AMD CPUs, SoC Rumors and Speculations Temp. thread 2

Page 36 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
Samsung shows off 10nm wafers at Techcon
http://semiaccurate.com/2015/11/12/samsung-shows-off-10nm-wafers-techcon/
Coreboot Adds CC6 To Improve AMD Power Savings
http://www.phoronix.com/scan.php?page=news_item&px=AMD-Coreboot-CC6




it also depends on how hard glofo fails. late 2016 paper launch and slow rollout throughout late 2016 -2017 is quite plausible.
 


not on a console. I was super surprised they put 8 gb gddr5/3 on this one! I would expect to see 8 gb and would be surprised again to see 16gb hbm. You have to remember that consoles have better ram management and 99% of games will play on 4gb of ram on 64bit win 7. 16 just seems amazingly huge jump for consoles.

@de5_Roy its always upto how hard glofo fails...
 


The problem is that not all that 8GB is used just for VRAM. From what I have read it seems that in most cases games have 3-4GB for VRAM while the rest is used for normal system RAM.

For some reason people (not super PC techy people but somewhat techy) seem to think that 8GB of RAM was able to be allocated to just VRAM when it is not. That is how 16GB of any, but mainly HBM, would benefit consoles. They would have more to be able to allocate to system processes and VRAM allowing for higher resolution textures.

That said, if they US HBM it does not mean that AMD automatically wins. Intel and nVidia could put a better offer on the table or they might instead use HMC (which will have a higher total bandwidth anyways).

The CPU will play a much smaller role, most likely it will be whoever gives them the cheapest option.
 


HMC spec2 is what was falsely compared to first wave HBM to produce those figures.

In all reality...the difference between the 2 will be a hill of beans by the time that 3rd generation HBM comes around, we will see what the numbers look like.

As for CPU being weaker...I am not sure I follow. Sure...the GPU will be the bigger chunk of the APU, though that is already the case.

I doubt they go with less power than they currently have in the APUs now from the CPU end.

I am not super familiar with XBone, however, I can confirm that PS4 allows 6 cores full time, 7 cores if you dare for game performance, and as far as system RAM, you are looking at the capability to be able to dynamically address as much as you need. So, you could feasibly have as much as 7 GB of VRAM or as little as 3-4 GB of VRAM I would imagine...
 


Yeah I think this business of 'oh well ok the consoles got 8gb of ram but 3-4gb is reserved for system' is kinda far fetched.

Boot up Windows 10 in clean boot mode (to disable unnecessary background applications) and look how much ram it uses at desktop... I'll think you'll find it's comfortably under 1gb. I can't see the OS on either consoles using *more* ram than Windows on a desktop.

Also memory allocation on a console is a complex issue. It's true in a PC you have a separate dedicated memory pool for the gpu, but on the other hand a lot of textures and such have to be streamed from memory effectively making them duplicated in the system ram which won't be necessary with one memory pool.

I can see that many games might use about half the ram for graphics, however I would imagine most of what's left is also used for the game, just maybe for other tasks such as networking, physics, AI and such like. Irrespective, 8gb of ram should still be ample for most games, even up to 4k if they keep the texture sizes in check. Don't get me wrong, I would expect this to increase at least a bit for next generation, maybe 12gb of HBM (3 channels) would be feasible? That would actually bring the console into line with a typical PC featuring 8gb system and 4gb graphics ram (although again having the benefit of being a single unified memory pool)....
 


He said that he doesn't have more information beyond old official slides. But we have linkedin profiles from his superiors at AMD mentioning 2017 Zen products. We have the interview to Zen team chief, and the interviewer says first Zen products appear in 2017. We have Lisa Su saying "sampling in 2016".

If the release is 2016, why the interviewer said twice that first Zen products appear in 2017?
If the release is 2016, why the Zen engineer wrote 2017 in his professional profile?
If the release is 2016, why Papermaster and Su said "sampling" in 2016?

I cannot claim with 100% certainty that Zen will not be released in late 2016, but all data I have suggests an early 2017 release.
 


The reserved memory on consoles is not used only by the OS, but also by the background applications.
 

