Speculation on Ryzen Overclocking

[CatMerc]

Wording is ambiguous. In fact "enable higher clock speeds" is actually... reaffirming my suspicion.

Damn, i need to have a meal and relax, that starts to worry me.

What is your suspicion?

 

[Majcric]

Ballpark figure of around 4.5. I just hope the scaling is there.

Would really like to see it go higher to help offset any ipc advantage Intel may have.

 

[lightmanek]

Hello everyone! I've been following a lot of discussions on this forum for quite a while (since XS went almost dead) and finally decided to register!

Couple of things I want to throw into discussion:
1. Assuming AMD will be outsourcing manufacturing of some Ryzen to Samsung, can they differentiate dies from different and supposedly a bit better process with letter X?
2. Is XFR really allowing stock CPU go above 95W TDP?
3. If not allowed to go above 95W TDP, then maybe XFR utilizes better cooling and lower thermals to hit advertised processor speeds like 1800X 3.6-4.0GHz?

I'm personally thorn between 1st (unlikely) and 3 (very likely). Pure speculation pleased

 

[The Stilt]

1. Assuming AMD will be outsourcing manufacturing of some Ryzen to Samsung, can they differentiate dies from different and supposedly a bit better process with letter X?

There are no guarantees that the silicon produced by Samsung at their own plants would generally be any better in terms of the quality than the one produced by GlobalFoundries. Despite the fact that Samsung originally developed the process. It is entirely possible, but far from certain.

 

[lolfail9001]

I wouldn't say that XFR carries any premium. It's the higher clocks that do, same as with Intel.

Fair enough... but the same argument works for K CPUs.

Some ppl here seem to be on a mission.

As the saying goes, it's all too good to be true, plus there are hints that AMD's marketing as usually fed the overhype machine.

What is your suspicion?

3930k has extended frequency range when compared to 3820.

 

[CatMerc]

3930k has extended frequency range when compared to 3820.

If I'm getting you right, that would be my worst case scenario for what XFR is lol

 

[bjt2]

Yep, but since i do not have access to IEEE article (did not bother with it yet), i have to ask. Is the part where conclusion made on turning these power savings into increased frequency... speculation or mentioned in the article itself? Because if it is the former then entire article describes power saving tech in Carrizo/Bristol/Mullins and then proceeds to make a jump inverting their purpose to increase frequency. Make no mistake, it is certainly possible to do. But the wording of the slide and Lisa's words definitely do not state outright that clock *increases*. Just that it can be increased.

Because XFR carries a hefty premium tag with it. Got to figure out why.
Inb4 AMD next leveled everyone and put a premium tag on barely useful feature.

Quoting the paper numbers:

For reliability tracker: "Thus, by dynamically monitoring the processor cores voltage and temperature, the overwhelming majority of users can realize an extra 100 MHz increase in FMAX while staying within long-term reliability targets, by allowing lower operating temperatures to enabled higher VMAX. In addition, even heavy users can realize an extra frequency boost with customized product cooling solutions. This increase is on top of the 4%–9% improvement in VMAX as afforded by the use of the static usage model."

On Digital LDO: "Summarizing, Pmin can be used to provide similar power savings as conventional CC6 but faster exit latency and cache/core state retention, reducing idle transition latency by removing the need to flush the cache. The DVR system enables maximum residency in Pmin, raising the performance of lightly threaded workloads, since more time is spent in the C-state boost state, resulting in net performance gain of ∼6% in these scenarios."

For BTC: "As a result, we can reduce the explicit aging guard band resulting in ∼20 mV of additional savings."

For shadow p-states: "In BR, shadow P-States enable peak boost frequencies, on average, to increase by 100 MHz over conservative traditional binning."

For STAPM: "Fig. 17 shows how the benefits from STAPM varies across workloads. We included measured data from the CineBench, 3-DMark, and PCMark suites, and have demonstrated energy savings in the range of 5%–13% made possible by STAPM. Fig. 18 outlines the underlying principle of STAPM. Essentially, by allowing the core(s) to run at boost frequencies for short periods of time, we can reduce the “time to completion,” The additional power from running the CPU faster is easily compensated, since it enables other system and SoC components to be put into low power sleep modes sooner. Thus, net platform energy savings are obtained overall from enabling STAPM."

 

[lolfail9001]

If I'm getting you right, that would be my worst case scenario for what XFR is lol

I am the glass 90% empty type of guy.

Quoting the paper numbers:

Thanks, i can see how that may be used, should it all be improved further, to enable some sort of larger and invisible at glance turbo clocks.

 

[JDG1980]

3930k has extended frequency range when compared to 3820.

The official slide deck makes it clear that XFR means more than just an unlocked multiplier. It will allow automatic, realtime overclocking beyond normal limits with "no user intervention required".

AMD, like all companies, has often presented their products in the most favorable light in their marketing materials. But they generally do not lie outright about what features are or will be included.

 

[lolfail9001]

The official slide deck makes it clear that XFR means more than just an unlocked multiplier

Permit -- to consent to expressly or formally/to give leave/to make possible/to give an opportunity

No, it does not quite tell us of automatic, real-time overclocking. Second part is non-sequitur because we already have that in form of manual OCs. The third part could suggest that it does oc automatically, or that the allowed multiplier is scaling automatically... that still can be interpreted in bad way, as you see.

But they generally do not lie outright about what features are or will be included.

Of course, but the best way about it are vague statements. And well, statements are indeed vague except the third part.

Now, i am probably wrong anyways, but not for a minute i believe AMD can launch a product, even a great one, without marketing blunder in recent years.

 

[.vodka]

Non XFR CPUs will boost up to maximum turbo boost described. For example R7 1700 will boost up to 3.7GHz and not a single MHz more, in 25MHz increments as described in the precision boost slide

If I'm interpreting the XFR slide right, well, its "extended frequency range". This applies to turbo. XFR enabled CPUs can and will boost further than the maximum turbo boost clock for these parts, the Ryzen die has an insane amount of monitoring going on plus who knows what else more sorcery that AMD has developed all these years of being limited to 28nm processes to make their APUs power efficient, I don't see how all of this can't be used the opposite way, to keep boosting up.

How does Ryzen boost work? Is it single core boost? Is it all core boost? Is there tiered boost as in single core, dual core, quad core boost? How much will XFR let boost go above advertised boost clocks? How far does XFR go with stock cooling, high end aircooling, custom loops and LN2? Can this be called "automatic overclocking" as many of us have described it since the slides were known?

Wait for reviews, February 28th is only 9 days away.


If it's another of AMD's fine marketing blunders like Fiji being an overclocker's dream, then you can save the extra money that AMD is asking for the XFR parts, and overclock manually to your heart's content as all Ryzen CPUs are multiplier unlocked. R7 1700 is $320, R7 1700x is $390. If clockspeed and XFR are the only features that differ, and XFR is a gimmick, then those extra $70 can go to a beefier motherboard with a decent VRM, for example.

 

[JDG1980]

It should be pointed out that Nvidia has been doing the equivalent trick on their video cards since at least Maxwell. The actual boost clocks observed are often beyond the "official" boost level, even for reference cards.

 

[bjt2]

Non XFR CPUs will boost up to maximum turbo boost described. For example R7 1700 will boost up to 3.7GHz and not a single MHz more, in 25MHz increments as described in the precision boost slide

If I'm interpreting the XFR slide right, well, its "extended frequency range". This applies to turbo. XFR enabled CPUs can and will boost further than the maximum turbo boost clock for these parts, the Ryzen die has an insane amount of monitoring going on plus who knows what else more sorcery that AMD has developed all these years of being limited to 28nm processes to make their APUs power efficient, I don't see how all of this can't be used the opposite way, to keep boosting up.

How does Ryzen boost work? Is it single core boost? Is it all core boost? Is there tiered boost as in single core, dual core, quad core boost? How much will XFR let boost go above advertised boost clocks? How far does XFR go with stock cooling, high end aircooling, custom loops and LN2?

Wait for reviews, February 28th is only 9 days away.


If it's another of AMD's fine marketing blunders like Fiji being an overclocker's dream, then you can save the extra money that AMD is asking for the XFR parts, and overclock manually to your heart's content as all Ryzen CPUs are multiplier unlocked. R7 1700 is $320, R7 1700x is $390. If clockspeed and XFR are the only features that differ, and XFR is a gimmick, then those extra $70 can go to a beefier motherboard with a decent VRM, for example.

Since BR shadow p-states allow clocking over base clock indipendently of the load if the conditions are met, so probabily also XFR is not bound to load...

 

[The Stilt]

The official slide deck makes it clear that XFR means more than just an unlocked multiplier. It will allow automatic, realtime overclocking beyond normal limits with "no user intervention required".

AMD, like all companies, has often presented their products in the most favorable light in their marketing materials. But they generally do not lie outright about what features are or will be included.

What do you think the red line represents?

 

[lolfail9001]

What do you think the red line represents?

Supposedly it should be the single core turbo, shouldn't it?

Spoiler

And if we are to read as deeply as possible, then usage of polyline clearly represents 25Mhz steps of Precision boost. Well, now i want this to be the exact truth just for laughs. Probably won't be.

 

[.vodka]

It should be pointed out that Nvidia has been doing the equivalent trick on their video cards since at least Maxwell. The actual boost clocks observed are often beyond the "official" boost level, even for reference cards.

Yes. I also think this is AMD's equivalent of GPU boost 3.0 applied to Ryzen. Non XFR parts have capped turbo, XFR parts have uncapped/unlimited turbo just like nV's GPU boost 3.0, limited by temperature and other parameters

From AT's review of the 1080

If it turns out to be this way, then I'd expect this technology to be available on Vega or later revisions of GCN as a new form of Powertune, too. But that's something for another thread. Oh, now that I think about it this is GDC, maybe we'll also get some Vega news.

Since BR shadow p-states allow clocking over base clock indipendently of the load if the conditions are met, so probabily also XFR is not bound to load...

Yes. I think this has more to do with temperature and keeping the entire chip at maximum clock speeds bound by temperature rather than load. Lisa said it herself in the new horizon live stream, "it clocks higher the cooler it gets".

edit: Well, isn't that the default behavior for every chip out there? Lower temperatures, higher clocks, lower voltage required, less power lost to leakage... Hmm...

Come on, why can't it be the 28th tomorrow?

 

[Erenhardt]

Anything above 4.3Ghz would be way above my expectations.

4.3Ghz+ with the claimed IPC would make Ryzen CPUs amazing value for all DYI market, where overclocking is the most common.

 

[Atari2600]

Now, i am probably wrong anyways, but not for a minute i believe AMD can launch a product, even a great one, without marketing blunder in recent years.

I guess the difference here is that the few tidbits of info regarding XFR have come from someone who has a phd in architecture design, they also happen to be CEO of the company that designed the chip....

So no marketeers with MBAs "filtering" the info!

 

[lolfail9001]

I guess the difference here is that the few tidbits of info regarding XFR have come from someone who has a phd in architecture design, they also happen to be CEO of the company that designed the chip....

Remember Raja, Crossfire scaling and "efficiency" at Polaris announcement at Computex? After that i lost any and all faith in regards to

So no marketeers with MBAs "filtering" the info!

Even when the speaker knows what he is talking about.

 

[CatMerc]

What do you think the red line represents?

All core turbo?

 

[Dresdenboy]

All core turbo?

It's written there, right in the middle.

 

[LTC8K6]

"Chart for illustrative purposes only"

So nothing on the chart is actually representative of anything. It's just a generic illustration.

 

[CatMerc]

It's written there, right in the middle.

I know, just trying to decipher what The Stilt is trying to say lol

 

[Atari2600]

What do you think the red line represents?

The 4.0 in a 3.6/4.0 moniker shirely...

 

[riggnix]

Anything above 4.3Ghz would be way above my expectations.

4.3Ghz+ with the claimed IPC would make Ryzen CPUs amazing value for all DYI market, where overclocking is the most common.

Only 300 MHz above the turbo clock would be really disappointing IMO. If XFR really increases max turbo clock (we'll see about that) it would be pretty pointless for 300 MHz.
I expect at least 4,5 GHz on air. Unless there is some bug preventing higher clocks. I think from a TDP point of view it should be more than possible.

Of course it depends on the meaning of "on air". I'm not talking stock cooler, I'm talking high end air cooling (running an FX-8350 @ 5 GHz on air )
4,5 GHz with stock cooling would be pretty nice actually.