New Zen microarchitecture details

[coercitiv]

Which is why they should release low TDP undervolted CPU's for small form factor pc builds.

It wouldn't be undervolted, just configured for lower TDP. Considering the advanced power management features AMD boasted for Zen (sensors, frequency control, XFR etc) it would be really odd if TDP control via BIOS were absent. This feature is present in modern Intel CPUs & chipsets, and AMD also has it in FM2+.

You buy a 65/95W CPU, set TDP limit via BIOS to 45W (or 25-35W if you so desire) and let power management choose optimum frequency for all kinds of loads. It's easy, efficient and flexible, since you can even tune max TDP based on your system characteristics (cooler size, case size and airflow, ambient temps etc). Why settle for a 45W TDP CPU when your case&cooler can take 55W?

 

[CentroX]

Since intel is already rolling out new cpus this august (gee i wonder why) amd needs to have ryzen out soon. A delay would be devastating at this point.

 

[KTE]

I'm not really interested in TDP. I couldn't care less about it for HP CPUs used at home.

But I do care about power for work builds, and in reality, if the performance@power is good, then it's good to know we can shrink that down to a nice laptop, like Intel has done since Conroe.

Probably my only concern with power.

Sent from HTC 10
(Opinions are own)

 

[KTE]

Since intel is already rolling out new cpus this august (gee i wonder why) amd needs to have ryzen out soon. A delay would be devastating at this point.

It's already heavily delayed, and it was always the case that IF AMD did get anywhere near Intel, Intel would launch a higher clocked CPU and price war the rest.

It's obvious Intel will try to ruin Ryzen launch/sales with faster, unexpected models so that it looks worse than it is.

Sent from HTC 10
(Opinions are own)

 

[CentroX]

It's already heavily delayed, and it was always the case that IF AMD did get anywhere near Intel, Intel would launch a higher clocked CPU and price war the rest.

It's obvious Intel will try to ruin Ryzen launch/sales with faster, unexpected models so that it looks worse than it is.

Sent from HTC 10
(Opinions are own)

I wouldn't say heavily delayed, but it is about 3 months delayed right now. Would we see Intels X models already in august if ryzen wasnt coming? No chance imo.

 

[lolfail9001]

I wouldn't say heavily delayed, but it is about 3 months delayed right now. Would we see Intels X models already in august if ryzen wasnt coming? No chance imo.

August would make it 3 years after Haswell-E, right on Intel's schedule.

 

[Sven_eng]

It's already heavily delayed, and it was always the case that IF AMD did get anywhere near Intel, Intel would launch a higher clocked CPU and price war the rest.

It's obvious Intel will try to ruin Ryzen launch/sales with faster, unexpected models so that it looks worse than it is.

Sent from HTC 10
(Opinions are own)

Do you think Intel will launch new CPUs that match AMD in perf/$$$? That will be quite the price cuts all around...

 

[KTE]

Do you think Intel will launch new CPUs that match AMD in perf/$$$? That will be quite the price cuts all around...

For sure, if need be.

Unless they have far more absolute performance. Then they won't bother.

Sent from HTC 10
(Opinions are own)

 

[krumme]

How can Intel comming new models be unexpected yet it is obvious ?
And the obvious unexpected cpus will only come if need be?
That is some funky stuff we have here
Is there some arguments or sources or just a tad more than obvious unexpected if needed cpu?

 

[coercitiv]

How can Intel comming new models be unexpected yet it is obvious ?

Examples:

• Binned 4+2 cores to squeeze extra performance. KBL-X can be launched "early" on 1151. Now with solder!
  
• HEDT 6c/12t chips at current mainstream i7 prices. They can drop all HEDT SKUs one tier down, maybe add a faster bin to maintain supremacy in charts.

All of this heavily depends on how Zen performs in all tiers it can address, availability also matters. You can rest assured Intel will not allow Zen to top the charts, even if that means pooping golden chips

 

[krumme]

Examples:

• Binned 4+2 cores to squeeze extra performance. KBL-X can be launched "early" on 1151. Now with solder!
  
• HEDT 6c/12t chips at current mainstream i7 prices. They can drop all HEDT SKUs one tier down, maybe add a faster bin to maintain supremacy in charts.

All of this heavily depends on how Zen performs in all tiers it can address, availability also matters. You can rest assured Intel will not allow Zen to top the charts, even if that means pooping golden chips

Meh they already have a 10c hedt cpu to top the charts. All is well known what they can do. There is no magic but as a the current core lineup is similar to 6 years ago its evident they can ajust pricing at move whatever into the segment they want especially when the 6c line in up and running.
Besides its Intel inside. So its pretty wellknown brand product.

 

[Sven_eng]

I think Kaby Lake and the HT Pentiums were the response...

 

[coercitiv]

Meh they already have a 10c hedt cpu to top the charts.

Not the "mainstream" consumer charts with $200 and $300 price tiers. That's what I was talking about, especially the $300 segment. Both bullet points in my list were actually addressing the same tier, but with different approaches based on possible strong / weak points in the the Zen offering.

On the other hand, we're just wasting forum ink on this subject, until we have decent info about Zen SKUs (configs, pricing, availability) we can just as well assume Intel may not need to adjust anything.

