AMD Zen Rumours Point to Earlier Than Expected Release

Hell the FX-9590 gets over 2500 for 2-bit and 12k for multi core with only 8 cores/threads. Scaled for CPU frequency of 5GHz on the 9590 and 2.5GHz on the eng sample the 9590 would be 1250 at 2.5GHz so clock per clock the zen would only be 20% faster. Pretty sure we should get a lot more IPC than that out of this whole thing...
 
Gideon said:
Its the only way you can build and sell things in China, you have to give the Chinese 50% of the company AMD will retain the other 50% a joint venture. Intel wont challenge it, Id put money on it.
Click to expand...

Intel has a similar joint venture with Rockchip in China.

Pieter3dnow said:
Something funny ;)
Not to sure about it but enjoy

AMD Eng Sample, ZS258045TGG54_34/25/20_2/16 @ 2.50 GHz
1 processor, 16 cores
Click to expand...

AuthenticAMD Family 21 Model 2 Stepping 0

Family 21 = Bulldozer, Zen is 23
Model 2 = Piledriver

So.... an engineering sample of a piledriver Opteron.
 
Quix said:
AuthenticAMD Family 21 Model 2 Stepping 0
Family 21 = Bulldozer, Zen is 23
Model 2 = Piledriver
So.... an engineering sample of a piledriver Opteron.
Click to expand...

Yeah I said it was funny ;)
There are to many people wasting their time hunting for these "gems" on the internet such a waste of time....
 
Spirit_Retro said:
Well, I'm not confident in your legal analysis, but I'll skip listing my differences since, unless it hits a lawyers office- it's all speculation.

However, you are right about doing business with China- it is problematic. If it were only a cheap labor issue we would not be having a conversation at all. Combine cheap labor with disrespect for intellectual property- well then you have a problem.

I'm tangentially involved in a business where the main manufacturing of the mass produced example of this product is made. Every innovation of the product has come out of crowd-sourcing (there are still patents involved). Mostly from the US/Canada/EU/Russia.

It takes about 6 months for the Chinese to copy any functional revision of the product- and return it to western markets at lower cost and seriously compromised quality. Where patents have been involved- they have simply ignored them.

So it's my opinion that moving any technology operation out of the west, specifically to China, is a huge mistake. Take it to Taiwan, Thailand, South Korea- but never China.

As an aside, I was involved with Apple computer in the 1990's, and was quite proud that when they began offshoring Ireland was the destination- a country that had cheap labor and IP protections.

These days- companies will eventually slit their own throats by doing business in China. They'll do the manufacturing, but the innovations will be stolen and and shot back into the western market. It's a bad deal.
Click to expand...


Well according to this its only packaging that the Chinese company will be doing. Anycase the reason they got money is because they sold part of their company to the Chinese firm.

AMD Announces Joint Venture with NFME | PC Perspective

If AMD is outsourcing this they might be laying more people off.
 
Last edited:
razor1 said:
Well according to this its only packaging that the Chinese company will be doing. Anycase the reason they got money is because they sold part of their company to the Chinese firm.

AMD Announces Joint Venture with NFME | PC Perspective

If AMD is outsourcing this they might be laying more people off.
Click to expand...

"Only"?

"Second, NFME now has access to some interesting packaging and testing technologies. NFME's website claims that this allows them to handle dies up to 800mm2, substrates with up to 18 layers, and package sizes up to 75mm. These specifications sound like it pulls from their GPU experience, which could bring all of that effort and knowledge to completely different fields."
 
Yeah Highly Highly unlikely (in complex designs, for GPU's or CPU's, yeah for other simplistic things they can, which many other foundries can do the same), the biggest chip Intel has made so far has been the latest Phi, knights landing, and that was 675mm2, we know Intel is much more capable then any other foundries when it comes to today's processes. Also package size of 75 is fairly common even the CPU world. And for any other foundries or in this packaging companies to say they can do more, seeing is believing, cause we haven't seen any foundries come close to what Intel is doing.
 
