"Eco-friendly" overclocking of Ryzen 1700?

TrackSmart

Member
Feb 6, 2011
36
8
81
Is there potential for an "eco-friendly", near-stock-voltage overclock of a Ryzen 1700 that won't cause a loss of single-threaded performance?
I'd like to maintain low idle power usage AND keep load power reasonable. What is "reasonable" to me? Generally no higher than the stock 1800X (about 90W package power -- see graph below). I'll be using a B350 board (Asus) and the stock cooler (Wraith Spire).

What I've heard around the web (which is full of conflicting information):
1) Some folks have reported that idle power usage increases substantially, even with a stock voltage overclock, because the CPU doesn't go into low power states (i.e. it stays in a P0 power state). Others say this is not true: The chip will still downclock and idle at a low power usage, even under a P0 power state. In other words, while "full base clock frequency" is reported, the chip itself will still "downclock" internally and achieve a low idle power usage. Which of these conflicting claims is true?

2) I know you can push clocks considerably higher on this chip (from 3.0 to 3.8-4.0 GHz), but only with considerably higher voltage and power usage at load (i.e. *way* above the 65 watt stock TDP). But is a move from 3.0 to 3.4/3.5 GHz possible/likely at stock voltage on a Ryzen 1700? Or nearly-stock voltage?

3) Related: Some say that modest overclocking is not worth it because you get lower single threaded performance. In other words, you get a slight rise in base frequency (e.g. to 3.4/3.5 GHz), but your chip will no longer boost individual cores to 3.7 GHz and you don't get XFR (another 50 - 100 GHz). Is this true?

4) Finally, if what I'm hoping to do is possible: Is the stock Wraith Spire cooler sufficient for the task of increasing base clocks modestly while maintaining some thermal headroom for Boost/XFR? It is rated for 95W TDP I'm hoping to use no more power than that at load.



THANKS!

==================

86267.png
 
Last edited:

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,328
4,913
136
1) You can use P-state overclocking which gives the best of both worlds. Only certain AM4 boards support this, so you will need to check forums to see who has had success. I don't use this method so I don't know the list off the top of my head.
2) 3.7GHz-3.8GHz is about the limit you can push with 1.2V-1.25V, depending on silicon lottery. Once you get beyond 1.3V+ (and especially 1.4V+), your power usage will necessarily skyrocket.
3) Depends on how much you need ST performance. If you OC to 3.7GHz+ it's a moot point.
4) Stock Wraith Spire RGB cooler is probably good up to 1.25V. 1.30V would be pushing it. People generally target 3.6GHz-3.8GHz overclocks with the Wraith Spire RGB.

Graph of voltage required for a given frequency, per The Stilt (individual chip may vary, subject to silicon "lottery"):
8Rch6JF.png


My best Ryzen 1700 does 4.0GHz at ~1.38V under water, for reference. My 1800X will do 3.8GHz @ 1.25V on air cooling. These are BOINC stable 100% loads.
 

XavierMace

Diamond Member
Apr 20, 2013
4,307
450
126
On the other hand, I haven't been able to get my 1700X stable above 3.8Ghz. So, yeah, lottery is a pretty big factor.
 

Reinvented

Senior member
Oct 5, 2005
489
77
91
I do use P-state overclocking on my Taichi. But, after watching this video it's quite true.

I also am using a 1700, and my comfort zone is 3.9 @ 1.31v. It's stable with all cores clocked, but I choose not to do it, cause less power is nice, but it's not by much.

https://www.youtube.com/watch?v=bzxn1hS7Nq4
 
  • Like
Reactions: IEC

PG

Diamond Member
Oct 25, 1999
3,426
44
91
I was going to post that nice graph above.
For a short time I had a Ryzen 1700 and Asus Prime B350-Plus. It has a nice auto overclocking feature that put the clock speed at 3.65Ghz for my specific cpu. Voltage was very low, I think around 1.1875 or 1.2V. No I didn't measure power consumption but I can't imagine it was too bad.
Moral of the story, stick to 3.65Ghz or lower and voltage can still be pretty close to stock. Multi threaded power should go up quite a bit but of course single thread isn't really going to be better.
Full load like this will really push the stock cooler. You will need to have great case airflow and even then you will be near the limit.
 
  • Like
Reactions: VirtualLarry

TrackSmart

Member
Feb 6, 2011
36
8
81
Thanks folks! This is a helpful start!

The video posted by Reinvented is quite worrying. *See table below, which is a screenshot with my annotations*. This particular system is showing a 93W increase in power draw under load at just 3.2 GHz. Mind you, this is full system power at the wall (including power supply inefficiencies, etc). My goal is to stay within a 95W total package. So...

