AMD CPUs, SoC Rumors and Speculations Temp. thread 2

Page 59 - 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.


AMD aren't daft- the prices will be high at first however:
I: There will be a range of models, with an entry 'Value' part with reduced core count (I guess 6 cores?) that will be priced to be tempting to the mainstream i5 / i7 users (probably offering a trade off of more cores, lower clocks, more PCIe lanes for SLI / xFire?).
II: The higher binned 6 and 8 core parts will start off priced up with the Intel E-parts
III: The prices on all parts will drop substantially after the initial interest dies down to a stronger competitive position, just like the FX 9XXX parts.

Long and short of it, as with all new tech launches probably worth waiting a couple of months before buying to avoid 'early adopters tax' 😛

 
Gentlemen, USAFRet made a good point on the last page, I suggest you adhere to it. And do not use the report feature because you don't like how someone disagrees with you. While I don't like what I see here, you are skirting the line of attitude problems in all aspects. Please tone down the arguments and attacks.

As he said, facts don't exist for products that don't exist. And tone down the attitudes, or we will
 


My remark was that the "octo-core Zen CPU" targets the "[/i]enthusiast market[/i]". The Zen core will be used in other markets, from semicustom embedded SoCs to large server clusters, and Zen cores will be used in notebooks and mainstream desktop as well. Note that Zen replaces both the Jaguar and the Bulldozer family.

It is likely that Zen uses the same Turbo technology seen on Bristol Ridge. Some people is expecting base clocks in the 2.6--2.8GHz and Turbo in the 3.2--3.6GHz range. The Stilt claims that Zen has a very sophisticated technology to manage power and voltage regulation.
 


Ok, point taken. I forget "Zen" is the code name for the whole uArch that will govern the new dies. My point still stands I guess. If they push -E levels in marketing, we will expect performance and price at those levels as well for the Octo/High versions of Zen.

And now that I remember... Which is also a very important topic IMO. Has there been any news around the Hub/NB? I want to see what the new MoBo chipset will have and all. Bazillion USB3 please be real!

Cheers!
 


I agree.
AMD needs two things; to make profit and gain some market share.
They will have to decide where that fine line "sweet spot" is, which may not be easy to read.
 


I don't agree with the bold part. It is very, *very* easy to spot: ~USD$300. That is where the ~50% of enthusiasts/gamers spend their CPU money. If you think about it, the i7 is just a 4 core CPU with HT, so you can just put it in the same bag averaging the 300 I mention.

http://store.steampowered.com/hwsurvey/cpus/

The other 50% is dual cores still, but that is budget and I would imagine escapes the scope of our speculation for now. AMD won't be showing Zen APUs until later on, which I would believe will attack the lower end of the spectrum. Still for sake of the point being made, that would be ~USD$100, right?

Cheers!
 


I am hearing they are shooting for 3.6/4.0, though they may end up closer to 3.3-3.4/3.7-3.8. Early samples are projecting well, there is optimism they will be close to target.

Using the same process the top Bristol ridge SKU is 3.6/4.0, and it only runs at 65W TDP, but contains 512 SPs @ 948 MHz.
 
Apparently details about Zen will be given officially at Hot chips conference. Check Day 2, the talk "A New, High Performance x86 Core Design from AMD"

http://www.hotchips.org/
 


On the other hand people as Thevenin, who claims to have access to Zen server engineering samples (40501415) is writing:

The relative power consumption is roughly the same as on Intel 14nm parts with similar configuration, but the clocks are quite low :/
Click to expand...

Therefore to whom I trust?



Bristol Ridge is not using the same process node. Bristol is a 28nm chip.

Power is a function of area and frequency. Mentioning 65W for an APU means nothing because about half the die is clocked very low (~1GHz), which reduces the power quadratically or cubically (depending of parameters of the process node used).

If you replace the iGPU by CPU cores, then the resulting hypothetical 8-core Excavator CPU @ 3.6/4.0GHz on 28nm would be rated above the 100W. To not mention you are ignoring the lack of L3 on Bristol (which adds to the power consumption).

If you cannot do the math by yourself check at least Intel chips to see the pattern. A quad-core APU can be rated at 65W or similar, but octo-core CPUs are rated at 140W or so.
 


Yes, I realized that, I was thinking about Zen APU, I had forgotten that Bristol Ridge is the Construction cores last hurrah.

As for power consumption...I am well aware how it all functions Juan...in fact, I think you seriously need your attitude adjusted. Back off, and stop talking down to everyone in the thread.

As for server parts...considering that I know, for a fact Server SKUs are only sampling early stages at the moment, and nothing near a completed product, I would say his opinion of the clockspeeds is probably prematurely formed...much like your opinion of everything AMD these days.
 


If you read what I wrote, I said that he has access to a Zen server engineering sample (and I gave you a numeric code). When he wrote that power consumption is fine but clocks are bit low, he means clocks are low for what he would expect for an engineering sample, otherwise he wouldn't do the remark. And his experience with the engineering sample is the reason why he wrote that he expects 8C desktop CPU to be clocked ~2.6GHz.

