New Zen leaks?

amd has created some hype. not as much as anything nvidia has done, and not as much as they did with other products like the 480, and not as much as the fans have done. but they are creating some hype.

@kewldude007 what i predict is AMD will sell a piece for 300 and the distributor will probably add their own margin of 50 or 100 more. So it may sell for 400 in total.

2 Likes

From what I'm hearing Zen is going to be a buggy mess, AMD is going to have a hard time selling it IMO. Prep for quick price drops after launch.

Is this bad?

very since the overclocks could make unstable to the point where the os cant boot.

+Windows only support maybe?

Driver based overclocking should be able to be used in linux as well, I would imagine. I don't think they would leave it out, but I've been wrong before.

im mostly worries about it being windows only and/or controlled by the OS

It is the driver, not the OS. This could be a very good thing, making it really easy to dial in overclocks. Or it could be a really bad thing, making things unstable and buggy. I have no idea. I don't know enough about it to have a good idea and it isn't out yet, so who knows? For now, it is just an interesting thing to note and look into once it is actually available.

1 Like

Could be, but I don't know what to think. Keep in mind this not official...

But there was rumor long before that zen cpu's would be soc what was strange to me (not strange for zen apu, because they must incorporate hbm somehow). So maybe it is logical for them that bios goes on cpu because of various stuff that they must support on AM4 platform (zen apu soc with hbm etc.). Motherboards would maybe be mess otherwise... And OS overclocking could just be consequence of that (they don't wan't user to mess on cpu itself).

I would like to hear what Wendell has to say on this...

IMO 480 was hardly hype. Hype was going on in peoples head (fury level of perfomance, 1500-1600mhz overclocks was Chinese "leaks" etc.). All amd said was minimum VR performance (~R9 290 level), better crossfire than gtx 1080 in AotS what happend to be true... Nothing spectacular

Driver overclocking, while a nice feature, defiantly shouldn't be the only method of overclocking. That being said, I'd be interested to see how motherboard manufacturers would try to promote their boards when the UEFI suddenly becomes a lot less useful.

honestly i'd always assumed about a 5960x, being cant beat a 3770k stock vs stock 1v1 thread, or oc vs oc most all of the time, but then 16 threads(assuming similar 16v16 performance, but single threaded is hard to extrapolate currently given only solid info is the blender test which assuming isnt tampered with, only tells about 16v16, if amd has better smt but shittier single thread performance could be same result and just hasn't been any info leaked at all about the smt that i've seen)

but for price i guessed entry level 2011 prices for atleast the lower end 8core(ala 8120/8320 to the 8320/8350 which could be higher priced) as if we go back to pile driver, when 8350 is 30~50% above 3570k which costs more at the time(about 230-45 over the 200) in 8vs4 threads like 7zip handbrake, gaming with fraps/streaming etc. no one bought it. so by going 3-450$ only competition is lower end 6cores.

also assuming the block diagram leaks are accurate, id assume they could sell at 8350 prices and still make money assuming similar yields, which probably not as good yields so more like 8150 prices which was the 300~ at the more lowest end prices for the top chip they could still make enough.

being 6 pipes for the int vs 8 in a module, fpu is a bit different, different front end, but mostly the massive 2mbl2 down to 512 in zen, would assume a zen core could be slightly to somewhat smaller than a module, and going from 32>14/16 if they could get twice as many transistors per die probably a similar silicon cost at best given same yields to the 32soi 8/9xxx parts. with a lower platform cost though, being soc, and 970/990fx being a pretty expensive platform itself, compared to say fm2+, or am1 which is a soc, but thats lower power stuff so cant really compare.

the hype usually isnt amd, ala bulldozer, piledriver. polaris seems about as advertised.

for soc: as far as i remember thats a function thing, to allow the platform to be more universal, being say a low end apu wont need as many pci lanes as the higher end cpu, reducing platform cost/size probably more efficient, potentially allowing longer board lifespan being pcie2 and 3 have been electrically compatible if you replace the pcie controller by getting a new processor you could gain access to pcie4 assuming that was also electrically compatible since, the motherboard would just be traces to the socket and a slot(in theory) not sure though since seems to be supplemental lanes in the 'chipset'

the bios thing not sure, might just be to keep up with the theme of soc, or could be part of the trusted platform stuff such as: http://amd-dev.wpengine.netdna-cdn.com/wordpress/media/2013/12/AMD_Memory_Encryption_Whitepaper_v7-Public.pdf but not sure really lol.

least as far as i can extrapolate.(sorry terrible grammar/wall of text as usual :/)

I hope there competitive. I will buy on price to performance not who is fastest. I cant afford nor need the fastest.

At worst it might make intel parts cheaper and I will buy an intel system. I am leaning to helping out the underdog however.

Those OC clocks are promising. And if it's anything like the Polaris launch, the first batch of chips aren't that well binned, that means we'll see the good bins later on the year of 2017, just like we're seeing Polaris GPUs now in the 80-90+ watts at stocks and below 150watts with an easy 1400mHz overclock.

It's probably why AMD is taking so long with their laptop versions. They've been waiting for the good yields.

@Fouquin This topic has been sitting in my head for a while, one of the reasons why I haven't pulled the trigger on a 6800k system.

http://techfrag.com/2016/05/24/amd-zen-summit-ridge-twice-fast-fx-8350-par-intel-i7-5960x-extreme/#blog

I know this was floating around a while ago and I would like to know how realistic it is.
It sucks needing the persormance but not having $2,700 to drop on a 6950X. Also would like to see skylake-E chips, Broadwell overclocks so poorly and no Broadwell-E chips will provide the single threaded performance I need :\

I could pick up a 5960X for roughly $1,200 But that's $500 more than a 6800k while still running Haswell

Zen doesn't have to blow our minds to be a success. It just has to be a viable alternative to Intel at a lower price point. Even if it's not a significantly lower price point than Intel, if you can get an equivalent of an i3, i5 and i7 or Xeon for $50 less than an Intel chip it will be a success.

1 Like

Everything is deleted now. That forum now needs registration. OK, I registered yesterday but it doesn't open (deleted)

Your user account may not have sufficient privileges to access this page. Are you trying to edit someone else's post, access administrative features or some other privileged system?
If you are trying to post, the administrator may have disabled your account, or it may be awaiting activation.

What is funny all posts on overclock.net in Zen topic is deleted also after someone posted that link. I'm not sure what that means. I was told that last post from source was "angry" because it get out (via some youtuber RedGamingTech allegedly), and he will have to do "a lot of explaining". WTF, what did he think that would happen... It's f-ing internet

If someone didn't copy his posts somewhere everything that has left is my summary in opening post. I can find only this quotes from him

About error

  • There are some errata issues present in the current testing samples, similar in a way to the TLB bug of the Phenom. The workaround right now is done via the BIOS. The workaround however, strips around 30 ~ 40% of the CPU performance.

And that benches are "with the error fix disabled"

Also this

AMD essentially has no Chipset for the CPUs. It's an SOC of sorts so everything is just about contained within the CPU or at least the CPU package. That does simplify the board in some ways, but the worst part is that it takes away the small differences boards had between them. Right now there's literally no difference at all between any board (all of which are far too early to be relevant). Essentially all control is in AMD's hands and in the user hands for overdrive.

You can still access traditional BIOS for SATA config, Audio, boot sequence etc but that's it really. Even DRAM frequency you'll have to set inside the OS and reboot as I don't think you can change multi once the mem training is done. We will have to see.

Again there are still way too many bugs to iron out, but the performance is there so they must work to iron out the rest of it. There's too much of a spread right now between CPUs as I said
4.1GHz vs 5.2GHz on some. But progress is being made rapidly so we will see. Overclocking is only on the OC SKUs, but as we have seen from AMD, that's almost all of them.

That's all that has left...

Well i personaly think thats highly unlikely to happen.
$700,- would sound more likely indeed.

Chris Hemmelgarn of Barclays :
“With Summit Ridge launching in Q1 of 2017, I guess how would you expect the channel to ramp that? Do you see it ramping pretty fully in the first couple of quarters of the year, or are you looking for more normal PC seasonality?”

President & CEO Dr. Lisa Su :
“You know, I would expect that there will be a relatively good initial demand for Summit Ridge that may be you know, not quite at the seasonal patterns. From where we see, Summit Ridge is playing in a space in the high-end desktop that we currently aren’t offering a product. So we believe we’ll be competitive certainly with Core I5 as well as Core I7 and we will be launching in those areas.”

Like I said earlier (regardless of leak) maybe not 300$ for 8c16t, but up to 400-450$ max (i7 6800K price range max). It will be as usual - lower singlethread, but competing with more core/thread (with heavily stronger multithread) than intel counterpart

Video about this

Yeah, a whopping 300 MHz, out of 3300 MHz. What a world of difference, against an Engineering Sample that isn't finished!

Don't be dense, that was a comparison for Instructions Per Clock, if AMD archives similar clock speeds to their past CPUs, the 8C/16T Zen CPU will eat the i7 6900K.