Last edited:
pxc said:
The CEO made statements about various products and the JV before the questions started. The first part of what I posted above was before the questions started. Absolutely zilch about progress on the desktop version, unlike the December investor conference call (where it hadn't taped out yet and was called a 2017 product).

BTW, when AMD's actual statements contract what inaccurate rumor sites report, it's not a misunderstanding of what was said in the conference call. The problem is not accepting a less than rose colored glasses view of reality, where Zen ain't coming out this year.
Click to expand...
Zen comes to desktops first

At the Q and A session after AMD's press conference, company CEO Lisa Su just confirmed that desktop Zen x86 chips are the first products to arrive for consumers. These will be followed by a server lineup, then a notebook lineup.
Click to expand...

Again .....
 
Looking forward to a system re-build at the end of the year I am.

Reaching for a moment of Zen. :)
 
Quix said:
That's actually a delay, Zen was scheduled for desktops in Q4 2016 with server chips coming before that. Hopefully they do get the chips out before the end of the year, it's becoming a do or die situation.
Click to expand...

Server was expected for Q4 2016 and desktop sometime in 2017. They moved up desktop and delayed server.
 
Was the server Q4, desktop 2017 time line ever official or was it just rumor? Server things usually require more tests to validate which tends to push them back. Desktop chips coming out first seems more natural to me.
 
  • Like
Reactions: N4CR
like this
My buddy is looking to upgrade but he's running AMD right now. We were talking about potential upgrades but I'm curious has AMD put out anything official in terms of release dates for Zen or is it all just speculation at this point?
 
Zen desktop 8 core / 16 threaded CPU only on AM4 - Q4 2016
Zen server / workstation ( up to 32 core with possible 8 channel ram ) on some other socket with enough pins (3000+?) to support additional ram channels and multiple cpus- 2017
Zen APUs on AM4 - 2017
Zen desktop 4 and 6 core on AM4 - 2017
 
Last edited:
Darakian said:
Was the server Q4, desktop 2017 time line ever official or was it just rumor? Server things usually require more tests to validate which tends to push them back. Desktop chips coming out first seems more natural to me.
Click to expand...

That was never the official line, in fact it was previously (before what I wrote in my last post) meant to release in late 2015.

SpongeBob said:
My buddy is looking to upgrade but he's running AMD right now. We were talking about potential upgrades but I'm curious has AMD put out anything official in terms of release dates for Zen or is it all just speculation at this point?
Click to expand...

Q4 2016, from their last press conference.
 
^ That's not what AMD said. It was thrown in the last few minutes of the Computex presentation and no release date was mentioned. What the CEO said during the press conference, not with any date mentioned at all, is...

"Zen is very, very special project and product for AMD," President and CEO Lisa Su said. "We are in the early stages of bring-up* but the product looks really good."
Click to expand...

Jargon translation: "We are in the early stages of [validation and testing] but the product looks really good."

Sampling is coming in the 3rd quarter, according to Su. It's very, very unlikely that it will come out this year. As a comparison, Kaby Lake is a processor likely to come out in Q4'16, and there were several demo systems on the floor at Computex. Zen is nowhere near the same development stage as Kaby Lake.

* yet again another confirmation that various rumor web sites have been spewing incorrect information. Zen, as of 3 days ago, was still in early testing and validation stages, not something ready for system launch in a couple of months.
 
Intel is also a larger company with far more resources and kaby lake is a smaller change from its' predecessor. If we assume AMD won't delay from what was stated at computex then OEMs will have samples in Q3 and will start launching products some time after that. If AM4 boards are around by then I don't see why AMD wouldn't also launch stand alone chips around the same time as the OEMs start launching products which might be this year.

I'm curious about where AM4 is. I know it was rumor, but I was expecting to see AM4 boards at computex.
 
Darakian said:
Intel is also a larger company with far more resources and kaby lake is a smaller change from its' predecessor. If we assume AMD won't delay from what was stated at computex then OEMs will have samples in Q3 and will start launching products some time after that. If AM4 boards are around by then I don't see why AMD wouldn't also launch stand alone chips around the same time as the OEMs start launching products which might be this year.