 

[Gideon]

What I'm worried about, is that there still are no leaks whatsoever about the SKUs. While I can understand no info about pricing (AMD might still be figuring it out), we should get leaks about the SKUs at least 3-4 weeks before, considering the need to distribute them, and the number of people involved.

Hopefully my concerns are unfounded, but a March launch looks more and more likely.

For comparison, we got info about the Kaby Lake SKUs in the very end of October, while the launch was in the beginning of January.

 

[krumme]

Amd messed up phenom and bd prices day one setting them way to high. Go read eg AT review. So its understandable that some people expect them to make similar idiotic mistakes this round. But my take on it is the entire organization is better now than then.
They need stable prices. Well known names. Predictable performance. Predictable functionality. Like core lineup. Thats how it should be done.

 

[iBoMbY]

What I'm worried about, is that there still are no leaks whatsoever about the SKUs. While I can understand no info about pricing (AMD might still be figuring it out), we should get leaks about the SKUs at least 3-4 weeks before, considering the need to distribute them, and the number of people involved.

Hopefully my concerns are unfounded, but a March launch looks more and more likely.

For comparison, we got info about the Kaby Lake SKUs in the very end of October, while the launch was in the beginning of January.

I guess they would need to start shipping this week, if they want to have the stuff in stores at the end of February. I also would've guessed to see something sooner, but there is no big reason of concern yet.

 

[TemjinGold]

For sure, if need be.

Unless they have far more absolute performance. Then they won't bother.

Sent from HTC 10
(Opinions are own)

Well it's settled then. No one should buy anything Ryzen. We should all just wait till August for Intel's obviously better stuff.

 

[lolfail9001]

Well it's settled then. No one should buy anything Ryzen. We should all just wait till August for Intel's obviously better stuff.

Well, we waited long enough for Zen that waiting few more months and see how Basin Falls turn out is only reasonable.

 

[LTC8K6]

Either you're being very sarcastic, or you know very little about electricity. A computer consumes a very modest amount of electricity compared to common house hold applicances. Anything that generates heat is easily in the 1000W+ range already. Your average toaster, hair dryer, water kettle, etc all consume over 1000W, some over 2000W. I don't get it why people are scared about 100W cpus. That's quite irrational.

How could it be anything but sarcasm?

I don't think it's odd to start looking at the power your computer system is using, and try to reduce it. Especially if you have a high end CPU and a high end GPU.

People generally don't run their hair dryers or water kettles for hours at a time every day like a gamer would run his 2.5Kw TDP cpu with external Ron Davis radiator...

As far as not caring about a 2000 watt TDP cpu, I think most people would soon quit gaming with a 2Kw+ computer system for hours a day when they saw what it added to the electric bill.

 

[LTC8K6]

Well it's settled then. No one should buy anything Ryzen. We should all just wait till August for Intel's obviously better stuff.

That's how the waiting game works, though.

And that's why it's difficult to play that game.

There is always something new on the horizon, and the mfg always claims it's better.

 

[krumme]

Its very simple. It depends on zen pricing. If you can get a base 8c for 400usd or less Intel will not follow it. If its 700 its imo worth waiting for the 6c response.

 

[Atari2600]

That's how the waiting game works, though.

And that's why it's difficult to play that game.

There is always something new on the horizon, and the mfg always claims it's better.

True...

Although if AMD stick to the AM4 platform* over the life of Zen, Zen+ and Zen++, then to a degree you can have your cake and eat it too.


*Unlike Intel which seemingly require a new MB with every generation.

 

[raghu78]

Well, we waited long enough for Zen that waiting few more months and see how Basin Falls turn out is only reasonable.

You know the problem with that is AMD has confirmed a tock every year for the next 3 years 2018-2020. I think we could easily see a Zen+ in Q1 2018 with higher IPC and clocks. Intel does not have any tock for 2018 and till well into 2019 (maybe mid year). The waiting game is not going to help. Just get the hardware you want when you need it or when the price perf is just too good to miss(if you value price perf).

http://www.pcworld.com/article/3155...hitecture-is-expected-to-last-four-years.html

When asked how long Zen would last, compared to Intel’s two-year tick-tock cadence, Papermaster confirmed the four-year lifespan and tapped the table in front of him: “We’re not going tick-tock,” he said. “Zen is going to be tock, tock, tock.”

 

[krumme]

True...

Although if AMD stick to the AM4 platform* over the life of Zen, Zen+ and Zen++, then to a degree you can have your cake and eat it too.


*Unlike Intel which seemingly require a new MB with every generation.

Yep and that would be great if am4 will do that. But if you can get a base 8c zen for 400usd one have to wonder if it really matters.
Let say a zen++ on 7nm is still 8c and have 12% higher ipc and 20% higher freq at say 65w tdp. Is that worth an upgrade then? Meh i would say. Better just get the 8c from the start if you want to upgrade anyway.

 

[krumme]

I think the upgrade path is far more important on server side. Think about replacing a 32c 180w tdp cpu with a new 48 or 64c cpu in the same tdp footprint and socket.
That would be a damn fine business proposal. Easy to calculate tco cost and therefore low risk investment.