Yes, Papermaster mentioned sampling in 2016, how does that mean no availability of chip in 2016?
Can you give link to to support your claim of 2017 release?
If AMD personnel points towards FAD's information, it means the aforementioned information still remains true.
 


Sampling means sending engineering samples to partners for testing purposes. An engineering sample is not a final chip.

Links to what Lisa Su and Papermaster said were posted several times before. The link to the interview to the Zen team head was also given. The only link I cannot provide now is that to the linkedin profile of the engineer that worked on Zen-based chips, because the information was deleted from his profile just after the leak appear in several forums there out.

The AMD personnel that you mentioned only said "launch" in 2016. I explained before that launch can mean "paper launch" and I provided a link to AMD doing that with a Kaveri chip. First a paper launch and then the real chip was released half-year latter. Check it.
 
I have yet to read those informations about zen being in 2017 you mentioned. And i can't find them in search.
 
Well I think Juan could be correct, I mean to hit the target and 'execute' AMD might paper launch in 2016 with actual availability in 2017, although I think it may be more a case they are keeping their options open atm depending on how glofo get on.

My personal prediction: they'll release the top end Zen scu at a high price with the launch in 2016, then make a range of more price competitive (and lower core count) parts available in 2017. I'm certain we'll see 4, 6 and 8 core Zen cpus by the time they flesh out the full line up, although probably not all at once (im assuming 2 dies with the 6 core being harvested 8 core parts).
 


Previous roadmaps wrote something as "4--8 core" for the Piledriver FX CPUs. But known Zen roadmaps only show 8-core CPU.
 


I would be surprised if they only offer it as a full 8 core. I mean they may only make the 1 die of course, but I'd be surprised if they got yields high enough to not have any defective parts.

If they do have defects then 4 or 6 core parts made from otherwise faulty 8 core chips would make a lot of sense (and would be great value options for gamers who are unlikely to need more than 4 cores / 8 threads for the majority of gaming uses)...
 


and to only have 8 core options limits to enthusiast level platform. Isn't zen about scalability and with the ability to put it together like Legos adding a core by just poping it onto the center piece.

What I think will happen here is amd's first venture to the intel scheme of mainstream, mobile, and enthusiast. meaning the 8 core chips (16 thread)? will compete with intel extreme series x99 chipset cpu's. then a quad core to compete with i7 6700K and dual core for i5 and i3. lastly apu's for mobile. they haven't said it yet but I expect to see such a scheme soon.

Think about it like this. amd has always tried to have something for every market intel makes, like 15W ULP devices, amd made Carrizo scalable to 15W and in the past with FX 8350 was to keep up with intel's 8 thread i7's. They have never had anything to compete with top end intel since 2010 at least. its a major gap in their market that the FX line never had a plan to fill. I think this x99 competitor is what amd refers to as the 8 core zen.
 


Throwing cores at Intel wont solve the problem. They also need to improve their IPC and single core performance.

As well, how would a dual core compete with an i5? And i5 is a quad core with no SMT. Their dual core would compete with Intels i3. If they don;t have a quad core no SMT then the i5 would have no real competition which is currently AMDs biggest problem since you can get a i5 that performs the same or better in most consumer applications as a FX 8350 at a lower clock speed and power draw.
 


IPC improvement of 40% remember... more cores helps with rendering so it doesn't hurt them. I forgot yes amd would need a quad core no smt to compete with i5.

Zen is a whole different animal than old BD cores. power consumption will be down due to better design and smaller process node as well as better performance as IPC will be "claimed" ~40% better.

if we use cinebench R15 as example for PD cores we get score around 100 for single core @4.0 ghz. if a overclocked zen core at 4.0ghz is 40% ipc over PD (advertised 40% over excavator) then we would see score of 140 single core. that puts ipc close to haswell. but amd claims another 10% as excavator is 10% better ipc than PD. meaning single core score of 150. putting us close to skylake. Yes it will be slower than skylake but only by 10-20% when you look at clock per clock.

Take that information and add to it the "8 cores (16 threads)" and we have a multi-core score around 1000 to 1400. that's pretty damn good. (150 x 10 = 1500) my assumption is that cores scale poorly getting only 90 points more per core, as all cpu's lose perf scaling to more cores. as a straight 1 to 1 comparison of a 16 thread pushing 150 each would put a multicore score of over 2000, which quite frankly is not going to happen.

the 8350 clocked at 5.0 would be lucky to get 800 as a multicore score. even the 4790K cant hit 1000.