I'm curious about where AM4 is. I know it was rumor, but I was expecting to see AM4 boards at computex.
Click to expand...
They stated that we should have seen in May this year for Bristol Ridge(?) before Zen. But I never saw anything on it.
 
Darakian said:
Intel is also a larger company with far more resources and kaby lake is a smaller change from its' predecessor. If we assume AMD won't delay from what was stated at computex then OEMs will have samples in Q3 and will start launching products some time after that.
Click to expand...
A processor does not go from early testing and validation 4 days ago to retail release in a few months. That's a many month process if AMD finds no big problems*. Given AMD's history with launching everything late, I'd put that at closer to a year at least. At this point, based on what Lisa Su said at Computex of the state Summit Ridge is at, it is a 2017 product. Q1'17 would be a huge miracle, Q2'17 is slightly possible and Q3'17 is fairly reasonable.

The reason I compared Kaby Lake's appearance at Computex to the state of Zen/Summit Ridge is that Kaby Lake is coming out in the Q4 time frame. It is much further along in validation and testing and is still several months away. If anything, this shows the flaw of clinging on to Q3'16 Zen predictions made back in 2014. AMD has a much longer testing and validation process than Intel did with Kaby Lake, yet Summit Ridge is somehow supposed to come out the same time or quicker than Kaby Lake? That Summit Ridge was not even shown running at Computex is a huge blow to the belief that it will launch this year. Let me explain why in this handy footnote:

* AMD's last new uarch, Bulldozer, was first shown **running** in November 2010. It was released in October 2011. Barcelona was first shown **running** in November 2006. It was released in September 2007. AMD first showed Hammer **running** in February 2002. It was released in June 2003.

Just to beat the point home, Summit Ridge isn't even at the point where it can be shown running an OS. Lisa Su explained Summit Ridge was still early in "bring-up", which is validation and testing.
 
GFXBench - Unified cross-platform 3D graphics benchmark database Carizzo-based AM4 (see bottom)?

AMD has not said Zen is coming in Q4, and not showing a working part suggests it's over a year away from release going by history as I showed above. Sampling in Q3 doesn't mean qualification samples, as in final processors. Sampling for processors covers a large range from the wobbly table leveler that Su showed at Computex to QS processors almost ready for release typically a year after initial samples.

It's been like this for over a year now. AMD has not met benchmarks which would suggest release X months away, but people still trust wild speculation from inaccurate rumors sites. The process isn't exactly transparent, but major milestones still give the best idea of what stage a processor is in. Early testing and validation without operational samples suggests Summit Ridge is a year-ish away. It's not the news people want to hear, but it's what the evidence supports.
 
Last edited:
xorbe said:
Saw this on Tom's last week, because TPU was talking about it iirc.

AMD CEO Teases Zen Processor Family, Summit Ridge Makes An Appearance

"AMD displayed a dynamic rendering of the Zen logo that was created, edited, rendered and played back with the new desktop Zen processor, code-named Summit Ridge, proving that the company has working silicon."
Click to expand...
That's interesting, but it's not at the point I listed for other prior architectures where AMD showed a running system. AMD might be getting close to that point, but it's not there yet. I'd guess it's still massively unstable and/or too slow for display. Even if AMD had brought a first live demo system last week, history shows there's an 11-15 month gap between that stage and product release. I have no doubt that AMD worked hard on preparing a live system for Computex, and it just didn't happen.
 
xorbe said:
Saw this on Tom's last week, because TPU was talking about it iirc.

AMD CEO Teases Zen Processor Family, Summit Ridge Makes An Appearance

"AMD displayed a dynamic rendering of the Zen logo that was created, edited, rendered and played back with the new desktop Zen processor, code-named Summit Ridge, proving that the company has working silicon."
Click to expand...

