• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Speculation on Ryzen Overclocking

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I'm guessing 100-200mhz additional with XFR.

It's interesting to speculate about that blue line, which also goes below the red line, suggesting that XFR can also hold clock speeds below PBR.
Presumably if the cooling is poor? Or other conditions are poor?
 
LTC8K6 said:
I'm guessing 100-200mhz additional with XFR.

It's interesting to speculate about that blue line, which also goes below the red line, suggesting that XFR can also hold clock speeds below PBR.
Presumably if the cooling is poor? Or other conditions are poor?
Click to expand...
I believe at that point it wouldn't be XFR, just normal Turbo range.
 
bjt2 said:
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."
Click to expand...


I quote myself, because i was re-reading the paper on bristol ridge, in particular the paragraph of the shadow p-states. I didn't reputed this possible, because otherwise the (few) reviews should have found it, but it seems that XFR is present also in bristol ridge. I report below the full paragraph on the shadow p-states of the paper, with a note...
Read it and tell me what you think...



SECTION V.
Shadow P-States
A critical path accumulator-based scheme to accurately assess true Si speed capability and address the problem of voltage margin reduction in traditional binning flows was briefly reported in [10 NDR: the Carrizo paper, where CPI is described]. While this original implementation was aimed at optimizing the per-part voltage required for target P-State frequency, in the BR implementation of AVFS, we use shadow Pstate to increase peak frequency on part-by-part basis directly when headroom is available. The peak Fmax of the product is generally limited by technology (Vmax) or by infrastructure (EDC) limits. In traditional power binning flow, each part frequency capability in the system is not precisely known. As a result, the peak frequency is set conservatively to a worst case value that can be met by target distribution of parts. Instead AVFS allows us to exactly characterize the part-specific Fmax capability, and BTC (NDR: Boot Time Calibration) allows us to characterize the platform-specific power delivery margin. So instead of restricting to worst case power binned Fmax, at boot-time, we combine AVFS and BTC (NDR: Boot Time Calibration) to build a frequency–voltage curve for a given part in a given platform. By solving along this curve, we can find the peak feasible frequency for this part under reliability Vmax constraints. Similarly, by combining the per-part unique leakage and active power fuses with the frequency–voltage curve, we can determine the highest feasible frequency that meets the regulator supply current specifications. We refer to these peak boost frequencies that meet the infrastructure limits (electrical design current and process Vmax), as shadow P-States. In BR, shadow P-States enable peak boost frequencies, on average, to increase by 100 MHz over conservative traditional binning.
 
It seems that it adjusts the Fmax also based on VRM and motherboard (and maybe PSU) capabilities... So probabily also XFR...
 
Moreover, from the description of shadow p-states, the frequency voltage table and maximum frequency, is computed at boot, supposedly with the coolest CPU. This is almost true, except when you use LN2 or helium or phase change. So it seems that for maximum frequency, you should turn on (or pour) special cooling BEFORE turning on the system...
 
bjt2 said:
Moreover, from the description of shadow p-states, the frequency voltage table and maximum frequency, is computed at boot, supposedly with the coolest CPU. This is almost true, except when you use LN2 or helium or phase change. So it seems that for maximum frequency, you should turn on (or pour) special cooling BEFORE turning on the system...
Click to expand...
So what happens when you suddenly heat up your LN2 chilled cpu? Seems like that might be dangerous to start the CPU up after it is cooled to LN2 like temps?
 
I'm not clear on the intent of XFR. Is it to get higher ST performance with one core boosting above Precision Boost?
 
LTC8K6 said:
So what happens when you suddenly heat up your LN2 chilled cpu? Seems like that might be dangerous to start the CPU up after it is cooled to LN2 like temps?
Click to expand...

It calculates a voltage frequency temperature table, but i suppose that it will compensate only for higher thatn boot temperature. But AMD could surprise us calculating the table in both higher and lower temperature.
But anyway the farther you are from the operating condition at boot, the less precise the calculations will be. I suppose that for far temperature from boot, the values will be more conservative.
So if you use the system with LN2 at -190C, it's advisable to boot with this temperature so that the VFT are the most accurate possible...
 
LTC8K6 said:
I'm not clear on the intent of XFR. Is it to get higher ST performance with one core boosting above Precision Boost?
Click to expand...

According to shadow p-state description, the technology is independent from load. While intel has semiprogrammed clocks for ST, 2 cores,4 cores etc, AMD has not preprogrammed clocks: the algorithm is "raise the clock until you reach one of temperature, tdp or vcore limit". Obviously with light loads the clock will raise more, but theoretically even with 16t you can surpass max turbo core (e.g. with a good cooling or liquid)...
 
It seems to me that when XFR is enabled, the CPU will clock above the factory turbo, and based on load it will keep increasing the clock above turbo till it hits a TDP that the HSF can no longer handle and down clock to a safe point when XFR detects it getting near that peak.
 
bjt2 said:
It calculates a voltage frequency temperature table, but i suppose that it will compensate only for higher thatn boot temperature. But AMD could surprise us calculating the table in both higher and lower temperature.
But anyway the farther you are from the operating condition at boot, the less precise the calculations will be. I suppose that for far temperature from boot, the values will be more conservative.
So if you use the system with LN2 at -190C, it's advisable to boot with this temperature so that the VFT are the most accurate possible...
Click to expand...
I meant damaging the CPU physically with the temperature changes.
 
LTC8K6 said:
I meant damaging the CPU physically with the temperature changes.
Click to expand...
The increase is not istantaneous. Moreover at -190C the CPU will draw much less power: leakage nearly zero, greater transconductance... It's difficult to damage the CPU...
 
1700 OC potential ~ 3.9-4.1 GHz
1800x ~ 4.1-4.3 GHz

we just tested a 1700, it hit 4.0GHz stable in everything, but ONLY in the Crosshair mainboard, the lower-end boards it was hovering around 3.80GHz as the VRM's were cooking with extra voltage. It however was maxing around 4050MHz, so I'd say 1700 can do 3.9-4.1GHz, of course the 1800X will probably do 4.1-4.3 as no doubt better binned, but if your clocking the motherboard has a big impact on the overclock and so far Asus Crosshair and Asrock Taichi seem the best two
Click to expand...

https://www.reddit.com/r/Amd/comments/5vqsqp/ryzen_1700_oc_40ghz_but_only_in_a_top_tier_mobo/
 
As expected, the Crosshair's VRM looks to be built like a tank and it shows.

Still from 3 GHz base to 3.8GHz on all eight cores for $320 it's not bad at all, but then you need a motherboard that can sustain that 24/7 without blowing the VRM up...


The cheaper boards shouldn't have problems dealing with 6 core Ryzen, even less with 4 core models.



I wonder if some manufacturer will release a no frills motherboard with a monster VRM without any of the extra bells and whistles that aren't needed...
 
Usually the low power chips are the one with low mean leakage. In turn those are the ones with weaker transistors (in terms of transconductance), that requires more volts for the same frequency. I am not surprised it overclocks only 1GHz from base...
 
.vodka said:
As expected, the Crosshair's VRM looks to be built like a tank and it shows.

Still from 3 GHz base to 3.8GHz on all eight cores for $320 it's not bad at all, but then you need a motherboard that can sustain that 24/7 without blowing the VRM up...


The cheaper boards shouldn't have problems dealing with 6 core Ryzen, even less with 4 core models.



I wonder if some manufacturer will release a no frills motherboard with a monster VRM without any of the extra bells and whistles that aren't needed...
Click to expand...
Yup, I tam seriously considering waiting a month and get the taichi and 1600x.
 
bjt2 said:
Usually the low power chips are the one with low mean leakage. In turn those are the ones with weaker transistors (in terms of transconductance), that requires more volts for the same frequency. I am not surprised it overclocks only 1GHz from base...
Click to expand...
Well, Ryzen has no high power chips. Wait, thought you were expecting 4.5Ghz?
 
.vodka said:
As expected, the Crosshair's VRM looks to be built like a tank and it shows.

Still from 3 GHz base to 3.8GHz on all eight cores for $320 it's not bad at all, but then you need a motherboard that can sustain that 24/7 without blowing the VRM up...


The cheaper boards shouldn't have problems dealing with 6 core Ryzen, even less with 4 core models.



I wonder if some manufacturer will release a no frills motherboard with a monster VRM without any of the extra bells and whistles that aren't needed...
Click to expand...

Crosshair
xzHWm5h.png


Taichi
b4cTA4x.png


I can't tell what all the little bits are rated for but the Taichi seems to have more little bits.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top