*** AMD "Zen" thread (inc AM4/APU discussion) ***

[Silent_Scone]

Yep I understand that but then the dynamic scaling always went from the base clock thus if it was a chip at 3.0GHz then it could dynamic scale (turbo) to say 3.4GHz and then the chip that was set at 3.2GHz could dynamic scale (turbo) to 3.6GHz and so forth.

That is basically showing the base clock is the limiting factor (as it is with Intel) depending on how it is set. It makes even less sense if all chips have a base speed but all can turbo as high as one another. It would also mean an even bigger lottery with people buying 10 cheaper chips, trying them all and then selling off/returning the chips they don't want because they have found a golden nugget and that would just inflate prices.

In regards to the thermal control and the system being similar to this, all that is going to do is that a chip that is too hot will not turbo or dynamic scale to it's maximum but it wont suddenly scale up to 4.2GHz as long as it is in the temp range.

This is only really a problem if not using a decent cooler and to protect the chip from a manual clock if you set the voltage etc higher from my view or maybe for builders like Dell & similar will have more use of this.

The CPU will turbo to whatever AMD deem they want it to...

The ASUS thermal control tool lets you set both light and heavy workload thresholds and voltages. That's what I was alluring to. Anyone with half a brain cell could work out the CPU isn't going to dynamically scale by default to a frequency that isn't stable lol.

 

[Curlyriff]

The scaling isn't fixed to 4 multipliers, it's a quite common choice but the are others too (I.E my laptops i7 has a 2.4 base and 3.5 turbo, my desktops i7 has a 3.4 base and 3.9 turbo).

Yeah sorry I am just using that as an example. My point was though that having something that can turbo to the same level at all chips would be pointless and that a base level from the SKU base frequency would be adopted.

You are comparing a laptop CPU to a Desktop CPU. That is different again. We are talking about all desktip SKU's with Ryzen at this time and ths the multiplier is likely to be locked similar between each step of range rather than be free floating and offering something that turbos by 1.2GHz at the lower scale but only 0.4GHz at the higher scale.

That would give further reason to never offer anything but 3 or 4 SKU's with different core/hyperthread counts and then just let the dynamic scale work on all of them and don't worry about it. It just doesn't make sense financially or with the product in discussion.

They will use a multiplier to give either a 0.3 or 0.4 GHz boost to them. This has already been shown in principle with the SR7 1800X that goes from 3.6GHz base to 4.0GHz turbo.

They will not then the standard 1800 with a turbo that offers 3.2GHz to 4.0GHz boost as that would kill the 1800X.

 

[Curlyriff]

The CPU will turbo to whatever AMD deem they want it to...

The ASUS thermal control tool lets you set both light and heavy workload thresholds and voltages. That's what I was alluring to. Anyone with half a brain cell could work out the CPU isn't going to dynamically scale by default to a frequency that isn't stable lol.

That is fine but we are still talking base terms a max turbo of 0.3 or 0.4GHz.

The principle of that boost requiring different voltages depending on workload wont change the figure otherwise though with what the turbo clock will be, just how it achieves the turbo speed. Which yes is good but not exactly anything an average user would be worried about.

And that still comes from the base clock which you was suggesting is irrelevant but it isn't. The thermal control doesn't allow the turbo to suddenly be go from 4.0GHz to 4.4GHz in that the CPU will already boost as high as it can at a thermal threshold set my Intel in the current case we are talking about. The Asus just allows more manual control because we are likely to have higher end cooling thus more headroom than what Intel spec at.

This wont be any different for AMD and it doesn't need to be other than they are offering users more manual control out the box.

 

[Silent_Scone]

That is fine but we are still talking base terms a max turbo of 0.3 or 0.4GHz.

The principle of that boost requiring different voltages depending on workload wont change the figure otherwise though with what the turbo clock will be, just how it achieves the turbo speed. Which yes is good but not exactly anything an average user would be worried about.

You're still not getting it. The TCT lets you define what these thresholds are. It's possible this will be available on AM4 also. That's what you want to take away from what I'm saying. That and the fact by default, the TDP is controlled much in a similar way to the tool (it's not patented).

What the turbo multiplier is at stock is of no concern to what i'm saying.

 

[ubersonic]

Intel releasing improved i7 and i5 chips to combat Ryzen makes me happy, they will probably have paid money for info on Ryzen and have a much better idea of its performance than us, so if Intel are feeling threatened we should feel happy

 

[Freddie1980]

Intel releasing improved i7 and i5 chips to combat Ryzen makes me happy, they will probably have paid money for info on Ryzen and have a much better idea of its performance than us, so if Intel are feeling threatened we should feel happy

I would be even happier if they just cut prices of their existing lines.

 

[humbug]

Lisa Su said Ryzen would be shipping at 3.4ghz onwards - look at the 8C clockspeeds,they state a base clockspeed of under 3.4GHZ!

The CEO is on record as saying that unless you are now saying she lied. Watch the December AMD event again.

Edit!!

Instead of thread thrashing the thread, it's even stated in the coverage here:

https://www.extremetech.com/computi...e-clock-speeds-additional-performance-details

So even if it were only for the 8C model it is contradicting what the CEO said - she said 3.4GHZ onwards not 3.0GHZ and is even stated in the AMD slides - any other info is not verifiable and really seem like you want to make sure at least the 8C Ryzen chips have low clockspeeds.

Thanks CAT, there were also more than one machine, at least two, one running Battlefront and another running Doom, the one running Battlefront was running 3.4Ghz the one running Doom was at 3.6Ghz.

Linus was there jumping up and down infront of a new Zen / Vega box once or twice a day.

 

[Curlyriff]

You're still not getting it. The TCT lets you define what these thresholds are. It's possible this will be available on AM4 also. That's what you want to take away from what I'm saying. That and the fact by default, the TDP is controlled much in a similar way to the tool (it's not patented).

What the turbo multiplier is at stock is of no concern to what i'm saying.

Maybe not for you but it will be for tens of thousands of others that will never touch the TCT. That is what I am saying.

And I am also suggesting that it will never let you set the threshold greater than what the chip above can do because it would eat into their own product line. This is why I can't see how all being unlocked would work as it would need extreme binning to make sure each one doesn't overlap.

If you put all your chips as unlocked with TCT then it would mean that all chips in theory could reach the fastest speed possible by turbo before thermal throttle and thus make the top end chip less desirable/waste of time because there is no reason to purchase anything that has a faster base clock. So in that theory then yes base & turbo clock become irrelevant.

However we have already got the information showing and stating that isn't the case by showing that there are different base clocks and that the chip will turbo to a set point of 4.0GHz for the R7 1800X.

With that at best what it is saying is that it will turbo to that when in light load and no thermal issue, however if thermals increase then it will dynamically reduce the turbo clock to say 3.9Ghz or 3.8GHz or whatever in the steps, it could be that it does it in 0.1GHz steps and thus more dynamic than current steps are with what Intel does.

Still none of what has been shown or suggested is what you are stating that base & turbo are not relevant. It would make 13 of the 17 SKU's pointless by suggesting that.

 

[Silent_Scone]

Maybe not for you but it will be for tens of thousands of others that will never touch the TCT. That is what I am saying.

And I am also suggesting that it will never let you set the threshold greater than what the chip above can do because it would eat into their own product line. This is why I can't see how all being unlocked would work as it would need extreme binning to make sure each one doesn't overlap.

If you put all your chips as unlocked with TCT then it would mean that all chips in theory could reach the fastest speed possible by turbo before thermal throttle and thus make the top end chip less desirable/waste of time because there is no reason to purchase anything that has a faster base clock. So in that theory then yes base & turbo clock become irrelevant.

However we have already got the information showing and stating that isn't the case by showing that there are different base clocks and that the chip will turbo to a set point of 4.0GHz for the R7 1800X.

With that at best what it is saying is that it will turbo to that when in light load and no thermal issue, however if thermals increase then it will dynamically reduce the turbo clock to say 3.9Ghz or 3.8GHz or whatever in the steps, it could be that it does it in 0.1GHz steps and thus more dynamic than current steps are with what Intel does.

Still none of what has been shown or suggested is what you are stating that base & turbo are not relevant. It would make 13 of the 17 SKU's pointless by suggesting that.

I think you're confused, you're basically saying AMD would disable overclocking as it would defunct their product stack. A lot of assumptions in that post. I don't think AMD would have any issues (assuming the tool works in a similar fashion, and possibly even employed by ASUS if not AMD) with letting users dynamically control TDP thresholds. It makes no odds what the CPU does at stock, or what it doesn't. That's not what the tool is about.
If all you care about is the maximum boost frequency out of the box, then you're making a very long winded read for very little point lol

 

[Curlyriff]

Thanks CAT, there were also more than one machine, at least two, one running Battlefront and another running Doom, the one running Battlefront was running 3.4Ghz the one running Doom was at 3.6Ghz.

Linus was there jumping up and down infront of a new Zen / Vega box once or twice a day.

No sorry but both you & CAT have taken both things out of context here. Even the evidence of the write up on the link you provided states the same as what I am in that there is at least one chip that will run at 3.4GHz base.

Lisa stated that indeed the chips will be at 3.4GHz base and they are. Nothing stated said all will be the case though. And as pointed all the range of Ryzen are showing at least one chip that is 3.4GHz base.

but the company has confirmed that it will offer an eight-core chip with at least a 3.4GHz base clock This is the sentence that confirms what I am saying. That AMD via Lisa confirmed at least one of their eight-core chips will be having a base clock of 3.4GHz.

In regards to what is at CES, that still doesn't change what is shown in the table or the information given. It just means they had two different SKU's on show at those speeds.

No one is wishing them to be slower but that is what has been provided.

 

[Curlyriff]

I think you're confused, you're basically saying AMD would disable overclocking as it would defunct their product stack. A lot of assumptions in that post. I don't think AMD would have any issues (assuming the tool works in a similar fashion, and possibly even employed by ASUS if not AMD) with letting users dynamically control TDP thresholds. It makes no odds what the CPU does at stock, or what it doesn't. That's not what the tool is about.
If all you care about is the maximum boost frequency out of the box, then you're making a very long winded read for very little point lol

Kinda, what I am not understanding from the information is that if you offer all unlocked chips that can all clock much higher than the SKU about it then that SKU becomes irrelevant. It is the reason Intel lock their chips and only offer a selection of unlocked otherwise a lower i7 is what we would all buy and manually clock to the best it can do.

The chips are not binned to overclock and not overclock, they are just locked or not locked when produced by Intel. That is it. If AMD do not lock any down then it is the same chip in every SKU and thus all could possibly clock as high as each other making having more than 4 SKU's irrelevant.

What then is the point in offering the different level SKU if you think they will all be able to use a TCT to do this for every chip and thus also offering manual overclock to all of them.

It would mean the more expensive chip would only be good for people who don't want to overclock a chip and just put it in the motherboard, and for anyone who manually overclocks their chips will buy the cheapest of the SR3/5/7 range they are looking at depending on their desired core count.

That is counter intuitive. And Intel are already disabling overclock with the different SKU's being locked at the alternative speeds.

What someone has suggested though is that all the X chips in each range are unlocked. All the ones that do not have this denotation are still locked chips at the lower speed. Now this would make sense as they are saying that they are offering each core count an unlocked version. But that is not what is stated by anyone previous with people all claiming that all chips are unlocked.

 

[Nutella33]

ISSCC notes point to a 3.4ghz base clock on the 8 core/16 thread designs and process advantages over Intel's current 14nm process.

http://www.eetimes.com/document.asp?doc_id=1331317&page_number=2

 

[humbug]

Looks like a certain company is starting to get rather worried

Is that a FAQ thing on Zen from Intel?

 

[Nutella33]

Is that a FAQ thing on Zen from Intel?

Briefing for front of store retail types.

 

[humbug]

Briefing for front of store retail types.

Lol... yeah that smacks of desperate.

"its ok everyone, Intel still win..... make sure you tell them that!"

Is there a link for this?

 

[Curlyriff]

That's the example I was going off, the 8300 has a lower base than the 6300 which has a lower base than the 4300. Generally AMD use the extra headroom, whereas Intel just ensure the i7 is clocked higher than the comparable i5 model which is clocked higher than the comparable i3 model, etc.

But see the FX-4/6/8 series go the opposite way to that.

The FX-4 4300 is 3.8GHz
The FX-6 6350 is 3.9GHz
The FX-8 9590 is 4.7GHz

Now with the R3/R5/R7 series though we don't know which ones correlate and so there are faster R3's than R5's however overall there is still a faster R5 in the series and so the statement that the 4 core & 6 core will be faster doesn't hold true.

And the above shows that. The point is the slide doesn't show there is going to be a faster model of the R5 than the R7 overall and there isn't going to be a faster model of the R3 than the R5 overall just because the core count decreases. So the slide is still correct in principle I feel.

The only thing I can see is that they have a locked and unlock at highest base clock or that the 1700 series is the highest we have seen and they have not firmed up but expect the 1800 series to come in with higher base clocks later.

 

[humbug]

I just spotted something...
-------------

Q: RyZen Processors are using the same 14nm Process as Intel, Does that mean AMD Processors are as good as Intel Processors?

A: Keep in mind that this is the first time AMD has created a 14nm product.................

Nope! Wrong, Polaris GPU's are 14nm.

 

[d_brennen]

That's the example I was going off, the 8300 has a lower base than the 6300 which has a lower base than the 4300. Generally AMD use the extra headroom, whereas Intel just ensure the i7 is clocked higher than the comparable i5 model which is clocked higher than the comparable i3 model, etc.

Depends though, all the launch CPUs with less than 8 cores are likely (not sure if confirmed but strongly hinted) derived from 8 core dies with defects. Defects can come in a variety of forms, from show stoppers which ruin a core/several cores to leaky transistors which can't make the desired base/turbo switching frequencies within the given voltage/thermal window.

With the latter you have two choices, up the power to get the clock speed or drop it to a lower SKU. Quad and hex core parts from the initial spin most likely do clock lower than a full 8 core part for that reason.

It will get more interesting if AMD ever does release a quad core specific die, but then again they have that coming in the form of APUs later in the year/early 2018 so may not bother, doubling down on the moar cores mantra. At the end of the day we are talking about ~200mm2 of silicon. That Intel have charged anything between $400-1100 for that size of chip is a byproduct of no competition more than anything. AMD have been selling die bigger than that for ever at vastly lower prices (nVidia too)

AMD Radeon RX 480 AMD Radeon R9 380X

Die Size 232mm2 366mm2

FX8350
Die Size: 315 mm²

 

[Nutella33]

Lol... yeah that smacks of desperate.

"its ok everyone, Intel still win..... make sure you tell them that!"

Is there a link for this?

From their internal "retail edge" program apparently.

 

[Armageus]

I just spotted something...
-------------

Q: RyZen Processors are using the same 14nm Process as Intel, Does that mean AMD Processors are as good as Intel Processors?

A: Keep in mind that this is the first time AMD has created a 14nm product.................

Nope! Wrong, Polaris GPU's are 14nm.

Surprised they didn't go with a cheap shot about AMD not even fabbing their own 14nm products

It will get more interesting if AMD ever does release a quad core specific die, but then again they have that coming in the form of APUs later in the year/early 2018 so may not bother, doubling down on the moar cores mantra.

Haven't heard that in a while