That quote is missing "AMD claimed" because all we have is their word there, it doesn't really prove anything. I'm not saying that they don't have working engineering samples, just that that doesn't prove anything. I hope they do have working samples otherwise we're probably a year out.
 
Quix said:
That's actually a delay, Zen was scheduled for desktops in Q4 2016 with server chips coming before that. Hopefully they do get the chips out before the end of the year, it's becoming a do or die situation.
Click to expand...

"Limiting Sampling" in Q4. Despite reports otherwise, the timetable has not changed: Consumer availability in Q1 2017. People got hyped about an early release for no reason, as I noted months ago.

EDIT

It's worth noting the ES sample was clocked at 2.4 GHz. While ES samples tend to be lower clocked, the indications have been for a while Zen will be clocked lower then BD/PD, and frankly, this was slightly worse then I was expecting. I'm now thinking about 3 GHz max at launch.

To put that in perspective, using the FX-8350 as a baseline, 3GHz would be a 25% reduction in clocks. That will mostly mitigate IPC gains (claimed 40%, which I consider "best case" until proven otherwise), leaving absolute performance gains at about 15-20%, or Ivy Bridge/Haswell level performance.

Honestly, early indications are just a little troubling.
 
gamerk2 said:
"Limiting Sampling" in Q4. Despite reports otherwise, the timetable has not changed: Consumer availability in Q1 2017. People got hyped about an early release for no reason, as I noted months ago.

EDIT

It's worth noting the ES sample was clocked at 2.4 GHz. While ES samples tend to be lower clocked, the indications have been for a while Zen will be clocked lower then BD/PD, and frankly, this was slightly worse then I was expecting. I'm now thinking about 3 GHz max at launch.

To put that in perspective, using the FX-8350 as a baseline, 3GHz would be a 25% reduction in clocks. That will mostly mitigate IPC gains (claimed 40%, which I consider "best case" until proven otherwise), leaving absolute performance gains at about 15-20%, or Ivy Bridge/Haswell level performance.

Honestly, early indications are just a little troubling.
Click to expand...

Keep in mind Base Clock verses Turbo... I always expected a 3-3.4Ghz base clock.
 
Are all of these mainstream chips coming with essentially Hyper Threading?

With the 8c/16t chips out first, I bet these will be extremely expensive.
 
iRUSH said:
Are all of these mainstream chips coming with essentially Hyper Threading?
Click to expand...

Yes. They will have significantly improved IPC and also what Intel calls HT. To achieve this level of performance I expect each core to have at least as many transistors as a bulldozer module had and similar to what Intel has.

With the 8c/16t chips out first, I bet these will be extremely expensive.
Click to expand...

Depends on what you consider expensive. I expect the 8c/16t Zen to compete with the lowest end 6 core / 12 threaded lga2011-3 CPU which is the i7-5820K which is currently priced ~$389. AMD will have a discount but may not be that big considering the lga2011-3 motherboards will cost more than the AM4 boards.
 
drescherjm said:
Yes. They will have significantly improved IPC and also what Intel calls HT. To achieve this level of performance I expect each core to have at least as many transistors as a bulldozer module had and similar to what Intel has.
Depends on what you consider expensive. I expect the 8c/16t Zen to compete with the lowest end 6 core / 12 threaded lga2011-3 CPU which is the i7-5820K which is currently priced ~$389. AMD will have a discount but may not be that big considering the lga2011-3 motherboards will cost more than the AM4 boards.
Click to expand...

Look at AMD price/performance if it can perform you pay more if it can't they most likely lower price ..
 
GothamsReckoning said:
I don't know where information about a low clock speed is originating from but the A12-9800 flagship is clocked at 3.8Ghz for Bristol Ridge. I doubt that Summit Ridge would see lower clocked variants for the non-APU processors. Also Zen should still be on-track for an October release for Summit Ridge including for the 8 core version. Keep in mind that Zambezei also launched in October 2011 for Bulldozer.
Click to expand...

Bristol Ridge is 28nm Summit Ridge is 14nm the process is different also the amount of transistors...
 
You must log in or register to reply here.
Back
Top