1) The good news is that low idle power draw is possible, even overclocked. And p-states don't seem to matter much (only a few watts difference at idle when disabled).
2) Unanswered: Realistically, assuming an "average" piece of silicon (I realize there is high variation), what kind of clock speed can one expect while staying within 95W of total package power and at reasonable temperatures using the stock Spire cooler?
3) Also unanswered: Is boost and XFR (i.e. 3.7 GHz on individual cores, plus 50 GHz XFR) still activated when mildly overclocked (e.g. 3.4 GHz instead of 3.0 GHz) on stock cooling? Or is that automatically disabled? Or does it fail because you are already above the stock thermal and power limits for boost to be activated due to being overclocked?

This guy in the video literally says "Basically, if you're interested in power efficiency, don't overclock at all!" If I have similar silicon, I'm looking at possibly ~3.3 GHz base clock (vs 3.0 GHz stock) at the voltages he used. And it is unclear if I lose the ability to boost on select cores, which would mean lower single and two threaded performance.

So, maybe that is my answer ("Don't overclock at all")? Or am I missing something here?

IMAGE: You'll have to click-through to Google Photos, my apologies...
https://photos.app.goo.gl/hKZvGb1N9lQESzBy1
 
Last edited:

VirtualLarry

No Lifer
Aug 25, 2001
56,326
10,034
126
Basically, Ryzen has a lot of carefully-designed power- and thermal- sensors, and can clock part of the chip up/down to meet thermal and power constraints.

Once you put the CPU into "overclocked mode", all of that power-management stuff gets basically disabled, including XFR. (*)

(*) But with BIOSes that allow P-state OC, this may not be true.
 

Reinvented

Senior member
Oct 5, 2005
489
77
91
Thanks folks! This is a helpful start!

The video posted by Reinvented is quite worrying. *See table below, which is a screenshot with my annotations*. This particular system is showing a 93W increase in power draw under load at just 3.2 GHz. Mind you, this is full system power at the wall (including power supply inefficiencies, etc). My goal is to stay within a 95W total package. So...

1) The good news is that low idle power draw is possible, even overclocked. And p-states don't seem to matter much (only a few watts difference at idle when disabled).
2) Unanswered: Realistically, assuming an "average" piece of silicon (I realize there is high variation), what kind of clock speed can one expect while staying within 95W of total package power and at reasonable temperatures using the stock Spire cooler?
3) Also unanswered: Is boost and XFR (i.e. 3.7 GHz on individual cores, plus 50 GHz XFR) still activated when mildly overclocked (e.g. 3.4 GHz instead of 3.0 GHz) on stock cooling? Or is that automatically disabled? Or does it fail because you are already above the stock thermal and power limits for boost to be activated due to being overclocked?

This guy in the video literally says "Basically, if you're interested in power efficiency, don't overclock at all!" If I have similar silicon, I'm looking at possibly ~3.3 GHz base clock (vs 3.0 GHz stock) at the voltages he used. And it is unclear if I lose the ability to boost on select cores, which would mean lower single and two threaded performance.

So, maybe that is my answer ("Don't overclock at all")? Or am I missing something here?





Ryzen%201700%20-%20Power%20versus%20Frequence%20and%20Voltage%20-%20Spreadsheet%20from%20youtube%20-%202.jpg
[/URL][/IMG]

Can't see the picture. But as Larry said, once you overclock any amount, XFR does get disabled and any power efficiency you had is gone completely.
So, if you want efficiency, you leave it at stock with your 1700. P-states aren't really worth the difference in power draw either, if that's your goal. I use P-states simply because I hate the weather here, and it's hot even in Fall. But, at the same time I'd love the performance.
 

TrackSmart

Member
Feb 6, 2011
36
8
81
Hi Folks,

One more try for the picture, which is a 1000 words, as the saying goes. Here is a link to Google Photos in case this doesn't work:
https://photos.app.goo.gl/hKZvGb1N9lQESzBy1

Things to consider:
1) Increasing the base clock to just 3.4 GHz on this 1700 resulted in a 109W increase in power at load compared to stock idle (measured at the wall). So let's say that this is an actual load power of 95 watts from the card itself. That's still below the 3.7 GHz boost speed for lightly threaded tasks. And you use 46% more power in heavy multi-threaded workloads.

2) Assuming you must give up boost and XRF when overclocked (is this definitive?), you would need to go up to at least 3.7 GHz to get similar single threaded speed to the stock CPU. For this particular Ryzen 1700, going to just 3.6 GHz yielded 143W difference between stock idle and load. That will mean a power hungry and hot card at load with the stock cooler, which is only rated for 95W TDP.

Overall: At 3.6 GHz you end up with more than double the load power consumption (65W --> 140W?) for a 20% increase in base clock speed (3.0 GHz --> 3.6 GHz). And you gain no speed in lightly threaded tasks (which would have boosted to 3.7 GHz at stock).

So, I think that this youtuber and Larry are correct. If you like efficiency, stay away from overclocking Ryzen. I guess I had higher hopes, given that Ryzen 1700 provides an amazing amount of compute power for 65W. But apparently when you ditch the sophisticated power saving features and brute-force an all-core increase in clock speed, you end up with very high power consumption. Even with modest overclocks.

I think I'll just enjoy the incredible efficiency of the stock chip, which does a ton of work in a 65W envelope.





jLkFfytqem_vUToEvLVC0inY1kXG35l7tAxa5b5gtwg25u755wqJgJNwPEVr9FBbmfsGzg3y-vLsuWydZimgFx5WGjl2r-tFo4vKYRPdbs4ZDhkI5QJy_Yp7r0B74BbUwmKBPM5q3os2Z8_TLBgrQgM44kuy7DHHjDxqNqxMsXQYFSVsJM0bii24IdbWNTmFnvsSqQenkIDFfpP0-FdNAdDGRfBGoy442RBKQWJY-Wq3yOnz-7GSAisfMVdOwQkkcBenjS_pTyb4k5Q3PbQfRBsCkCgqHV96U-HQmhBi7FKVpPNQX1Fp635NRNo0sMcYkA4G0dSsCMUFbWl5EC2KKtwvVNX9I357RnK44MYnSSOvVFVb_5R6RfExF69N2E-9fV6Oqzw588thS1CtvAxnxFRnqEPtMHx-dutd0o3Hl5NW1cXbMpMg3_32rpQCMaponU8aDFBUbtxhHrIeb12S6zTcxIqFTM2cIWNOGJhZa6sDjTaISA_6rNlRhKc_KNaC0-j2hO7uzESc6yEKVzIsuNYmpscOjsQfk4rQ_8WwZknH01bQAqcVraKeI_6ifmOfXLuQhptdHVdNPRSPx9GIhqDgDq7W9ALawr6uGYSI3PNmIB0t9LQZpN2Efy4v6Ta_QQdtdj0XAZAWYVkGvJHvCloEH0s-M4EJ6y-v=w1775-h850-no

fly6MuWZk6OrFXD42
 
Last edited:

TempAcc99

Member
Aug 30, 2017
60
13
51
Basically, Ryzen has a lot of carefully-designed power- and thermal- sensors, and can clock part of the chip up/down to meet thermal and power constraints.

Once you put the CPU into "overclocked mode", all of that power-management stuff gets basically disabled, including XFR. (*)

(*) But with BIOSes that allow P-state OC, this may not be true.

Hm so OCing on Ryzen actually sucks? I don't have a recent intel CPU but I have one. I can just up the multiplayer and keep turbo working. So for a 8700k I would up the multiplier by 2 and get 4.9 GHZ single-core turbo. Maybe voltage needs to be upped a tiny bit and then I'm done with the OC.
 

Reinvented

Senior member
Oct 5, 2005
489
77
91
Hm so OCing on Ryzen actually sucks? I don't have a recent intel CPU but I have one. I can just up the multiplayer and keep turbo working. So for a 8700k I would up the multiplier by 2 and get 4.9 GHZ single-core turbo. Maybe voltage needs to be upped a tiny bit and then I'm done with the OC.

That's not what was said at all...not even in the slightest...
 

VirtualLarry

No Lifer
Aug 25, 2001
56,326
10,034
126
Hm so OCing on Ryzen actually sucks?
If "actually sucks", is some sort of Intel fanboy term for "AMD overclocks differently than Intel", then yeah...

Anyways, I can't think of a platform, in which I would want to max out my all-core voltage and frequency, and then have some nebulous "turbo" feature push things out of spec on me and have the CPU core crash. So, it's really not an issue.

With Intel's "K" CPUs, you can either adjust the base clock multi, or the turbo multis. Generally, you don't tweak both of them.
 

TrackSmart

Member
Feb 6, 2011
36
8
81
Two things in response:
1) At stock speeds, the Ryzen chips are binned to have an impressive amount of performance per watt -- if you have an application that can use 8 cores and 16 threads. But you definitely trade single-threaded performance compared to Intel's highest clocked chips.
2) Regarding overclocking: It seems that this generation of Ryzen chips have limited overhead for overclocking before you need to apply quite a bit of voltage (which = a lot more power). So in that sense, yes, these are not amazing CPUs for overclocking if you want to stay power efficient. That said, I don't know that things are so much rosier on the Intel side in recent years. For Ryzen, people are regularly getting 25%-30% overclocks (e.g. 3 GHz to 3.8/4.0GHz) on Ryzen 1700 chips with fairly simple tower air coolers. How many modern day Intel chips let you push base clock much higher (percentage wise) above stock? I don't think we are still in the days of the i5-2500k, where that was a common occurrence. But correct me if I'm wrong on that.
 

kjboughton

Senior member
Dec 19, 2007
330
118
116
Let me just say that Intel claims max efficiency at 1.2GHz. This has not changed for many generations of they current Core architecture.

