New Zen microarchitecture details

[DrMrLordX]

Do you really think *any* amount of data is enough to stop the arguing in these forums?

No, but it can at least bring us back to the deliberately obtuse misinterpretation and/or twisting of data to make points that prevails is in the usual arguments around here. The present arguments are based on mere fragments of data.

 

[swilli89]

No, but it can at least bring us back to the deliberately obtuse misinterpretation and/or twisting of data to make points that prevails is the usual arguments around here. The present arguments are based on mere fragments of data.

Ha! So true. I think Anandtech forums are so unique because they've been around for so long and people have drawn lines in the sand and refuse to alter their opinion on certain things. Maybe not.. Maybe this place is just haunted by tech company shills. Either way it's always an entertaining read

 

[bjt2]

3. Still ES, again, remeber haswell ES clocking to the sky 5+ and early retail stumbled on 4.3-4.4? (could be fabbed elsewhere or differently, tim instead of solder or whatever)

But here we are not talking of OC. We are talking of an early ES that clocks higher than a retail chip, and still with less consumption and comparable (if not slightly higher) performance.
Maybe you are right and Zen will not overclock over 4.5Ghz, but you should admit that retail Zen will have same or better perf and perf/W than these ES, otherwise AMD would have not shown to us...

 

[bjt2]

Actually what make him confident are his tests of an early sample, that s what he explicitely state.

Now he measured 93W on the 12V ATX rail, wich accounting for losses makes about 82W for the CPU, but he dont know to what extent the chip was overvolted since this is an ES and an early one.

If we take his number of 82W as basis and the fact that he state that the chip did turbo to 3.3, at least occasionaly, and that it s at this frequency that he indeed measured those 82W max power then he s uncautious to pretend that the chip exceed 95W at 3.6 (he didnt state the frequency@95W+..), because the chip wont consume significantly more than at 3.3, expect something like 20%, so if there s an excess it s a couple watts, and still, likely due to a slight overvolting..

I am wondering if the CPU 12V rail powers also the DRAM VRMs...

 

[superstition]

Ha! So true. I think Anandtech forums are so unique because they've been around for so long and people have drawn lines in the sand and refuse to alter their opinion on certain things. Maybe not.. Maybe this place is just haunted by tech company shills. Either way it's always an entertaining read

If not for the investment people in forums make in knowing the tech situation the products we're given to buy would be even worse. Imagine that for a moment.

Accountability is noisy but, really, we could do with more of that. There are tech practices that are really shady and higher standards should be demanded in those cases.

The more people feel invested in something the more likely they'll leave their objectivity at the door, unfortunately. So, just as we need to be vigilant about companies' products, we need to be vigilant about the quality of what is said here.

 

[cytg111]

..Maybe you are right and Zen will not overclock over 4.5Ghz, but you should admit that retail Zen will have same or better perf and perf/W than these ES, otherwise AMD would have not shown to us...

- I am right about what? I've never stated that it wont. I've just stated that you cant put forth an absolute *right now* stating that it will. I think there's a good chance it will, if not at first then a few months down the road, rev1.1 or whatever.. But who knows.
I should admit something? I dont exist in this black and white universe you describe . Ill admit to nothing, nothing I tell ya.

 

[Dresdenboy]

Well, that is nice, and maybe he did a really excellent job at balancing the compromises in the final product. But the statement itself is kind of asinine. I mean duh, doesnt every cpu designer aim to come to the best balance between frequency and power usage? I seriously doubt a group of engineers has ever had a design goal to make a cpu reach 4ghz and use as much extra power as possible.

For most bigger designs of the last years this should be true. But there might be differences in how power consumption is being tracked from the start.

But think of designs pre 2K and maybe even the P4 variants (which went with domino logic and other fun stuff). There likely was an inherent power limit induced by area, while leakage was relatively small. And area has been kept low due to costs.

