AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[swilli89]

yea, that was real informative.

I'm sorry.. are jokes and good cheer not allowed according to you?

 

[KTE]

There are not advertised boost clocks. The top turbo (4GHz for the top model) is guaranteed with 32C of ambient and stock cooler. If you don't have air conditioner, then you are right. But with 25/26C (sane summer air conditioning setting) XFR should kick in even with base cooler...

What does XFR change the VID to?

How does it decide this?

Where is the predefined table for Clocks@VID kept?

Is this editable?

Is there a max power cap for XFR?

What's the time skew for max power/temperature after being hit?

Is XFR response based on Hotspots priority first?

What's next on priority?

Plenty of questions...

Sent from HTC 10
(Opinions are own)

 

[PPB]

XFR is better than that type of OC. Is a continuos self calibration of p-states, based on temperature and load. No other OC mean can beat it, at same TDP.

I never talked about oc. I just said i would rather have pstate control than uncontrolable "infinite boost" a la nv.

Obviously the latter will give yo more bench pr like nvidia got with their 680 in their first itineration. Just state a silly clockspeed and let the actual core clock kick in until thermals get high enough, and see the benchmarks fly. Core m also do this in a limited fashion (no unlimited part here)

Sent from my XT1040 using Tapatalk

 

[Agent-47]

What does XFR change the VID to?

How does it decide this?

Where is the predefined table for Clocks@VID kept?

Is this editable?

Is there a max power cap for XFR?

What's the time skew for max power/temperature after being hit?

Is XFR response based on Hotspots priority first?

What's next on priority?

Plenty of questions...

Sent from HTC 10
(Opinions are own)

I think there are answers to few of them.
1. XFR would probably increase VID to maintain stability.
2. AMD said based on load on per core basis.
3/4. There is no table. Its calculated on the fly.
5. Every silicon has a max voltage. Its probably that as long as temps and vrm are good enough.
6. I think amd said its 100 ms.
7/8. No idea.

To add more questions:
How does the CPU know if the vrm is good enough to supply enough power without burning out the VRMs.

How did they account for scenarios where threads get bounched from core to core.

 

[DrMrLordX]

Elaborate. In the worst case you just end up stuck on what you've had.

Other responses have more than sufficed, but really any time you wind up paying too much for too little, you lose. The worst example? Probably whoever purchased one of those Cyrix 5x86 chips that burned itself out at stock settings using stock cooling.

Expediting desktop Cannonlake probably isn't feasible, because from what we know so far, Intel's 10nm process is a disaster. It might be more reasonable to expedite Coffee Lake, which (leaks indicate) is basically just Kaby Lake with 6 cores instead of 4. The problem is that Intel is not at all an agile company; something that would take Apple 6 months would take Intel 2 years, or more.

I don't expect expedited anything, to be perfectly honest. Maybe some tweaked prices here and there.

Whether they ship it or not has nothing to do with what they should or shouldn't do. They have major problems with the 10nm process, and even if they ship CL this year, it's already planned to be exclusively very LP mobile chips.

What I meant is, Intel is on pace to ship Cannonlake this year. It's allegedly going to be another Broadwell-like release since - as you say - the process is insufficiently mature for anything but low-power operation. It will be awhile until we see Icelake or anything of its ilk. 2018 would be my guess. It's going to be a bit odd. Also, with AMD's prices being as low as they are, it's really impossible for Intel to counter with their HEDT lineup without significant brand devaluation. Say goodbye to those ph4t margins, Intel.

 

[PPB]

I'm sorry.. are jokes and good cheer not allowed according to you?

Its a very serious time for him. For a lot of people in this forum actually.

Or are you the one that laugh at bubbles being burst in your face?

 

[hotstocks]

Enough with the hype. We need real overclocked speeds and benchmarks. I (granted am mostly a gamer, but heavy multi-tasker) have not seen a reason to upgrade my 6 year old 4.7ghz i7-2600k yet, unbelievable and sad but true. Now Ryzen has my interest, but ONLY if single thread IPC = Haswell or better AND stable 24/7 overclock of 4.7ghz or above is also possible on air or closed water loop. If not there is still no reason to upgrade unless your work demands a lot of slower cores.

 

[Mockingbird]

Enough with the hype. We need real overclocked speeds and benchmarks. I (granted am mostly a gamer, but heavy multi-tasker) have not seen a reason to upgrade my 6 year old 4.7ghz i7-2600k yet, unbelievable and sad but true. Now Ryzen has my interest, but ONLY if single thread IPC = Haswell or better AND stable 24/7 overclock of 4.7ghz or above is also possible on air or closed water loop. If not there is still no reason to upgrade unless your work demands a lot of slower cores.

Well, you can go home because an overclock of 4.7 GHz is probably out of the question.

 

[wembley]

Enough with the hype. We need real overclocked speeds and benchmarks. I (granted am mostly a gamer, but heavy multi-tasker) have not seen a reason to upgrade my 6 year old 4.7ghz i7-2600k yet, unbelievable and sad but true. Now Ryzen has my interest, but ONLY if single thread IPC = Haswell or better AND stable 24/7 overclock of 4.7ghz or above is also possible on air or closed water loop. If not there is still no reason to upgrade unless your work demands a lot of slower cores.

I also want that. And for free, with the mother board too

Sent from my HTC One_M8 using Tapatalk

 

[zinfamous]

Enough with the hype. We need real overclocked speeds and benchmarks. I (granted am mostly a gamer, but heavy multi-tasker) have not seen a reason to upgrade my 6 year old 4.7ghz i7-2600k yet, unbelievable and sad but true. Now Ryzen has my interest, but ONLY if single thread IPC = Haswell or better AND stable 24/7 overclock of 4.7ghz or above is also possible on air or closed water loop. If not there is still no reason to upgrade unless your work demands a lot of slower cores.

For a 1.4 post-per-year rate, you could have spent this precious time sounding less desperate and unreasonable, my man.

 

[itsmydamnation]

Well, you can go home because an overclock of 4.7 GHz is probably out of the question.

I think 4.5 is the magic number, I'm pretty hopeful it can hit it

 

[thedavexp]

I think 4.5 is the magic number, I'm pretty hopeful it can hit it

4.5 on an 8 core would be rather fantastic

 

[hotstocks]

I wouldn't mind paying $600 for the processor and $600 for a good mobo/ram, but I won't pay Intel $1000+ for a processor. I own multiple Intel, AMD, and Nvidia parts. My point is if Ryzen can't beat a 6 year old Intel processor in IPC OR Ghz, then it is going to be a loser even at $400. Why release something pretending to compete with Intel 8 core chips favorably if they really can't except in one canned benchmark? I WANT to upgrade, but both Intel and AMD have made that a non-reality over the last 6 years of releasing 5% faster chips/year, that is not significant. Can you imagine using a 6 year old graphics card (intel or amd), lol. Maybe I am hopefully optimistic because it is not hard to cool a .9-1v 95W cpu AND AMD says the turbo will be unlimited only depending on cooling. So I can see people with a good air or water cooler designed for a 140W Intel cpu putting Ryzen at 1.1, 1.2, or 1.3v and maybe hitting 4.5Ghz or higher. Guess we will see soon enough.

 

[CentroX]

4.5 on an 8 core would be rather fantastic

how high can 6900K hit?

 

[Justinus]

how high can 6900K hit?

I think not quite 4.5. My 5960x has no issue at 4.5 on an AIO loop, though, and if I ever get around to the custom water loop I want to build I think it will make 4.75 without too much trouble.

 

[snarfbot]

I would be surprised if it doesn't do 4.7 at least. I was kinda hoping for 5ghz..

 

[CentroX]

I would be surprised if it doesn't do 4.7 at least. I was kinda hoping for 5ghz..

reminds me of fury x fiasco. AMD: oh this cooling is great for overclocking! turned out you couldnt OC shit.

 

[.vodka]

Enough with the hype. We need real overclocked speeds and benchmarks. I (granted am mostly a gamer, but heavy multi-tasker) have not seen a reason to upgrade my 6 year old 4.7ghz i7-2600k yet, unbelievable and sad but true. Now Ryzen has my interest, but ONLY if single thread IPC = Haswell or better AND stable 24/7 overclock of 4.7ghz or above is also possible on air or closed water loop. If not there is still no reason to upgrade unless your work demands a lot of slower cores.

Wait 2 weeks for reviews or go get a 7700k, delid + apply liquid metal TIM, there you have your easy >5GHz Sandy OC capability + healthy IPC increase over Sandy

I would be surprised if it doesn't do 4.7 at least. I was kinda hoping for 5ghz..

Single core with downright suicidal voltages and beefy custom WCing, probably. CanardPC didn't put that ZenOC@Air=5G easter egg for nothing in their magazine. As reported by their source, it was a single core out of 8 at an unspecified voltage on a flaky VRM equipped motherboard and that was December. There's been a lot of tweaking and refining these past three months until launch. This could've changed for the better regarding all core OC.

24/7 on all 8 cores with reasonable voltages, you're probably looking at 4.4/4.5GHz tops as hinted by buildzoid a while ago (he said X.X GHz which can be interpreted as 4.4GHz). This is no delidded+metal TIM'd 7700k doing >5GHz with ease.

Even then, AMD/Glofo processes are known to be tweaked with time. Early Phenom IIs were dogs overclocking vs later revisions, same for the first FX8350s out of the fab vs later revisions like the FX8370e, in voltage required and other stuff.


At least we know 8c16t @ 4GHz is possible. Question is, how far it'll go on all eight from that and how much will XFR do by itself. 2 weeks to go.

 

[coercitiv]

I wonder if people who casually talk about 4.7Ghz overclocks on 8c/16t Zen realize they're gonna have to handle a 200W+ CPU.

 

[bjt2]

What does XFR change the VID to?

How does it decide this?

Where is the predefined table for Clocks@VID kept?

Is this editable?

Is there a max power cap for XFR?

What's the time skew for max power/temperature after being hit?

Is XFR response based on Hotspots priority first?

What's next on priority?

Plenty of questions...

Sent from HTC 10
(Opinions are own)

I answered it many times. The famous papers on ieeexplorer. I should save in the bookmarks the link to the post instead of writing it dozen of times...

AVFS construct periodically (probabily more than 100 times a second) a table of optimal frequency/voltage based on statistics collected on the 1300 replica circuits. Then calculates the Vmax applicable to the CPU based on the temperature and the CPU usage: there is a function called FIT that estimate the failure probability based on those two parameters and the Vmax is defined as the maximum voltage that let this FIT function be under a threshold of dependability.

Given power draw, temperature, this calculated Vmax and the frequency/voltage table calculated below, the CPU increase its frequency (and set the voltage accordingly to the table) if the power draw is under the limit, the temperature is under the limit and the required Vcore for this frequency is under the calculated Vmax. If the situation changes and the frequency is not sustainable anymore (one on more violations of those rules), the frequency will be lowered.

I suspect that TDP and temperature limits can be configurable and a Vmax offset (at your own risk) can be specified...

Let's see how many times further i must write these information...

EDIT: for the time skew: on Bristol Ridge the STAPM uses a filtered power limit, with an exponential mean, and acknowledge and higher boost state if the FILTERED power is under the limit. So this could also be the case for XFR... We will see...

 

[Lovec1990]

I wonder if people who casually talk about 4.7Ghz overclocks on 8c/16t Zen realize they're gonna have to handle a 200W+ CPU.

i belive they are aware of it plus if my theory about XFR is right everyone who will want too use this function fully will have too buy stronger PSU than needed

 

[bjt2]

I think there are answers to few of them.
1. XFR would probably increase VID to maintain stability.
2. AMD said based on load on per core basis.
3/4. There is no table. Its calculated on the fly.
5. Every silicon has a max voltage. Its probably that as long as temps and vrm are good enough.
6. I think amd said its 100 ms.
7/8. No idea.

To add more questions:
How does the CPU know if the vrm is good enough to supply enough power without burning out the VRMs.

How did they account for scenarios where threads get bounched from core to core.

The VRM probabily have a limit. Anyway since carrizo the CPUs have a voltage droop protection, that kicks in. The full paper specifies also that this voltage droop protection is in use even on the Pmin state, with voltages as low as 0.7V, to stretch the clock if there is a surge. There are plenty of information in this paper on the stability issues. The LDO that repurpose the power gating PMOSes to do a digital VRM to lower the Vcore indipendently core for core, and also have a sort of protection on Vdroops, and so on...

 

[bjt2]

I wouldn't mind paying $600 for the processor and $600 for a good mobo/ram, but I won't pay Intel $1000+ for a processor. I own multiple Intel, AMD, and Nvidia parts. My point is if Ryzen can't beat a 6 year old Intel processor in IPC OR Ghz, then it is going to be a loser even at $400. Why release something pretending to compete with Intel 8 core chips favorably if they really can't except in one canned benchmark? I WANT to upgrade, but both Intel and AMD have made that a non-reality over the last 6 years of releasing 5% faster chips/year, that is not significant. Can you imagine using a 6 year old graphics card (intel or amd), lol. Maybe I am hopefully optimistic because it is not hard to cool a .9-1v 95W cpu AND AMD says the turbo will be unlimited only depending on cooling. So I can see people with a good air or water cooler designed for a 140W Intel cpu putting Ryzen at 1.1, 1.2, or 1.3v and maybe hitting 4.5Ghz or higher. Guess we will see soon enough.

XFR will raise the Vmax if the temperature is low (good cooler), automatically.

 

[bjt2]

i belive they are aware of it plus if my theory about XFR is right everyone who will want too use this function fully will have too buy stronger PSU than needed

There is voltage droop protection: if the Vcore is noisy as shit, due to low quality VRMs or PSU, the performance penality is under 1% and the stability is pretty much guaranteed (if there is not too much droop) since there is BTC: boot time calibration. Even if the Vcore were off tolerance, this calibration can offset this problem...

 

[Lovec1990]

There is voltage droop protection: if the Vcore is noisy as shit, due to low quality VRMs or PSU, the performance penality is under 1% and the stability is pretty much guaranteed (if there is not too much droop) since there is BTC: boot time calibration. Even if the Vcore were off tolerance, this calibration can offset this problem...

sorry but can you explain this boot time calibration?