I doubt AMD could do better.

So, no, there is no way to "eco-friendly" overclock your system. In fact, you should probably place it in low power mode if that's a concern of yours.
 

coercitiv

Diamond Member
Jan 24, 2014
6,187
11,853
136
How many modern day Intel chips let you push base clock much higher (percentage wise) above stock? I don't think we are still in the days of the i5-2500k, where that was a common occurrence. But correct me if I'm wrong on that.
If Intel were to make an unlocked 65W SKU, it would push much higher than stock.

AMD offer is excellent not because of it's relative overclock potential (stock vs. max) but rather because all chips can overclock. Their only limit currently is the A chipset which is in fact a healthy way to tell consumers that those boards are built with low power requirements in mind, hence unfit for 95W+ TDP CPUs.
 

TrackSmart

Member
Feb 6, 2011
36
8
81
Let me just say that Intel claims max efficiency at 1.2GHz. This has not changed for many generations of they current Core architecture.

I doubt AMD could do better.

So, no, there is no way to "eco-friendly" overclock your system. In fact, you should probably place it in low power mode if that's a concern of yours.

Indeed, that's what I've arrived at: Doubling the power usage yields only a 15-20% increase in clock speed. And you lose turboboost for single threaded tasks. And you lose all advanced power management. There also appears to be some legitimate binning happening, given how much voltage (and power) need to be increased on Ryzen 1700 that reaches the same frequencies as a stock Ryzen 1800x.

As for underclocking: 65W at load is pretty darn thrifty given the performance provided by the 1700. With an efficient power supply, the system should idle at about 40-45 watts and push no more than about 120 watts at full load on 8 cores.
There are diminshing returns to underclocking such an efficient chip. Just like doubling the gas mileage of a Prius from 50 to 100 MPG won't change your gas bill very much, but doubling the mileage of your SUV from 18 to 36 MPG will save you a lot of money at the pump.
 

Reinvented

Senior member
Oct 5, 2005
489
77
91
Indeed, that's what I've arrived at: Doubling the power usage yields only a 15-20% increase in clock speed. And you lose turboboost for single threaded tasks. And you lose all advanced power management. There also appears to be some legitimate binning happening, given how much voltage (and power) need to be increased on Ryzen 1700 that reaches the same frequencies as a stock Ryzen 1800x.

As for underclocking: 65W at load is pretty darn thrifty given the performance provided by the 1700. With an efficient power supply, the system should idle at about 40-45 watts and push no more than about 120 watts at full load on 8 cores.
There are diminshing returns to underclocking such an efficient chip. Just like doubling the gas mileage of a Prius from 50 to 100 MPG won't change your gas bill very much, but doubling the mileage of your SUV from 18 to 36 MPG will save you a lot of money at the pump.

Stock with my R7 system, it doesn't pull more than 145-150W at full load on stock. That's DDR4-3200, R7 1700 with CnQ/C6, etc and NVME as well as a RX 480 8GB. Idle, the lowest it'll go is 123-125W. There really is no easy way, period. It's silly to think you can overclock and draw less power.
 

TrackSmart

Member
Feb 6, 2011
36
8
81
Stock with my R7 system, it doesn't pull more than 145-150W at full load on stock. That's DDR4-3200, R7 1700 with CnQ/C6, etc and NVME as well as a RX 480 8GB. Idle, the lowest it'll go is 123-125W. There really is no easy way, period. It's silly to think you can overclock and draw less power.

123-125 watts at idle sounds high. These guys measured 48W at idle on their gaming desktop (with GTX 1080 no less) and 125W during "content creation" loads (i.e. CPU heavy, not GPU). I've seen similar 40-60W numbers [system idle at wall] in other Ryzen reviews. Maybe you have some power hungry additional components? Or don't have full power saving options enabled for your CPU and GPU when at stock?

Check out their graph at the top of the page:
http://www.guru3d.com/articles_pages/amd_ryzen_7_1700_review,7.html

65W is what the CPU draws a load alone. I was hoping (and clearly that hope is gone) that I could coax a bit more performance, while keeping CPU (alone) power relatively modest (e.g. 90-ish watts at load). No such luck. So I agree that was wishful thinking.
 
Last edited: