 GHz range. Intel will be selling Kaby Lake when Zen comes out, and Haswell will be it's ancient ancestor. My experience tells me Zen is too little too late. It is what AMD needed several years ago and will not significantly alter the
PC or server marketplace in my opinion.
 
So, no HWBot for Zen, lol!

What I took from the article is not that the older operating systems will not work with the newer hardware it just will not be officially supported. I am sure we will be able to make older versions work on newer platforms just no updates or support. Think about Windows XP on a Skylake build.
 
Why would anyone pick second-best when the overall costs aren't that significant to have the better option available to them.

Assuming Zen is second best (and that second best doesn't mean insanely overpriced garbage), I would - and most likely will - still buy it for the simple reason that it's not Intel. I want to support the underdog, I want to help maintain at least some form of competition in the CPU market and so far I've liked AMD's approach to CPU design more, even though there's no denying Intel is getting better results. I know, the majority of potential customers only care about price/performance, so I do think that AMD needs to really beat Intel in that metric if they want to increase their market share. In order to get my money, though, they only need to make a decent CPU.
 
That's why I stuck with AMD for so long. My last AMD rig went in a new case and is getting a SSD this week. If the graphics card I want gets back in stock my RX 480 will go in it and the new owner will be a happy camper. I finally had the funds to build a new PC and just couldn't justify buying another FX. I agree AMD is good for the market, and I have a fondness for "the little guy" in life, but for once I wanted good, not good enough. I'm already wondering what's after Zen.
 
I just want to De-lid ZEN. Pretty much where my interest stands at this point. When I buy, it may be one of the first things I do with it :p
 
Something I haven't considered is what if Zen clocks like the early FX octocores? So it keeps up with Haswell at 3.5 GHz. What if they happily OC to 5500 MHz or more? So many questions. Patience may be a virtue, but I'm not feeling particularly virtuous. LOL
 
http://hexus.net/tech/news/cpu/95914-amd-provides-first-glance-zen-summit-ridge-performance/

For its comparative performance demo AMD ran an 8-core, 16-thread Summit Ridge desktop processor (engineering sample) against a "similarly configured 8-core, 16-thread Intel Core i7 Broadwell-E processor" launched in June. To provide a level playing field, both system CPUs ran at 3.0GHz. The two machines ran multi-threaded Blender rendering software to complete a specific task. The result? AMD's Zen-based processor outperformed Intel's chip. No chart or detail about the margin of the win has been shared by AMD at the time of writing.

What type of instruction mix does Blender use?

http://www.pcper.com/reviews/Processors/AMD-Zen-Architecture-and-Performance-Preview

Even more info here.

The point of the demonstration was to showcase the IPC improvements of Zen and it worked: the render completed on the Zen platform a second or two faster than it did on the Intel Broadwell-E system.

If only they indicated even approximately how long the overall run was, so we can estimate a % difference.

The IPC of Zen is on-par or better than Broadwell.

At that task. I tried to research Blender, and it doesn't seem to use FMA which is still speculated to be a weakness for Zen. Blender even seems to gain from HT. This still has many unknowns, like ram or cache dependence.

we were left with a vague but understandable remark of “comparable TDPs to Broadwell-E”

At least that is something...
 
Last edited by a moderator:
Great posts/links mack! That is good news if it all works out! Maybe I will get to play with an AMD again!
 
I know Blender will utilize AVX2.

Edit: These bug fixes suggest FMA is supported.
https://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.74/Bug_Fixes

I will see if I can run some Blender benchmarks of my own at some point. My earlier statements were based on looking up benchmarking done by others elsewhere, which suggested SB was ball park 20% lower IPC than Skylake. This is not unreasonable for non-FMA workloads, and much higher would be expected for FMA. It may be that some instructions are only used on some features. For my CPU needs I am biased towards high FMA performance so that is of particular interest to me.
 
AMD have released a video on youtube, which shows the speed demo in progress. I don't have accurate time measurement capability at the moment, but it shows the start of the test. After 3 or 4 seconds, they overlay x4 speed, and in total about 18 or 19 seconds pass from start to finish. There is a gap of less than a second between them finishing. Given the inaccuracies in my time measurement, I wont try to give an exact difference, but it is sufficient to estimate a total run time of the magnitude of a minute or so. If we take the earlier comment of "a second or two" relative to a minute, we're looking at low single digit % differences.
 
Tha IS promising, can't wait to get my hands on one. Here's the Vid

 
Again a very rough timing, that looked about 47 seconds overall and not that big a gap between them, I'd estimate a second or so, still in the low single digit %.
 
Just the fact that it's keeping up with the broadwell is good news and it's still in the engineering stages. They might be able to wring out a bit more performance per clock yet. Just needs a few more things be priced around $300 CDN or less, OC to 5.0 on ambient cooling and take a full pot for serious benching.( unlike the 7870k )
 
8 cores is fine, but it will be interesting to see how 4 cores stack up as it will be going against Skylake or Kaby Lake by then, tougher opponents than Broadwell.

To try and put things in context, I looked up earlier testing I did using Cinebench, which I guess could be argued a similar category of application. I have no evidence to say they will behave similarly, so Blender may well be totally different. With that disclaimer, I found IPC of Skylake was 7.4% faster than Broadwell, which was in turn 3.9% faster than Haswell, which was 8.5% faster than Sandy Bridge. I don't have Ivy Bridge to fill in the gap there. Benefit from having HT compared to not was somewhat variable but in the ball park of 30%.
 
Back