[Overview] Ryzen CPU & AM4 Mainboard Lineup

[lolfail9001]

What exactly is it that you think XFR means or does?

Literally extended frequency range: ability to set arbitrary multipler above some arbitrary value.

 

[Mopetar]

If that's the case, you'd have to believe that this slide is a complete fabrication:

It basically spells out that XFR includes additional boost scaling based on cooling solution. Obviously this implies that the chip is binned well enough to be physically capable of reaching those higher frequencies.

The only way your explanation makes sense is if its just a fancy marketing term for unlocked chip, in which case all Ryzen CPUs have XFR since AMD has already said they'll be unlocked and you can manually OC assuming the motherboard supports it.

 

[lolfail9001]

If that's the case, you'd have to believe that this slide is a complete fabrication:

No, if you interpret them by the letter, they are true about Intel's unlocked multiplier as well. Graph is a trickier subject, but if we assume it actually thermal throttles like that, then it's still all fine.

The only way your explanation makes sense is if its just a fancy marketing term for unlocked chip, in which case all Ryzen CPUs have XFR since AMD has already said they'll be unlocked and you can manually OC assuming the motherboard supports it.

i7-3820 had unlocked multiplier.... but that unlocked multiplier was capped at 42. Get my comparison yet?

Of course, this is the absolutely worst case scenario for what XFR is.

 

[richierich1212]

i7-3820 had unlocked multiplier.... but that unlocked multiplier was capped at 42. Get my comparison yet?

Of course, this is the absolutely worst case scenario for what XFR is.

The i7-3820 is "partially/limited unlocked":

Let's hope Ryzen is fully unlocked and XFR isn't linked to a "limited unlock"

 

[Mopetar]

No, if you interpret them by the letter, they are true about Intel's unlocked multiplier as well. Graph is a trickier subject, but if we assume it actually thermal throttles like that, then it's still all fine.

i7-3820 had unlocked multiplier.... but that unlocked multiplier was capped at 42. Get my comparison yet?

Of course, this is the absolutely worst case scenario for what XFR is.

Your explanation still ignores the main points of that slide which state an automated solution that scales based on cooling solution used. What I interpret this to mean (and seems most logical) is that the listed values for stock and boost are based on their included cooling solution. However, if you use a better air cooler or a water cooler, then XFR will automatically boost beyond the configured frequencies if it detects that the chip is running cool enough.

Yes, you can do all of that manually, but Intel doesn't sell any chips that will automatically do it for you. I don't know why you keep trying to draw that comparison with Intel when the slides make it pretty obvious what it's doing. The only reason to believe XFR works the way you suggest is because the slide is a fabrication.

Personally I don't consider the feature to be at all useful, if only because most people who buy the high-end chips with the functionality enabled are going to do their own manual overclocking and will probably be more aggressive than this functionality would be by itself. I guess it's maybe useful for the idiot that spends $500 on a top-end chip and another hundred or two on a water cooling kit but doesn't overclock it manually, but I would imagine that's a small market. Maybe it's something that could be more useful for notebooks, but even that might be on the niche side.

 

[lolfail9001]

state an automated solution that scales based on cooling solution used.

AMD already has that with Precision boost and their GPUs: what it translates to is automated throttling based on temperature/power draw.
Anyways, i prefer the slightly-better scenario of it simply lifting the stock boost cap on ACT/SCT clocks. Since they would be probably thermals/power draw limited without OC, it just makes it a marketing gimmick, but at least it does not harm lesser SKUs. AMD probably aggressively bins them, anyways, with such pricing scale.

Yes, you can do all of that manually, but Intel doesn't sell any chips that will automatically do it for you.

Intel sells a ton of chips that automatically throttle on power draw caps and have turbo clocks so inflated, lifting the limits on them serves no purpose whatsoever. Matter of fact, so does AMD. And yes, clocks scale automatically.

The only reason to believe XFR works the way you suggest is because the slide is a fabrication.

Fabrication? Nah, more like spin. The looncraz/bjt's descriptions combined paint a much more coherent picture, so i stick to it now.

 

[Mopetar]

You're still missing what the feature is purported to do I think.

Intel (and AMD) CPUs already have turbo boost, and for unlocked chips you can configure this. What XFR supposedly does is detect that the CPU is running cool and automatically exceed the predefined boost limits. If I buy an i7 CPU that's unlocked (e.g. 7700k) and put it under water cooling and do nothing else, it will run at 4.2 GHz and boost up to 4.5 GHz as is advertised. Assuming it had XFR (or equivalent) it would notice that it's running at exceptionally low temperatures and automatically boost above 4.5 GHz without adjusting it manually. That's supposedly what XFR is.

So no, it's not something that Intel has, but I don't think it really matters since it isn't useful to enthusiasts and it's seemingly only something enabled on the more expensive chips.

Fabrication? Nah, more like spin. The looncraz/bjt's descriptions combined paint a much more coherent picture, so i stick to it now.

How is it spin? The slide gives a pretty basic description of the feature and nothing else. Spin would be painting it as some kind of must have or actually useful. It's about the most non-spin slide you could get as it doesn't even try to compare it to Intel's solution or anything else. It sounds like you're trying to say that the slide is real, but what it's saying is somehow untrue. If that's the case, which of the 3 bullet points (or any other part of the slide, not that there's much) isn't true?

I'm not sure what either of those people posted as its not in this thread or why you'd take their word over a pretty basic and easy to understand slide that a few tech sites believe is legitimate enough to write an article about unless they have good reason to believe it's a fabrication, but apparently you don't believe it's a fabrication so I don't even know why you continue to believe what you believe because it doesn't make any logical sense to me.

 

[moonbogg]

It sounds to me like XFR only effects the regular boost function, making it boost higher. If that's the case, most overclockers won't care about it because it won't effect all cores. Most people will end up overclocking on all cores anyway. Business as usual in the overclocking area is what I expect.

I'm also a little worried about the apparent lack of PCI-E lanes. So what if I decide to go SLI? I'm stuck with 8X on each GPU? Maybe less than that?

 

[unseenmorbidity]

It sounds to me like XFR only effects the regular boost function, making it boost higher. If that's the case, most overclockers won't care about it because it won't effect all cores. Most people will end up overclocking on all cores anyway. Business as usual in the overclocking area is what I expect.

I'm also a little worried about the apparent lack of PCI-E lanes. So what if I decide to go SLI? I'm stuck with 8X on each GPU? Maybe less than that?

If it was really smart, and boosted the cores I wanted, then I would leave it alone. For instance, if I am playing WoW, and it only boosts those 2 cores higher than I could boost all the cores, then that would be great.

 

[moonbogg]

If it was really smart, and boosted the cores I wanted, then I would leave it alone. For instance, if I am playing WoW, and it only boosts those 2 cores higher than I could boost all the cores, then that would be great.

I don't trust any chip to be that smart. I feel the need to slap it around and tell it how to behave.

 

[Elixer]

I'm also a little worried about the apparent lack of PCI-E lanes. So what if I decide to go SLI? I'm stuck with 8X on each GPU? Maybe less than that?

Each card would be 8x PCI-E gen 3 slots, and you won't notice a difference on current gen vid cards.

 

[inf64]

Each card would be 8x PCI-E gen 3 slots, and you won't notice a difference on current gen vid cards.

Exactly, the difference would be negligible even on a pair of Titan X cards.

 

[lolfail9001]

What XFR supposedly does is detect that the CPU is running cool and automatically exceed the predefined boost limits.

That's basically what i said i use as current working hypothesis. Except that i do not expect it to push any far, because not at a single point AMD specified whether XFR will work off of default TDP rating or something specific. In which case all it relies on is keeping temperature so low that it actually affects power consumption in meaningful way.

So no, it's not something that Intel has

But effectively it would produce same result as putting 4.8Ghz turbo clock on mobile Intel CPU. Or Carrizo CPU.

Exactly, the difference would be negligible even on a pair of Titan X cards.

That's one thing oc.net have empirically established: on pair of Titan XMs or XPs, it does make a difference on minimums.

 

[moonbogg]

Exactly, the difference would be negligible even on a pair of Titan X cards.

I gained several hundred points in firestrike by enabling 3.0 16x on my board with 980ti's. I know in games it won't be a big deal, but I do feel some performance is being left on the table with only 8x support. I don't actually plan on doing SLI next round, but if I change my mind I'd like to know I am covered. I'll do some more reading and benchmark searching on this topic. This alone wouldn't stop me from going with Ryzen though. I don't think anything can stop me at this point.

 

[Ancalagon44]

Rumour is that Zen doesn't clock that high anyway, like getting above 4.0Ghz is difficult. So, XFR might not really matter that much.

Might have to wait for Zen 2.0 for CPUs that can match Intel's clocking capabilities.

 

[beginner99]

Your interpretation is wrong and your conclusion is wrong, here are the chips sets.
http://hothardware.com/ContentImages/NewsItem/39819/content/big_amd_am4_summary.jpg.ashx?maxwidth=1170&maxheight=1170

its disk IO is 4 SATA off the chipset + 4PCIE gen 3 broken up into whatever the MB maker wants it to be.
Majord covered it here:

http://i.imgur.com/Luc2HEW.png

Can you elaborate on that? As far as I read the diagram, if I choose a B350 board I can get for example a 4xNVMe port (m.2) directly to the SOC and then the chipset offers 2xSATA3. Can the SATA-e ports also be used as normal SATA3 ports? (If mobo maker decides to do that?). If not number of sata ports are pretty low. I mean you need 1 for a bluray drive and 1 for a hdd at least. I currently have 3 HDDs...

 

[Mopetar]

Rumour is that Zen doesn't clock that high anyway, like getting above 4.0Ghz is difficult. So, XFR might not really matter that much.

Might have to wait for Zen 2.0 for CPUs that can match Intel's clocking capabilities.

Could also be a limitation of the process node that AMD is using. Intel has had the best process technology for a long time now and I suspect that if they were making their chips at Global Foundries they wouldn't clock as high either.

Some of the 4-core chips might be able to hit 5 GHz just because there's dark silicon to help spread the heat around, but I'll be surprised if the 8-core chips regularly see about 4.5 GHz.

 

[unseenmorbidity]

Rumour is that Zen doesn't clock that high anyway, like getting above 4.0Ghz is difficult. So, XFR might not really matter that much.

Might have to wait for Zen 2.0 for CPUs that can match Intel's clocking capabilities.

Based on? I have never seen anything about that.

 

[unseenmorbidity]

updated,

https://www.reddit.com/r/Amd/comments/5v4hqt/overview_ryzen_cpu_am4_mainboard_lineup_updated/

 

[StrangerGuy]

Rumour is that Zen doesn't clock that high anyway, like getting above 4.0Ghz is difficult. So, XFR might not really matter that much.

Might have to wait for Zen 2.0 for CPUs that can match Intel's clocking capabilities.

I have zero interest in any of the auto OC schemes from anybody because by the very nature of their operation their freq/V curve needs to be set very conservatively, and/or they boosting pointlessly tiny overclocks in transient loads above sustained 100% load turbo.

I'm leaning more towards the 1700X over the non-X because of the better stock cooler and possibly better binning.

 

[bjt2]

Rumour is that Zen doesn't clock that high anyway, like getting above 4.0Ghz is difficult. So, XFR might not really matter that much.

Might have to wait for Zen 2.0 for CPUs that can match Intel's clocking capabilities.

Zen has already 4.0 GHz default turbo clock. I don't think that at 4.025GHz suddenly it will explode. Actually with XFR it should surpass 4GHz quite well even at default.

 

[itsmydamnation]

Can you elaborate on that? As far as I read the diagram, if I choose a B350 board I can get for example a 4xNVMe port (m.2) directly to the SOC and then the chipset offers 2xSATA3. Can the SATA-e ports also be used as normal SATA3 ports? (If mobo maker decides to do that?). If not number of sata ports are pretty low. I mean you need 1 for a bluray drive and 1 for a hdd at least. I currently have 3 HDDs...

My understanding is the Sata E ports on the chipset can only be 2x SATAe or PCIe v3, but that shouldn't stop a MB making use of those 4x PCI-E3 lanes to go to a chip that has lots o SATA/whatever, or they could use the PCI-e 2.0 lanes to do that if they wanted.

 

[lolfail9001]

I don't think that at 4.025GHz suddenly it will explode.

Well, Broadwell sort of does exactly that, but at 4.2->4.3 jump instead.

Actually with XFR it should surpass 4GHz quite well even at default.

So far we have seen only hints of 100mhz over stock boost.

 

[bjt2]

Well, Broadwell sort of does exactly that, but at 4.2->4.3 jump instead.

So far we have seen only hints of 100mhz over stock boost.

Broadwell starts from 400MHz of base clock lower and with an higher TDP. If we shift the Broadwell behaviour (even ignoring the TDP differences) by 400MHz, probabily we are not so far from reality of Zen OC behaviour...

 

[lolfail9001]

Broadwell starts from 400MHz of base clock lower and with an higher TDP.

You know enough about CPUs to understand that Broadwell WILL have higher TDP. Disable 256-bit AVX and look at tdp then. Going to bet it will be 95W too, with Turbo clock too, making the clocks between 1800X and 6900k virtually identical.

If we shift the Broadwell behaviour (even ignoring the TDP differences) by 400MHz, probabily we are not so far from reality of Zen OC behaviour...

So, 4.2Ghz it is?