Jimmy you have a point as single core perf will still be lower than intels top offering, but those high core counts will save zen. this cpu has all the potential to be a complete rendering encoding beast. as long as its priced under $1000 😛

EDIT: Please remember that all my math is always relative and I never actually look up numbers. I happen to be fairly certain as to the cinebench R15 scores as I looked them up two weeks ago or so, and have an 8350 that does a single core score of 115 when clocked to 4.8 ( a clock speed I doubt zen will be able to reach.)
 
That is why I have been saying that we are missing a very important piece of information, what are they considering the IPC gain. Is it in single core? Per core? Or per CPU? If it is the latter (CPU to CPU) then it is a overall IPC gain and will not be as much as if it were a per core gain.

Another piece of information we are missing, does that IPC gain include SMT? It most likely does. And is this a 8c 16t Zen vs a 4m 8t Excavator? Or is this on a per "core" level?

There is just too much to hypothesize over right now with too little information to be even close to accurate. Too many variables.
 


Hmm. always assumed they ment single core performance gains in a direct ghz to ghz comparison.

kinda assumed that as amd wanted to show they are massively improving in what was their weak spot of single core perf.

Now that I think about it, it may mean overall ipc like this shows a cumulative improvement of every task the cpu can compute. meaning may not hit full 40% in just single core improvements.
http://images.anandtech.com/doci/9483/01%20-%20Gains%20over%20Sandy.png
 
This graph shows that:

Sandy Bridge to Ivy Bridge: Average ~5.8% Up
Ivy Bridge to Haswell: Average ~11.2% Up
Haswell to Broadwell: Average ~3.3% Up
Broadwell to Skylake (DDR3): Average ~2.4% Up
Broadwell to Skylake (DDR4): Average ~2.7% Up

Oh dear. Typically with an architecture update we see a bigger increase in performance than 2.7% IPC. Looking at matters purely from this perspective, Skylake does not come out well. These results suggest that Skylake is merely another minor upgrade in the performance metrics, and that a clock for clock result compared to Broadwell is not favorable. However, consider that very few people actually invested in Broadwell. If anything, Haswell was the last major mainstream processor generation that people actually purchased, which means that:

Haswell to Skylake (DDR3): Average ~5.7% Up.

EDIT: link to original review. http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/9
 


Assumptions though are just that, assumptions. We don;t know anything about Zen beyond what we are told.

As for the Skylake, a lot is in the power improvements. That is one of the largest focuses right now since, well AMD doesn't have much to threaten their dominance. Instead they are focusing on improving their iGPU and power so they can get more into the UMD market and I am all for a full x86 CPU in a smartphone vs the ARM meh we have had.
 


Don't forget clockspeed. Increasing IPC by 10% per core is pointless if clocks decline 10% in the process; net improvement ends up being 0%. And clocks are almost certainly going to be lower then BD, so that 40% IPC improvements is going to be reduced by however much the clock is. The actual performance improvement is going to be on the order of 20-30%, if I had to guess. And that's assuming the claimed 40% number is typical case, per core, rather then best case, CPU as a whole, otherwise I'd expect 10-20% performance gains at best.
 


No, guys, wait. IPC has to be in single threaded loads. Think about it, they claim a 40% improvement over Excavator, but I'll take Steamroller as an example:

Let's say the A10-7890k does 100 performance points. The A10-7890k is a 4-core CPU. If we consider IPC as a result of the CPU as a whole, then the 8 core / 16 thread Zen will have 140 performance points. That's 8.75 points per thread, or 17.5 per core, against 25 points for the A10. By that logic, Zen would be three times slower than Steamroller in single-threaded loads. That can't be right.

But if you take a quad-core Zen, and assume it does 140 points vs 100 for the A10, remember the Zen handles 8 threads, so that's 17.5 per thread, which is again slower than the A10.

The only way it makes sense is if we consider it in a single-thread workload. Otherwise cores and threads and CMT/SMT make it nearly impossible to compare in a fair way.

The info is 40% IPC increase in single threaded loads.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom