Discussion: AMD Ryzen

[manleysteele]

Does anyone know if zen is set to use a different location or method to report thermals? As it's been for quite some time they use some sort of mystery algorithm which makes it difficult to get accurate temp readings with other programs aside from their overdrive software and even that can be a bit off at times.

I'm just wondering if it's something they've addressed since zen is finally a 'do over' moment for them to perhaps fix some things that have been a bit of a pain for users.

http://www.tomshardware.com/faq/id-2122665/understanding-temperature-amd-cpus-apus.html

Have you read this yet? It explains what you are seeing with a load on AMD.

 

[os2wiz]

Except this isn't Kaveri and its not some other design its Zen and its using 14nm by Global foundry. 400 series hasn't even been known to have a great overclocking experience and the prices on the 400 series built on the same tech as Zen have been way higher then what they should be hinting at a yield issue. I highly doubt anything above 3.2Ghz will be achieved from a 8 core Zen. I still think we will see Zen hit higher targets but not until yields improve.

Edit even Intel doesn't have those types of targets for their 8 core CPU just seems a bit much to me to assume. Even more so when they are trying to hit a 95 watt TDP target.

Actually while GloF is implementing the process, it is licen s ed by them from Samsung. Just wanted to be 100% accurate.

 

[os2wiz]

Except this isn't Kaveri and its not some other design its Zen and its using 14nm by Global foundry. 400 series hasn't even been known to have a great overclocking experience and the prices on the 400 series built on the same tech as Zen have been way higher then what they should be hinting at a yield issue. I highly doubt anything above 3.2Ghz will be achieved from a 8 core Zen. I still think we will see Zen hit higher targets but not until yields improve.

Edit even Intel doesn't have those types of targets for their 8 core CPU just seems a bit much to me to assume. Even more so when they are trying to hit a 95 watt TDP target.

Actually while GloF is implementing the process, it is licensed by them from Samsung. Just wanted to be 100% accurate.

 

[juanrga]

I just wanted to add 1 more thought. When Kaveri was in development, I distinctly remember similar talk about an engineering sample. I believe it also was 2.8 GHZ. I believe the initial Kaveri release was the 7850 with a 3.5 GHZ clock speed.

The latest engineering samples of Kaveri were clocked at 3.7GHz, which is the same frequency than got final silicon.

We should no far more by the end of October after product starts to ship to the channel.

Zen doesn't start to ship in October. The final silicon is not still ready.

 

[os2wiz]

I just wanted to add 1 more thought. When Kaveri was in development, I distinctly remember similar talk about an engineering sample. I believe it also was 2.8 GHZ. I believe the initial Kaveri release was the 7850 with a 3.5 GHZ clock speed.

The latest engineering samples of Kaveri were clocked at 3.7GHz, which is the same frequency than got final silicon.

We should no far more by the end of October after product starts to ship to the channel.

Zen doesn't start to ship in October. The final silicon is not still ready.

When is the production date of that 2.8 GHZ sample???

 

[os2wiz]

I just wanted to add 1 more thought. When Kaveri was in development, I distinctly remember similar talk about an engineering sample. I believe it also was 2.8 GHZ. I believe the initial Kaveri release was the 7850 with a 3.5 GHZ clock speed.

The latest engineering samples of Kaveri were clocked at 3.7GHz, which is the same frequency than got final silicon.

We should no far more by the end of October after product starts to ship to the channel.

Zen doesn't start to ship in October. The final silicon is not still ready.

More than one website, including semi-accurate mentioned first processors shipping later in October. Not sure why there all these conflicted reports.

 

[jdwii]

Except this isn't Kaveri and its not some other design its Zen and its using 14nm by Global foundry. 400 series hasn't even been known to have a great overclocking experience and the prices on the 400 series built on the same tech as Zen have been way higher then what they should be hinting at a yield issue. I highly doubt anything above 3.2Ghz will be achieved from a 8 core Zen. I still think we will see Zen hit higher targets but not until yields improve.

Edit even Intel doesn't have those types of targets for their 8 core CPU just seems a bit much to me to assume. Even more so when they are trying to hit a 95 watt TDP target.

Actually while GloF is implementing the process, it is licen s ed by them from Samsung. Just wanted to be 100% accurate.

Sure but that still doesn't make the points i made any less relevant i guess we won't know for sure until it comes out but as of now i have no doubts really that it will be clocked much closer to 3Ghz then even 3.6Ghz, i do not expect Turbo to hit 4.0 or above either.

 

[juanrga]

I just wanted to add 1 more thought. When Kaveri was in development, I distinctly remember similar talk about an engineering sample. I believe it also was 2.8 GHZ. I believe the initial Kaveri release was the 7850 with a 3.5 GHZ clock speed.

The latest engineering samples of Kaveri were clocked at 3.7GHz, which is the same frequency than got final silicon.

We should no far more by the end of October after product starts to ship to the channel.

Zen doesn't start to ship in October. The final silicon is not still ready.

More than one website, including semi-accurate mentioned first processors shipping later in October. Not sure why there all these conflicted reports.

More than one website claimed 2016 launch for Zen, because they got it wrong. More than one website is now claiming that Zen launches in February, when AMD didn't say that.

 

[os2wiz]

One quite important question that I asked you did NOT answer. I had asked when the engineering sample of 2.8GHZ base frequency, mentioned in all the benchmark tests, was produced. Without a date it is impossible to say if it is one of their latest samples. If it is not a very recent sample then it is possible there is more room for improving base frequency.

I just wanted to add 1 more thought. When Kaveri was in development, I distinctly remember similar talk about an engineering sample. I believe it also was 2.8 GHZ. I believe the initial Kaveri release was the 7850 with a 3.5 GHZ clock speed.

The latest engineering samples of Kaveri were clocked at 3.7GHz, which is the same frequency than got final silicon.

We should no far more by the end of October after product starts to ship to the channel.

Zen doesn't start to ship in October. The final silicon is not still ready.

More than one website, including semi-accurate mentioned first processors shipping later in October. Not sure why there all these conflicted reports.

More than one website claimed 2016 launch for Zen, because they got it wrong. More than one website is now claiming that Zen launches in February, when AMD didn't say that.

 

[jdwii]

Spoiler

One quite important question that I asked you did NOT answer. I had asked when the engineering sample of 2.8GHZ base frequency, mentioned in all the benchmark tests, was produced. Without a date it is impossible to say if it is one of their latest samples. If it is not a very recent sample then it is possible there is more room for improving base frequency.

I just wanted to add 1 more thought. When Kaveri was in development, I distinctly remember similar talk about an engineering sample. I believe it also was 2.8 GHZ. I believe the initial Kaveri release was the 7850 with a 3.5 GHZ clock speed.

The latest engineering samples of Kaveri were clocked at 3.7GHz, which is the same frequency than got final silicon.

We should no far more by the end of October after product starts to ship to the channel.

Zen doesn't start to ship in October. The final silicon is not still ready.

More than one website, including semi-accurate mentioned first processors shipping later in October. Not sure why there all these conflicted reports.

More than one website claimed 2016 launch for Zen, because they got it wrong. More than one website is now claiming that Zen launches in February, when AMD didn't say that.

Anything is possible but when Amd showed Zen they down clock the comparable Intel CPU instead of just picking some Zen chip that could already do 3.2Ghz, if anything that would tell me that 8 core Zen they showed off was probably hand picked already. I seriously can't understand how one could think a 8 core Zen would have a frequency at 3.6Ghz or more when its going to be a 95 watt part. Intel's own superior 14nm with a 8 core Broadwell is only doing 3.2Ghz out of the box with a TDP of 140 watts. At this point in time i don't even see Amd hitting 2.8-3.2Ghz at 95(maybe 125) watts using their 8 core Zen chip on Global foundry's "14nm".

That's not to say their 4 core Zen won't have a clock speed of 3.6Ghz out of the box.

 

[juanrga]

One quite important question that I asked you did NOT answer. I had asked when the engineering sample of 2.8GHZ base frequency, mentioned in all the benchmark tests, was produced. Without a date it is impossible to say if it is one of their latest samples. If it is not a very recent sample then it is possible there is more room for improving base frequency.

In the first place, you asked about an unidentified chip which you refereed to as "2.8 GHZ sample" in a post where you previously introduced in the discussion some inexistent 2.8GHz Kaveri sample and after I replied you that the last engineering samples of Kaveri did run at 3.7GHz.

I guess now you are really asking about the several 2.8GHz Zen samples (plural) that AMD has been shipping for a while. My answer is that I don't know.

In the second place, the question is irrelevant, it doesn't matter when an engineering sample is made. What really matters is what silicon level uses the sample. I have stated multiple times than the current octo-core and quad-core engineering samples are A0 stepping silicon. I have stated this is silicon very close to the final production. I have also stated that people as The Stilt expect final silicon to get 200--400MHz higher clocks than current A0 silicon. This means about 3.0--3.2GHz base clock.

In the third place, as mentioned also here in the forums, the first engineering samples of Zen did run at 2.4GHz. Your own 20--25% rule implies 2.88--3.0GHz for final silicon.

My own prediction about clocks can be found in this thread.

 

[juanrga]

I seriously can't understand how one could think a 8 core Zen would have a frequency at 3.6Ghz or more when its going to be a 95 watt part. Intel's own superior 14nm with a 8 core Broadwell is only doing 3.2Ghz out of the box with a TDP of 140 watts. At this point in time i don't even see Amd hitting 2.8-3.2Ghz at 95(maybe 125) watts using their 8 core Zen chip on Global foundry's "14nm".

Many claims one can find in the Internet about Zen just ignore physical and technological aspects.

Personally, I find very weird to see certain tech reporters (like BitsandChips) that claim that Summit Ridge 8C will get base clocks higher than 4GHz on a mobile oriented process like GF-14LPP, when IBM just had to choose GF-14HP to get 4GHz for its own chips. Basically what those reporters claim is that the High Performance process node provides less performance than the Low Power Plus process. They are also basically claiming that engineers at IBM are idiots.

It is just about so weird as reading the latest news from WCCFTECH where the author makes the nonsensical claim that the 32C Zen chip for servers has 512MB of L3 cache. Not only everyone knows that this CPU has 64MB of cache, but that the author doesn't even get that 512MB of cache would occupy about one half of the whole motherboard size.

 

[jaymc]

Call me optimistic guy's but I think this Phoenix is gonna rise from the ashes.... with Zen and with Vega which is now looking like it's getting announced at the supercomputer show albeit a version that's for the workstation market and may also be just a paper launch this year.

But they are keeping their cards very close to their chest indeed when it comes to details about Vega itself...

Well ya can't blame me for loving the underdog... In a perfect world AMD would do great on both fronts. But in this world I hope Glofo doesn't give us supply problems.

I think Intel need to release a chip that's redesigned from the ground up like AMD are doing. Well to get their comfortable lead back... maybe they have already done so and have one waiting in the wings to release... (hell maybe 2) maybe they were just milking the market slowly without releasing new chips designs because of no competition.

Anyone any thoughts what's Intel's next fully redesigned chip from the ground up... that wont just be a 5-10% increase in IPC ? An when can we expect to see it ??

J

 

[os2wiz]

Spoiler

One quite important question that I asked you did NOT answer. I had asked when the engineering sample of 2.8GHZ base frequency, mentioned in all the benchmark tests, was produced. Without a date it is impossible to say if it is one of their latest samples. If it is not a very recent sample then it is possible there is more room for improving base frequency.

I just wanted to add 1 more thought. When Kaveri was in development, I distinctly remember similar talk about an engineering sample. I believe it also was 2.8 GHZ. I believe the initial Kaveri release was the 7850 with a 3.5 GHZ clock speed.

The latest engineering samples of Kaveri were clocked at 3.7GHz, which is the same frequency than got final silicon.

We should no far more by the end of October after product starts to ship to the channel.

Zen doesn't start to ship in October. The final silicon is not still ready.

More than one website, including semi-accurate mentioned first processors shipping later in October. Not sure why there all these conflicted reports.

More than one website claimed 2016 launch for Zen, because they got it wrong. More than one website is now claiming that Zen launches in February, when AMD didn't say that.

Anything is possible but when Amd showed Zen they down clock the comparable Intel CPU instead of just picking some Zen chip that could already do 3.2Ghz, if anything that would tell me that 8 core Zen they showed off was probably hand picked already. I seriously can't understand how one could think a 8 core Zen would have a frequency at 3.6Ghz or more when its going to be a 95 watt part. Intel's own superior 14nm with a 8 core Broadwell is only doing 3.2Ghz out of the box with a TDP of 140 watts. At this point in time i don't even see Amd hitting 2.8-3.2Ghz at 95(maybe 125) watts using their 8 core Zen chip on Global foundry's "14nm".

That's not to say their 4 core Zen won't have a clock speed of 3.6Ghz out of the box.

I said at least 3.5 not 3,6 as you said and probably I am wrong, but you are certainly wrong to say they can't produce a 3.2 GHZ 8 core chip when they easily overclocked the 2,8 GHZ sample without water.. What makes Intel 14m better than Samsung's 14nm ?????? You can't make those assumptions since none of you know the date of that engineering sample obviously.

 

[jdwii]

Spoiler

One quite important question that I asked you did NOT answer. I had asked when the engineering sample of 2.8GHZ base frequency, mentioned in all the benchmark tests, was produced. Without a date it is impossible to say if it is one of their latest samples. If it is not a very recent sample then it is possible there is more room for improving base frequency.

I just wanted to add 1 more thought. When Kaveri was in development, I distinctly remember similar talk about an engineering sample. I believe it also was 2.8 GHZ. I believe the initial Kaveri release was the 7850 with a 3.5 GHZ clock speed.

The latest engineering samples of Kaveri were clocked at 3.7GHz, which is the same frequency than got final silicon.

We should no far more by the end of October after product starts to ship to the channel.

Zen doesn't start to ship in October. The final silicon is not still ready.

More than one website, including semi-accurate mentioned first processors shipping later in October. Not sure why there all these conflicted reports.

More than one website claimed 2016 launch for Zen, because they got it wrong. More than one website is now claiming that Zen launches in February, when AMD didn't say that.

Anything is possible but when Amd showed Zen they down clock the comparable Intel CPU instead of just picking some Zen chip that could already do 3.2Ghz, if anything that would tell me that 8 core Zen they showed off was probably hand picked already. I seriously can't understand how one could think a 8 core Zen would have a frequency at 3.6Ghz or more when its going to be a 95 watt part. Intel's own superior 14nm with a 8 core Broadwell is only doing 3.2Ghz out of the box with a TDP of 140 watts. At this point in time i don't even see Amd hitting 2.8-3.2Ghz at 95(maybe 125) watts using their 8 core Zen chip on Global foundry's "14nm".

That's not to say their 4 core Zen won't have a clock speed of 3.6Ghz out of the box.

I said at least 3.5 not 3,6 as you said and probably I am wrong, but you are certainly wrong to say they can't produce a 3.2 GHZ 8 core chip when they easily overclocked the 2,8 GHZ sample without water.. What makes Intel 14m better than Samsung's 14nm ?????? You can't make those assumptions since none of you know the date of that engineering sample obviously.

For one Its obvious that Global foundry has yield issues second Intel can not even get 3.5 or more at stock with their 8 core. Its a bit much to expect Amd to do it. Anything in the universe is possible however its possible Intel gives up and gives everything to Amd but is it likely no. We all could be living in a bubble with a 5 year old alien taking care of us. That simply isn't something i'll assume without proof. I can assure you Amd will be closer to 3.0Ghz then 3.5Ghz stock. To claim Amd can do something Intel can't even do at 95 watts is a major claim something that needs LOTS of evidence to even support the claim. Possibly if Amd had the 8 core at a 165TDP or more i might claim otherwise.

 

[jaymc]

Haha... have to say I love the bubble with the Alien. Think you've solved there mate... maybe we should the Alien about clock speeds 🙂

 

[juanrga]

Call me optimistic guy's but I think this Phoenix is gonna rise from the ashes.... with Zen and with Vega which is now looking like it's getting announced at the supercomputer show albeit a version that's for the workstation market and may also be just a paper launch this year.

But they are keeping their cards very close to their chest indeed when it comes to details about Vega itself...

Well ya can't blame me for loving the underdog... In a perfect world AMD would do great on both fronts. But in this world I hope Glofo doesn't give us supply problems.

I think Intel need to release a chip that's redesigned from the ground up like AMD are doing. Well to get their comfortable lead back... maybe they have already done so and have one waiting in the wings to release... (hell maybe 2) maybe they were just milking the market slowly without releasing new chips designs because of no competition.

Anyone any thoughts what's Intel's next fully redesigned chip from the ground up... that wont just be a 5-10% increase in IPC ? An when can we expect to see it ??

J

Intel (and other companies) always have teams working in the next big thing. A fully new microarchitecture from Intel is expected by 2020 or so, but no one would expect miracles from it (more about this below).

It is not true that Intel doesn't have competition. Intel has a very fiery competition, IBM and Sun on servers, Nvidia on HPC, Apple and Qualcomm on mobile, ARM in microservers,... In fact Intel has lost billions of $ trying to compete in some of those very competitive markets.

The reason why we don't see huge increases in performance on Intel desktop chips is because x86 hit a performance wall. We have known that x86 is a dead-way since decades ago. This is the reason why Intel tried to use the transition from 32bit to 64bit to design a new ISA was scalable. Then AMD messed it all by extending x86 to 64bits and giving support to legacy x86 software. Now both AMD and Intel are cornered around a nonscalable ISA as x86-64.

The only reason why AMD is designing Zen from the ground is because the K11 family was, in the own words of AMD, a "disaster".

Also the only reason why AMD Zen produces a much higher IPC than Excavator is because Excavator is a speed-demon microarchitecture (designed for high clocks at the expense of low IPC) whereas Zen is a brainiac microarchitecture (designed for high IPC at the expense of low clocks).

AMD is obtaining a 40% IPC gain from a (speed-demon --> brainiac) transition.

Further designs (e.g. Zen+) will bring the usual 5--10% gain because this will be an usual (brainiac --> brainiac) transition. Same history about Intel chips. Intel moved from speed-demon to a brainiac designs when left the Pentium 4, since then any transition on Intel CPUs is (brainiac --> brainiac) and the IPC gains are small, and any future design will bring a (brainiac --> brainiac) transitions. This is why no one would expect miracles around 2020 when Intel introduces a fully new microarchitecture. In the next 5 years or so any x86 chip will bring 5--10% gain per gen and frequencies probably will not pass 4.5GHz.

 

[juanrga]

I said at least 3.5 not 3,6 as you said and probably I am wrong, but you are certainly wrong to say they can't produce a 3.2 GHZ 8 core chip when they easily overclocked the 2,8 GHZ sample without water..

You mentioned higher frequencies as 3.7GHz and 3.8GHz here

http://www.tomshardware.co.uk/forum/id-2986517/discussion-amd-zen/page-15.html#18669372

Current samples have 2.8GHz base and 3.2GHz turbo. The sample they overclocked at 3.0GHz was with turbo disabled, which suggests the silicon cannot pass 3.2GHz. Producing a 3.2GHz chip would imply having silicon able to support 3.6GHz for the turbo.

Mentioning watter is rather irrelevant. All recent chips produced by AMD hit a frequency wall and using watter cooling doesn't increase frequencies significantly.

What makes Intel 14m better than Samsung's 14nm ??????

density, yields, max frequency,... also Samsung 14nm and Glofo 14nm aren't exactly the same.

You can't make those assumptions since none of you know the date of that engineering sample obviously.

As explained previously the date is irrelevant, what matters is the silicon used; early lab silicon is not the same than near-production silicon. As explained before the current engineering samples are A0 steeping, i.e. silicon is very close to production silicon.

 

[os2wiz]

I said at least 3.5 not 3,6 as you said and probably I am wrong, but you are certainly wrong to say they can't produce a 3.2 GHZ 8 core chip when they easily overclocked the 2,8 GHZ sample without water..

You mentioned higher frequencies as 3.7GHz and 3.8GHz here

http://www.tomshardware.co.uk/forum/id-2986517/discussion-amd-zen/page-15.html#18669372

Current samples have 2.8GHz base and 3.2GHz turbo. The sample they overclocked at 3.0GHz was with turbo disabled, which suggests the silicon cannot pass 3.2GHz. Producing a 3.2GHz chip would imply having silicon able to support 3.6GHz for the turbo.

Mentioning watter is rather irrelevant. All recent chips produced by AMD hit a frequency wall and using watter cooling doesn't increase frequencies significantly.

What makes Intel 14m better than Samsung's 14nm ??????

density, yields, max frequency,... also Samsung 14nm and Glofo 14nm aren't exactly the same.

You can't make those assumptions since none of you know the date of that engineering sample obviously.

As explained previously the date is irrelevant, what matters is the silicon used; early lab silicon is not the same than near-production silicon. As explained before the current engineering samples are A0 steeping, i.e. silicon is very close to production silicon.

The 14nm process that GLoFis using is licensed from Samsung. So tell me how it is different from Samsung's since it is Samsungs? You have taken my on water remark out of context. I made the point the overclock to 3.2 on that 2.8 engineering sample was made without water . There is no reason they could not issue the silicon at 3.2. Of course that would limit is overclocking capability. I really doubt the cpu would be issued at 2.8 GHZ. It would pyschologically undermine marketing of the product.

 

[jaymc]

I presume Glofo must of made a slight change to the process themselves...

 

[juanrga]

The 14nm process that GLoFis using is licensed from Samsung. So tell me how it is different from Samsung's since it is Samsungs?

Samsung has licensed to Glofo the IP behind the process, nothing more. This is like if I license you a recipe of Galician style Octopus, this doesn't mean you will be able to produce tapas dishes of "Pulpo a feria" so delicious as mine.

It is worth to mention a bit of history. Glofo started developing its own 14nm process, named 14XM. After lots of marketing talks they canceled it because it was outdated even before releasing. Then they licensed the IP from Samsung. Notice that Samsung licensed the 14nm node because wanted Glofo to become a second-source for Apple chips. The original idea was to build Apple chips on Samsung Foundry and then use Glofo for all extra production couldn't give Samsung Foundry. Apple tested chips on both foundries. Chips from Samsung foundry were ok, chips from Glofo were terrible. Apple didn't accept the low quality of the Glofo-produced chips and rescinded the original contract with Samsung, returned to the draw table, redesigned the chips and sent them to TSMC. This is the reason why about half the Apple A9 chips are made by Samsung 14nm whereas the rest are made by TSMC 16nm. Once Glofo lost production of Apple chips and its fab Nº 8 was empty without anything to produce then them and AMD had to renegotiate the WSA which forced AMD to leave Glofo to do the Polaris GPUs.

Of course after this commercial fiasco Samsung has broken relations with Glofo, which has forced Glofo to design by itself the 10nm and 7nm nodes... and history is repeating once again. A pair of days ago Glofo announced cancellation of its 10nm node and jump directly to their 7nm node, which is another disaster.

From better to poor

Samsung 14nm > TSMC 16nm > Glofo 14nm.

You have taken my on water remark out of context. I made the point the overclock to 3.2 on that 2.8 engineering sample was made without water . There is no reason they could not issue the silicon at 3.2. Of course that would limit is overclocking capability. I really doubt the cpu would be issued at 2.8 GHZ. It would pyschologically undermine marketing of the product.

The engineering sample was overclocked to 3.0GHz and with turbo disabled. If it could be overclocked to 3.2GHz then there is no reason why AMD wouldn't do it, instead they underclocked the Intel chip for the Blender demo.

I already explained that last AMD chips hit a frequency wall from the process node choosen and that watter cooling is of little help to hit higher frequencies. The overclocking capabilities of Zen will be very limited. If you want know my predictions about the overclocking frequencies I expect with watter cooling, check this thread.

I really doubt the cpu would be issued at 2.8 GHZ. It would pyschologically undermine marketing of the product.

As stated multiples times in this thread I expect 3.0Ghz for final silicon. Other people is expecting 3.2GHz. No one I know really expect final product to be clocked at 2.8GHz.

Psychology and marketing don't change the laws of physics. If your process node doesn't achieve 4.3GHz then you cannot release a chip with 4GHz base and 4.3GHz turbo, period.

 

[jaymc]

@juanrga 2020 that gives AMD few years to make some mulla !!!

@everyone else: Stop arguing with juanrga, will yis never learn... it's like brigning a knife to a gun fight. Your gonna loose. He didn't achieve his "guru" status for no reason ya know. I'd be very surprised if he's wrong. Come on we all wanna see faster clock speeds but it is what it is... Maybe they can increase them when production ramps up Juan ???

Anyway... here's some gossip an random speculation...maybe even some lies:

http://www.itechpost.com/articles/36082/20161004/amd-zen-cpus-set-ces-2017-launch.htm

http://www.guru3d.com/news-story/amd-prepares-for-zen-clearing-inventory.html

Jay

 

[juanrga]

@juanrga 2020 that gives AMD few years to make some mulla !!!

@everyone else: Stop arguing with juanrga, will yis never learn... it's like brigning a knife to a gun fight. Your gonna loose. He didn't achieve his "guru" status for no reason ya know. I'd be very surprised if he's wrong. Come on we all wanna see faster clock speeds but it is what it is... Maybe they can increase them when production ramps up Juan ???

Anyway... here's some gossip an random speculation...maybe even some lies:

http://www.itechpost.com/articles/36082/20161004/amd-zen-cpus-set-ces-2017-launch.htm

http://www.guru3d.com/news-story/amd-prepares-for-zen-clearing-inventory.html

Maturing a process can help to increase yields and adjust better the node parameters to extract some extra MHz. See that AMD did with Trinity --> Richland or that Intel did with Haswell --> Haswell Refresh.

Regarding the news they are related to my former posts in this thread where I said that a February launch was not confirmed. This is the relevant part of the news:

However, since Zen-based products are unlikely to start mass shipments until after February, they are unlikely to start contributing profits until the second quarter 2017, according to sources from the upstream supply chain.

This means in plain terms a paper launch in February, with Zen being available on stores latter.

 

[jimmysmitty]

Call me optimistic guy's but I think this Phoenix is gonna rise from the ashes.... with Zen and with Vega which is now looking like it's getting announced at the supercomputer show albeit a version that's for the workstation market and may also be just a paper launch this year.

But they are keeping their cards very close to their chest indeed when it comes to details about Vega itself...

Well ya can't blame me for loving the underdog... In a perfect world AMD would do great on both fronts. But in this world I hope Glofo doesn't give us supply problems.

I think Intel need to release a chip that's redesigned from the ground up like AMD are doing. Well to get their comfortable lead back... maybe they have already done so and have one waiting in the wings to release... (hell maybe 2) maybe they were just milking the market slowly without releasing new chips designs because of no competition.

Anyone any thoughts what's Intel's next fully redesigned chip from the ground up... that wont just be a 5-10% increase in IPC ? An when can we expect to see it ??

J

The last two time Intel did that people didn't want it or I failed.

Itanium was a brand new fully 64bit design. People didn't want to lose their x86 performance and make the move and AMD swopped in with the easier and cheaper option so it won out sticking us with x86.

Then we got NetBurst. Not horrible but man was it not good. Core was vastly better in IPC and with 65nm made it a much better power part as well.

I will wait till it is out and we have full details to see how much of it is "all new" and how much of it is re-used. I would not be surprised if they didn't go back to a K8/K10 design and enhance it with new ideas like SMT.

 

[gamerk316]

The problem is now that x86-64 has taken off, we're probably stuck with it for at least two decades. There really isn't a reason to go beyond 64-bit computing, and now that 64-bit OS's are in the majority, there's enough architecture lock-in where simply deciding to move to a new architecture isn't an option.

Itanium was a FAR better solution, but that 20% loss to legacy x86 performance, coupled with AMD creating x86-64 basically killed it in the Desktop space. But we'd be in a far better place right now if x86-64 hadn't come along.