 

[krumme]

You can set the strategic goals as you want. And bad management set unrealistic goals because they dont have the professional insight to give clear directions. I think we have all have experienced a top manager with lofty braindead goals and thinking every moron could go on stage rambling like that.
What you need to set goals and get there is competences already present in the organization. A good manager knows what competences is at hand and he talks from there.

Now what Keller is saying is not what fo4 is or if zen will oc to 4.3 or 4.9.
What he is stating in that vid is that amd have the competences for high freq design and low power design and that those competences was in play. Bd and jag help develop those competences.

We know zen is 3.6 base 4 turbo in a 95w tdp. So no crap there. It is a high freq low power design. We have a 180w 32 server core comming. I would say they pretty much build an near impossible bridge here. With a meager 300 man working on the core part.

If you can do that for those cost the impact will be huge going forward.

 

[Atari2600]

Could be worse... I can think of Juan person that is sporting an impressive case of cognitive dissonance.

teeheeheehee

 

[bjt2]

- I am right about what? I've never stated that it wont. I've just stated that you cant put forth an absolute *right now* stating that it will. I think there's a good chance it will, if not at first then a few months down the road, rev1.1 or whatever.. But who knows.
I should admit something? I dont exist in this black and white universe you describe . Ill admit to nothing, nothing I tell ya.

You were casting doubt on 5Ghz OC, because intel couldn't. But intel couldn't do a lot of things that AMD did, like 4.7GHz/5GHz 8 core on 32nm or 65W 4.3GHz APU on 28nm BULK... So this is not a valid reason to say that AMD can't do a 5GHz OC chip.
I said that maybe you are right, but by chance... Not by strong reasoning...
I gave reasoning that maybe AMD can reach 5GHz OC. The reasons are following:
1) Zen is an high frequency design: indeed it has +400MHz base frequency than INTEL with at most slightly less power consumption
2) 5GHz OC was even obtained on air, albeit on one core, on a preproduction MB and on an early ES
For those two reasons i would expect at least the same OC than the 6900K, but probabily more and perharps in the 5GHz range...

 

[guskline]

Until RyZen is released retail all we can do is speculate.

I really am tempted to do a Ryzen 8c/16t build that is water cooled to see what effect good cooling (EK Supremacy EVO cpu block-I have a need block with the correct AMD4 plate) can do and whether or not the good cooling will improve overclocking.

 

[Atari2600]

I really am tempted to do a Ryzen 8c/16t build that is water cooled to see what effect good cooling (EK Supremacy EVO cpu block-I have a need block with the correct AMD4 plate) can do and whether or not the good cooling will improve overclocking.

Yep. Same thinking.

I've currently an old 1100T running in a big Cosmos II box. Its now pretty tired and some of the SATA ports are becoming erratic.


If Zen delivers on the promise to date, I'll eagerly do an organ transplant on the box. If it doesn't, I'll probably begrudgingly have to go get a KL.

 

[cytg111]

You were casting doubt on 5Ghz OC, because intel couldn't..

And I repeat my self. I will admit to nothing. .
I did no such thing, Intel never entered into that equation.
I put forth haswell as an example where performance regressed from ES to final. AMD has introduced new shiny products in the past that regressed performance as well. Thats not the point.
My point is that observing this "Unidentified Flying Object" called Zen and suddenly I see people posting "It will clock at 4.6179 GHz and sip 95.05 watts of power". How can you possibly know that - when its still unidentified? I am not saying its aliens but.. <insert that image here>
And ill stop adding to this debate with this; I think some of you are sort of searching for these AMD vs. Intel trenches more than you realize.

 

[coercitiv]

I think some of you are sort of searching for these AMD vs. Intel trenches more than you realize.

Fan or Foe

 

[Glo.]

Guys, do you think there will be 4C/8T version of Ryzen CPU with 45W TDP?

I want, fast, but power efficient computer, and if there is no 45W version of Ryzen then only option for me is 4C/8T 95W Raven Ridge APU.

 

[AtenRa]

Guys, do you think there will be 4C/8T version of Ryzen CPU with 45W TDP?