On the other hand you contradict him and claim that the 8C desktop CPU will hit much higher frequencies: " they are shooting for 3.6/4.0, though they may end up closer to 3.3-3.4/3.7-3.8. Early samples are projecting well, there is optimism they will be close to target." Both you and me know that samples are clocked at lower frequencies. What is the frequency of the current 8C desktop CPU samples? 1.8GHz? 2.1GHz?
 


Last I heard was 2.4 currently.
 


Ok, that is close to the 2.1GHz that I wrote. Now the question is if current sample silicon for Zen is running at 2.4GHz why do you believe that final silicon for the 8C Zen CPU will hit 3.6GHz base and 4GHz turbo late this year? How do you know that the process node will not hit a frequency wall near 3GHz as others are expecting?

Note I also see that you are lowering expectations for clocks. Months ago you claimed 4.0--4.1GHz base clocks. A pair of days ago you reduced that to 3.6GHz base clock and lately you are even considering lower frequencies of 3.3GHz. That is a 20% reduction in frequency!!!!
 


You have got to be kidding me, you've already been warned today, now its time for a vacation.

NO personal attacks, period
 
Cmon guys, we're all interested in the same things and I think everyone wants AMD to succeed with this, even those with a more sceptical stance.

We were having a really nice discussion a couple of pages back, no need for animosity :)
 


Considering PD also sampled at this stage at 2.4, I am not concerned that a ceiling will arise...however...what I hear is optimism that they are trying to hit 3.6 base. What I am hearing is that they expect to be within 300 MHz of that target assuming current trends continue.

I expected 4.0 GHz in 2014. That was what I was told was their goal, at that time. Things change, and expectations change, and at this time, the current information I have says they expect to be close to 3.6 + turbo on the top SKU in HEDT. Server parts will be lower...in fact, I would be amazed if they hit 3.6 on a 16c/32t part...I would expect server chips to be much closer to 2.8-3.1. Even if you review the older opterons, lots of those were still relatively low clocks on that uarch with higher core counts.
 


Piledriver is not a valid comparison. Piledriver is made on 32SHP node --with SHP meaning Super High Performance-- derived from a cooperation with IBM foundries which fabricated 32nm IBM CPUs clocked at 5.5GHz base and higher.

Zen is made on a mobile-oriented process: the 14LPP node --with LPP meaning Low Power Plus--. Thevenin wrote that 14LPP is optimized for 2.5GHz and lower frequencies:

The 14nm LPP performs best (efficiency wise) up to 2.5GHz, and takes a significantly nose dive at frequencies above 3GHz.
Click to expand...

This is not a process designed to break worldwide overclocking records.



Back in 2014, I said that the 4.0--4.1GHz was a pipe-dream and I wrote then that 3.0--3.5GHz was a more realistic target. But you are right that things change, and lately people is claiming that the process is poor than expected. This is why they are reducing the expected based clocks to the 2.6--2.8GHz range.
 

Maybe they expect ~3.6 on the top 4-core option? Not even Intel has 3.6 GHz on their 8-core CPUs.
 


http://www.intel.com/buy/us/en/product/components/intel-core-i7-6900k-32ghz-20mb-smart-cache-box-515529

They do hit 3.2
 


But Intel is experienced in the architecture, have their own process, and have much more money. I doubt AMD will get such high clocks on the first generation. I mean, I sure want them to, but ~3.6 I only expect at most on 6-cores, but most probably on 4-cores.
 
Well, we already saw something similar before: Phenom I. They missed the target speeds by a lot and we had to wait for Phenom II to pick up the ball in several fronts.

I don't think Zen is going to be that bad, but it has happened before, so there's that.

Cheers!
 
To sum some things up in the last few pages. It would seem like clock speeds for ZEN will be around the low 3 GHz clock range or even 2.8 GHz range. To be a positive person I will say Bulldozer was first getting clock speeds extremely low (was wildly reported by motherboard manufactures) and they ended up within a 10% lower target then they originally wanted.

As for turbo mode Gamer made a good point in that Windows likes to keep threads from getting locked into keeping the work in one or two cores. This would mean even if a reported 1 GHz turbo boost was possible for 1 core or so it would be in a perfect situation. Real targets would probably be lower.

The real question will be the price and who is this for? Gamers still want high clock high IPC CPUs and to be honest a lot of applications still need this over higher core count (look at 6700K vs. 6950X). This I fully expect a 6 core Zen will probably be a poor choice over a 4 core Intel CPU with HT unless it’s priced low. As for price I highly expect a 450$ price range for their top in part. It’s obvious that they will probably end up comparing their top end CPU to Intel’s 1700$ CPU (I can see the marketing slides now).

I fully expect Ivy-sandy IPC but with 8 cores myself but at a 95 watt power target. I still kind of expect 125 watt CPUs as well cause why not? They can increase clock speed by 200-400 MHz and enthusiast wouldn’t care about the extra power draw.

But this is just my opinion I could be wrong after all this entire forum does support speculation.
“Forming a theory or conjecture without firm evidence” or pure speculation.
 
I just noticed that AMD is probably going to call Zen, Zen. Since thats what they had in the reveal video. I don't think they use codenames for promotional videos like that. Wonder if that will become a new branding.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom