• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Fudzilla: New AMD Zen APU boasts up to 16 cores (plus Greenland GPU with HBM)

Page 41 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Eh...10-20 % isnt going to make me upgrade.

Im a cheap ass when im upgrading anyway, so i doubt il be upgrading from this Xeon in the next couple of years.

I managed to stick with my Shitdozer for 3 years before this Xeon 🙂.

Was hoping AMD would blow intel out of the water tough, they need it.

But i'm also thinking they would make a competitive chip even if they can't touch Skylake, if they keep their low pricing + almost intel performance + high number of cores that would be an attractive offer for those on a tight budget.
 
Enigmoid said:
7- zip is a terrible indication of IPC.

Why?

Because its so branchy and latency/bandwidth focused that large cores absolutely suck at it.

In most other programs the E3-1240 leads by a factor of ~3x. In 7-zip its around 2x.

http://www.anandtech.com/bench/product/836?vs=1229

7-zip is a really 'small-core friendly' test.
Click to expand...

I could have used Fritzchess but you re against this soft despite it favour Intel, using more recent chess games the scores are the same as 7 ZIP, so what s next, we should use Sysmark instead, or better, 3D Particle movement that you advocated as being relevant before it was debunked as being eveything safe a bench..??.

I guess that people are so much addicted to the Cinebench kool aid that they dont grasp that Haswell has not as much Integer IPC than in this ICC compiled bench...

If AMD can deliver thoses 40% this chip will have significantly better Integer IPC than Haswell, for FP we dont know but we can expect a doubling of a Piledriver core throughput, wich would amount to 70% "IPC", wich it isnt, it is rather throughput.
 
Last edited:
geoxile said:
If geekbench single-thread benches are any indication, Zen should be on par with Haswell overall. No idea how the improvements are distributed though.

Dropping CMT also means no more penalty on multi-threading.

This is assuming XV has a small improvement over Steamroller btw. Haswell's geekbench score is about 1.5x that of Steamroller (calculated using 4670k vs 7850k, single thread scores from HWbot normalized for frequency)
Click to expand...

Also consider the L3 cacheless tax the 7850k pays being an APU and all.


All in all the 40% figure seems realistic. The moment we will be able to see XV benches we will be able to tell just how much this stacks against AMD's last effort on the HEDT space (8xxx and 9xxx products).
 
mikk said:
You are wrong. Check the slides. Zen is CPU only. APU is Carrizo or a Carrizo refresh apparently.
Click to expand...


You re right, slides should be checked before posting...

amddesktoproadmap2016.jpg



I think this is clear.
 
IPC claims on marketing slides are problematic because they could be cherry picked. We don't know what benchmarks or estimate this is based on. Especially because Zen is probably more than a year off from release.
 
frozentundra123456 said:
That roadmap looks incomplete. If it is accurate, it seems like a disaster for AMD. If all they have from Zen is an FX without an igp, how much market can there be for that?

Does that mean the APUs are still on steamroller, not Zen, and no HBM? That is what the slide implies, since they specifically mention Zen for FX and DDR3/DDR4 for the APUs.
Click to expand...

ShintaiDK said:
K12 also delayed to 2017. After the Skybridge collapse I wonder if its just waiting to be axed.
Click to expand...

Welcome to AT forums where people first read all the information before jumping to conclusions.

"Meanwhile AMD has confirmed that Zen will be shipping in 2016, and that it will be produced on a yet-to-be-named FinFET process. Our bet would be that AMD continues to use traditional partner (and spin-off fab) GlobalFoundries, who will be ramping up their 14nm equipment for next year as part of their licensing/partnership with Samsung to implement Samsung’s 14nm FinFET process. Zen at this time is AMD’s priority, to the point where the company is willing to push back the ARM K12 in order to get Zen out the door first.

Desktop users will be happy to know that the first Zen processor out the door will be AMD’s high-end desktop CPU (AMD was very deliberate in this, it’s not an APU). AMD will be aiming high and then cascading Zen down into APUs and lower-end products.

Said Zen CPU will use a new AMD platform – AM4 – which will also support DDR4. Unlike the Dozer family, all of AMD’s desktop CPUs will use the same AM4 platform/socket, so when AMD does ramp up their Zen APU, it too will be on AM4 and not its own socket, simplifying the process."
http://www.anandtech.com/show/9231/amds-20162017-x86-roadmap-zen-is-in
 
Looks like it'll be at least SB/IB performance level then, and possibly Haswell territory depending on type of workload. Very nice! 😎

The 8 core Zen will by a very significant margin beat anything Intel has in the mainstream CPU segment in MT performance in 2016. And it'll come close in ST performance. You'll have to get a $1000 Haswell-E to beat it in MT performance!

I think this will be the top consumer CPU to get in 2016 for those requiring high performance, such as gamers and computer enthusiasts.

Zen will most probably also get a nice share of the x86 server market, and the high margins that come with that.
 
This is what everyone wanted, right? Drop the construction cores, drop(ish) the APU, focus on high end x86 CPU and GPU desktop performance parts on a new process node. I guess not as good as closing up shop and donating all the money to intel so intel could drive the industry forward a little more but this is probably the 2nd best thing.
 
According to Q&A the 40% IPC increase is purely from the architecture change and does not take into account the shift to the FinFET process.
 
mikk said:
IPC claims on marketing slides are problematic because they could be cherry picked. We don't know what benchmarks or estimate this is based on. Especially because Zen is probably more than a year off from release.
Click to expand...

Of course they are suspect of being biaised.

As much as your prejudices.?.

Because it s prejudice to say that one may not be honnest without having the slightest evidence about it.

They use 7ZIP + Winrar, Blender, CB R15, Povray, X264, TrueCrypt and so on, that s what they displayed in their Kaveri reviewer guide, much better than their competitor s Sysmarks or other inhouse benches a la WebXPRT or ultracompiled Antutu...

Noctifer616 said:
According to Q&A the 40% IPC increase is purely from the architecture change and does not take into account the shift to the FinFET process.
Click to expand...

IPC improvements can only be done with uarch enhancements, node shrinks only bring density and perf/Watt improvement through higher frequencies at same power.
 
Last edited:
mikk said:
IPC claims on marketing slides are problematic because they could be cherry picked. We don't know what benchmarks or estimate this is based on. Especially because Zen is probably more than a year off from release.
Click to expand...

I don't find the %40 claim to be unreasonable however. Because of just how bad they screwed up with Bulldozer.
 
erunion said:
That doesn't seem to match the leaked roadmap.
The leaked roadmap was just about the best case scenario for AMD, with pretty much the entire current product stack retained and updated. So I'd consider this bad news.
Click to expand...

Said '"leak" has been officialy stated as being fake slides, i think we can stop with this non sense that is commenting junk info.

http://www.planet3dnow.de/cms/14941-erklaerung-von-amd-zu-den-zen-folien/

erunion said:
The lack of the Zen on the APU is very suspicious. It suggests that the rumors of a Carrizo refresh in 2016 were true!
Click to expand...

It suggest that you have trouble reading, Carrizo with Excavator is 6th gen in the slide, the replacment is 7th gen.

mikk said:
Zen for APUs is coming 2017 I've read.
Click to expand...


FUD unless you can point us where you did read this..
 
Last edited:
Abwx said:
It suggest that you have trouble reading, Carrizo with Excavator is 6th gen in the slide, the replacment is 7th gen.
Click to expand...

Nope. You're wrong as always.

The rumor was that Bristol Ridge would be a Carrizo refresh in 2016.

http://wccftech.com/amd-bristol-ridge-apu-2016/

Edit to address your edit:
Yes, the previously leaked slides not matching the real roadmap means they were fake. And like I said, its bad news. Because the fake roadmap represented a much stronger lineup than the real roadmap.
 
Last edited:
I think that the research papers posted earlier reflect more of AMD's direction with GPGPU development going forward. The paper that concerns a more flexible scalar logic in the compute unit sounds like they are bringing it inline with the ACE function, aswell as improving scalar performance in each CU.

The other paper on channel elements seems to be about making the architecture more friendly towards less finely structured compute kernels.
From what I have read about Blender Cycles, alot of their performance optimisation has been focused in more fine grained workloads, which I imagine takes significantly more time to program for. The work in this paper may lead to GPGPU programming on AMD GPU's being more tolerant to compute kernels that are not as finely optimised.
 
AM4? Not FM3? Hmm! I certainly hope the HSF mounting system is the same one as on AM3+. That means I get to keep my nh-d14 awhile longer! Yay!

edit: okay, so the big question is Bristol Ridge. Originally, it was said that Bristol Ridge would be 2M Excavator + 512 GCN shaders.

Then there were the more recent slides that said that Bristol Ridge would be 4C Zen + 512 GCN Shaders + HBM (maybe).

Now they are saying, "7th Gen APU"

Does that mean we are basically back to the 2C Excavator APU for Q3 2016?

Also, what does this mean for Basilisk?
 
Last edited:
erunion said:
Nope. You're wrong as always.

The rumor was that Bristol Ridge would be a Carrizo refresh in 2016.

http://wccftech.com/amd-bristol-ridge-apu-2016/
Click to expand...

Lol, who does need a rumour when we have now actual info from AMD s mouth..?..

6th gen is Excavator based and 7th gen is based on what.?.Excavator again..?.

kapulek said:
Papermaster confirmed it in Q&A.
Click to expand...

A quote and a link eventualy..?.

All i m reading in some posts are things like, "i did read", " they said', "they confirmed", but curiously no quote or link to thoses alleged statements...
 
Abwx said:
Lol, who does need a rumour when we have now actual info from AMD s mouth..?..

6th gen is Excavator based and 7th gen is based on what.?.Excavator again..?.
.
Click to expand...

But the words from AMDs mouth dont mention what core 7th gen is using. If its Zen, why didn't they say so like they did for FX?
 
What interests me most about the AMD presentations was the mentioning of a new high speed network on chip interconnect, and the high bandwidth cache system that was mentioned.
The presentations lacked details except that the interconnect was based on experience with coherent hyper transport, so presumably HT is not quite as dead as believed.
 
Abwx said:
I could have used Fritzchess but you re against this soft despite it favour Intel, using more recent chess games the scores are the same as 7 ZIP, so what s next, we should use Sysmark instead, or better, 3D Particle movement that you advocated as being relevant before it was debunked as being eveything safe a bench..??.

I guess that people are so much addicted to the Cinebench kool aid that they dont grasp that Haswell has not as much Integer IPC than in this ICC compiled bench...

If AMD can deliver thoses 40% this chip will have significantly better Integer IPC than Haswell, for FP we dont know but we can expect a doubling of a Piledriver core throughput, wich would amount to 70% "IPC", wich it isnt, it is rather throughput.
Click to expand...

I never said all this stuff and I never gave an alternative. I'm saying, pick what you want but 7-zip (and winrar and the like) are poor indicators of IPC.

Again I said everything depends on the context. Some tests may be terrible general purpose performance indicators but good in a specific context.

mikk said:
IPC claims on marketing slides are problematic because they could be cherry picked. We don't know what benchmarks or estimate this is based on. Especially because Zen is probably more than a year off from release.
Click to expand...

Most definitely.

Look at footnote 1.

Based on internal AMD estimates for “Zen” x86 CPU core compared to “Excavator” x86 CPU core.
Click to expand...

If AMD can reach 40% IPC gains that will be fantastic. But this is AMD marketing. I have no idea if I can trust it. I don't trust anyone's marketing without huge caveats. I want the goods delivered before I will start getting excited.

IMO based on the IPC disclaimer I would be extremely leery of actually believing a straight up 40% IPC improvement. AMD is using 'internal estimates' that quite possibly are not accurate. It would go a long way towards peace of mind if they were using specific tests and gave specific performance comparisons.

IMO 'based on internal estimates' is very often - "If everything is working properly you might see at most..."
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top