I want, fast, but power efficient computer, and if there is no 45W version of Ryzen then only option for me is 4C/8T 95W Raven Ridge APU.

Down-clock/under-volt and you are set to go, also there may be a TDP setting in BIOS like in FM2+.

 

[Abwx]

A 4C/8T should be within 45W@3.4GHz and 2.5GHz for a 8C/16T, no need of dedicated SKUs since all these are unlocked.

I am wondering if the CPU 12V rail powers also the DRAM VRMs...

Dunno, that s possible, but this wouldnt be cautious to feed the RAM regulator from a rail that has huge residual noise due to the highly varying CPU loading...

 

[Atari2600]

Guys, do you think there will be 4C/8T version of Ryzen CPU with 45W TDP?

Talk seems to be of the 6C12T squeezing in under 65W. No word of 45W quads.

I'm sure, as others say, you'll be able to downvolt - infact, with XFR etc, it may be even easier as you should be able to tune it all in software.

If you want a power sipping 45W TDP for battery life, I'd like to think two profiles (battery or mains) would allow you the best of both worlds.

If you want a low cooling solution for HTPC, then I expect while watching movies etc, Zen will be in a high Pstate anyway, so a 65W CPU will only be sipping a few watts.

 

[imported_jjj]

Guys, do you think there will be 4C/8T version of Ryzen CPU with 45W TDP?

I want, fast, but power efficient computer, and if there is no 45W version of Ryzen then only option for me is 4C/8T 95W Raven Ridge APU.

CPC was saying the quads are 65W , i think the 45W might be a WCC... invention , their excessively simplistic way of dividing 95W by 2 for half the cores.

Chances are the SKUs are limited at launch, maybe they add lower power versions over time.

 

[Abwx]

CPC was saying the quads are 65W , i think the 45W might be a WCC... invention , their excessively simplistic way of dividing 95W by 2 for half the cores.

.

That s not as simplistic as one would think...

FI a FX4350 clocked at the same frequency/voltage than a FX8350 would use 55-60% of the latter s power, and the 4350 has a full 8MB cache, with half the cache to match the core count we would be closer to 50%.

 

[ecogen]

CPC was saying the quads are 65W , i think the 45W might be a WCC... invention , their excessively simplistic way of dividing 95W by 2 for half the cores.

Chances are the SKUs are limited at launch, maybe they add lower power versions over time.

Slight correction, CPC said the 6C is @65W.

 

[Glo.]

A 4C/8T should be within 45W@3.4GHz and 2.5GHz for a 8C/16T, no need of dedicated SKUs since all these are unlocked.

3.4 GHz/3.7 GHz, 45W TDP CPU? Sign me up for it.

Perfect situation would be when that CPU would also cost much lower than Intel 35W i7-7700T.

 

[MrTeal]

That s not as simplistic as one would think...

FI a FX4350 clocked at the same frequency/voltage than a FX8350 would use 55-60% of the latter s power, and the 4350 has a full 8MB cache, with half the cache to match the core count we would be closer to 50%.

In power used yes, but both the FX8350 and FX4350 are 125W TDP CPUs. Until they actually announce it or there's a leak, we probably won't get an idea of whether there will be any 45W TDP Ryzen quads even if you can make them consume that much.

 

[imported_jjj]

Slight correction, CPC said the 6C is @65W.

CPC did mention 65W for 6 cores but also for 4 cores.

"I'm not aware of any Ryzen ES w/ 45W TDP. All current 4C samples are rated at 65W. 45W TDP would require a max. frequency @ 2.8G (w/ 4C)

"https://twitter.com/CPCHardware/status/819435019588567040

 

[ecogen]

CPC did mention 65W for 6 cores but also for 4 cores.

"I'm not aware of any Ryzen ES w/ 45W TDP. All current 4C samples are rated at 65W. 45W TDP would require a max. frequency @ 2.8G (w/ 4C)

"https://twitter.com/CPCHardware/status/819435019588567040

Huh. Hadn't seen that, thanks.