Ryzen 1700 overclocking

Just out of curiosity Shrimp, what's your opinion of the 1800X? It basically runs @3.8ghz on all cores assuming you aren't approaching the thermal limits for XFR. I'd wager that the voltage is above the sub 1.2V as well, but I cannot test that due to not owning one [emoji14]

I haven't conducted 2hr P95 stability testing @3.8ghz yet, but plan to this weekend. It did pass the 20min stability testing I do when bumping clocks @1.25V w/ LLC set to 1 for my CPU. My thoughts on it are that it would make my 1700 a lesser 1800X: a stock 1800X in more than 2 cores, and slightly less single threaded performance, yet still better than a stock 1700 across the board.

ShrimpBrime said:

Probably would be wise to follow that 1.35v max with FinFet. One main reason is the transistors stop responding to the increase of voltage.

Shorter pipelines help create this issue and makes difficult for a higher clock speed. So we are stuck to a lower clock speed, saving power and relaxing the restriction on some of the less critical transistors. So you increase the frequency and loose that efficiency, add heat and really put more stress on the processor in general whether we look at it from a power savings stand point or sheer throughput, your not going to gain much or any at all pushing 1.45v and trying to accomplish maybe a personal goal.

So you look at how you want to build, or engineer if you will, a processor. Now both top makers have tried long pipelines. It's a great selling feature, but throughput is lost. Here's where old sayings are born. Faster isn't always better. Ask the wife . Just because you can run 1.41250v stable and has a clock speed of 3846.4mhz, you've totally lost all efficiency to gain maybe 1% or so.

A wise man would tell any average OC'er and Gamer'er to run 1.35v or less with a Ryzen processor and get what you can from it. 3.7ghz to 3.8ghz is not worth 1.35v. Keep that sub 1.2v and 3.7ghz... (that is my 2 cents and sticking to it)

Click to expand...

It would seem that 3.7ghz is the sweet spot. Statistical analysis says only time will tell.

I look at cooling ability. If you can keep the core cool under load at slightly higher voltages then a slight bump isn't going to hurt. I would never suggest any overclocker to run their CPU so high that the efficiency of the CPU is compromised. That would be counter productive anyways. We all should be running benchmarks to test the CPU. I run them every single time I bump the speed up even for 5MHz. If I take the CPU up too far and the benchmarks drop in score then obviously the chip is struggling despite the fact it was able to boot to Windows and run the test. Average overclocker should probably stick to manufacturer's recommendations. Someone looking to push and has applied the correct cooling solution should be able to go a tad higher. If this were not so then extreme cooling wouldn't work so well.

I have ran a Core 2 Quad 6600 G0 under phase change for over a year that was way overvolted. I still have the damn thing and it won't die. The Core line had a shortened pipeline compared to the P4.

In the end benchmark the CPU and if the score drops then why bother. Watch your temps, if it is running too hot then obviously the cooling isn't cutting it.

Bluefalcon13 said:

Here's to hoping I win the silicone lottery!

Click to expand...

Come on man! Silicone, really?

Silicone
From Wikipedia, the free encyclopedia
Not to be confused with the chemical element silicon.
Silicone caulk can be used as a basic sealant against water and air penetration.

Silicones, also known as polysiloxanes, are polymers that include any inert, synthetic compound made up of repeating units of siloxane, which is a chain of alternating silicon atoms and oxygen atoms, frequently combined with carbon and/or hydrogen. They are typically heat-resistant and rubber-like, and are used in sealants, adhesives, lubricants, medicine, cooking utensils, and thermal and electrical insulation. Some common forms include silicone oil, silicone grease, silicone rubber, silicone resin, and silicone caulk.

Silicon
From Wikipedia, the free encyclopedia
This article is about the chemical element. For other uses, see Silicon (disambiguation).
Not to be confused with the silicon-containing synthetic polymer silicone.

Silicon is a chemical element with symbol Si and atomic number 14. A hard and brittle crystalline solid with a blue-gray metallic luster, it is a tetravalent metalloid. It is a member of group 14 in the periodic table, along with carbon above it and germanium, tin, lead, and flerovium below. It is not very reactive, although more reactive than germanium, and has great chemical affinity for oxygen; it was first purified and characterized in 1823 by Jöns Jakob Berzelius.

Electronic grade
Monocrystalline silicon ingot grown by the Czochralski process.

The use of silicon in semiconductor devices demands a much greater purity than afforded by metallurgical grade silicon. Very pure silicon (>99.9%) can be extracted directly from solid silica or other silicon compounds by molten salt electrolysis.

DaveB said:

Come on man! Silicone, really?

Silicone
From Wikipedia, the free encyclopedia
Not to be confused with the chemical element silicon.
Silicone caulk can be used as a basic sealant against water and air penetration.

Silicones, also known as polysiloxanes, are polymers that include any inert, synthetic compound made up of repeating units of siloxane, which is a chain of alternating silicon atoms and oxygen atoms, frequently combined with carbon and/or hydrogen. They are typically heat-resistant and rubber-like, and are used in sealants, adhesives, lubricants, medicine, cooking utensils, and thermal and electrical insulation. Some common forms include silicone oil, silicone grease, silicone rubber, silicone resin, and silicone caulk.

Silicon
From Wikipedia, the free encyclopedia
This article is about the chemical element. For other uses, see Silicon (disambiguation).
Not to be confused with the silicon-containing synthetic polymer silicone.

Silicon is a chemical element with symbol Si and atomic number 14. A hard and brittle crystalline solid with a blue-gray metallic luster, it is a tetravalent metalloid. It is a member of group 14 in the periodic table, along with carbon above it and germanium, tin, lead, and flerovium below. It is not very reactive, although more reactive than germanium, and has great chemical affinity for oxygen; it was first purified and characterized in 1823 by Jöns Jakob Berzelius.

Electronic grade
Monocrystalline silicon ingot grown by the Czochralski process.

The use of silicon in semiconductor devices demands a much greater purity than afforded by metallurgical grade silicon. Very pure silicon (>99.9%) can be extracted directly from solid silica or other silicon compounds by molten salt electrolysis.

Click to expand...

Lol... It was past my bedtime...

Bluefalcon13 said:

Just out of curiosity Shrimp, what's your opinion of the 1800X? It basically runs @3.8ghz on all cores assuming you aren't approaching the thermal limits for XFR. I'd wager that the voltage is above the sub 1.2V as well, but I cannot test that due to not owning one [emoji14]

I haven't conducted 2hr P95 stability testing @3.8ghz yet, but plan to this weekend. It did pass the 20min stability testing I do when bumping clocks @1.25V w/ LLC set to 1 for my CPU. My thoughts on it are that it would make my 1700 a lesser 1800X: a stock 1800X in more than 2 cores, and slightly less single threaded performance, yet still better than a stock 1700 across the board.

Click to expand...

You fellas would examine P-state values.

1800X is topped out from the box. Highly doubt any 1800X is going to require more than 1.35v. Alas I don't own an 1800X either. What's the point? There's no OC headroom.....

ShrimpBrime said:

You fellas would examine P-state values.

1800X is topped out from the box. Highly doubt any 1800X is going to require more than 1.35v. Alas I don't own an 1800X either. What's the point? There's no OC headroom.....

Click to expand...

I don't think it runs over 1.35V either, but you had recommended staying as low (under 1.2V) as possible on the 1700. That's why I was asking, just wanted your thoughts regarding that. Believe it or not, you're respectable folk around here with more experience than me.

I do agree 3.7ghz is a good speed for this CPU, but a minimal cost 3.8ghz also seems like a reasonable proposition to me, assuming staying under the 1.35V threshold.

Considering the recommendation for a typical user......

If we are looking for top clocks, air and liquid cooling is useless. The processor just isn't designed for high clock speeds.

If more IPC was desired, we would turn off SMT.

ShrimpBrime said:

Considering the recommendation for a typical user......

If we are looking for top clocks, air and liquid cooling is useless. The processor just isn't designed for high clock speeds.

If more IPC was desired, we would turn off SMT.

Click to expand...

I wish the bios in the Prime had this option. I could actually test this in Linux with the program I plan on writing this weekend and running. Could see if it would negatively or positively impact it on a well built scheduler [emoji14]

Can't wait to see some results.

J-Bo said:

I look at cooling ability. If you can keep the core cool under load at slightly higher voltages then a slight bump isn't going to hurt. I would never suggest any overclocker to run their CPU so high that the efficiency of the CPU is compromised. That would be counter productive anyways. We all should be running benchmarks to test the CPU. I run them every single time I bump the speed up even for 5MHz. If I take the CPU up too far and the benchmarks drop in score then obviously the chip is struggling despite the fact it was able to boot to Windows and run the test. Average overclocker should probably stick to manufacturer's recommendations. Someone looking to push and has applied the correct cooling solution should be able to go a tad higher. If this were not so then extreme cooling wouldn't work so well.

I have ran a Core 2 Quad 6600 G0 under phase change for over a year that was way overvolted. I still have the damn thing and it won't die. The Core line had a shortened pipeline compared to the P4.

In the end benchmark the CPU and if the score drops then why bother. Watch your temps, if it is running too hot then obviously the cooling isn't cutting it.

Click to expand...

There's different factors to consider. Your of the 1% whom may run SS on a daily rig out of the 1% of all PC users being an overclocker.

If your on a Bench team, your always going to go full all out. It's the nature of winning. Getting the highest score w/e.

My point is cooling helped FX. Cooling helped Phenom and yes Core 2. It will help even with Ryzen, but your looking at maybe 5ghz on LN2. 4.5ghz'ish DICE and 4ghz'ish on an expensive water loop...... but looking at big time voltage, the worth isn't there just to play some games, web browsing or risking that fail during a long render.

I agree with you. I am was just saying that 1.38 Volts or even 1.4volts with excellent cooling will not kill the core in short order. This is of course if the temps are under control during heavy and full loads and the CPU isn't struggling.

J-Bo said:

I agree with you. I am was just saying that 1.38 Volts or even 1.4volts with excellent cooling will not kill the core in short order. This is of course if the temps are under control during heavy and full loads and the CPU isn't struggling.

Click to expand...

I'd disagree based solely on my less-than-complete semiconductor knowledge. I recall some of the equations from my semi class resulted in a Volts/area measurements, and those can get astronomically large which can damage the semiconductor material itself, or fundamentally change it, especially if you consider we are looking at a 16nm process. To put it into perspective, a silicon crystal atom is approximately 5.4Å (10Å = 1nm, so 16nm = 160Å). We are dealing with a TINY area that the voltage traverses.

If AMD says keep it under 1.35V when overclocking, I'm going to listen. I don't have $330 laying around to replace my CPU.

^^^^QFT^^^^^

Bluefalcon13 said:

If AMD says keep it under 1.35V when overclocking, I'm going to listen. I don't have $330 laying around to replace my CPU.

Click to expand...

Can you provide a link to exactly where AMD specifies 1.35V as the maximum Core Voltage?

DaveB said:

Can you provide a link to exactly where AMD specifies 1.35V as the maximum Core Voltage?

Click to expand...

I can't, but ATM asked them for that info post review. They gave him that info. It's posted in the review thread if you'd like to see it. I believe it's page 2 or 3 with 10 posts per page on the various tapatalk apps

Well damn: I was wrong, that was for temperatures. ATM has it posted in his review on the frontpage, unless you are saying that ATM lied to us..... and the world.

Another edit: yes, they didn't say maximum voltage, they said max recommended voltage for overclocking. They also stated 1.45V with 'reduced lifespan' which could be weeks or 10 years vs normal. Either way, 1.35V is a good, solid, safe number.

I do understand that the die keeps shrinking along with transistor material and size. I do understand that there are electron mobility limitations and electron hole mobility problems with shrinking them. I don't agree that AMD has given us the absolute limit to the voltage the core can handle. If they have then AMD are very foolish and run a risky business. This is because cooling has a direct correlation with electron hole mobility through silicon. SiGe is next but not required until silicon is down to 10nm which is where silicon has electron mobility and hole mobility problems. SiGe is on the way though along with 10nm silicon, FinFET transistors.

As I stated earlier though, average overclockers should just stick with the 1.35v that AMD has suggested. What ever makes one feel safe and there is nothing wrong with that. Anytime you overvolt there are risks. The question is simply whether or not to take the risk. I am not so broke that I sit in worry about cooking a CPU by overvolting or I too would simply opt not to.

Information from AMD Regarding Overclocking

As a general guideline: a CPU voltage of up to 1.35 V is acceptable for driving everyday overclocks of the AMD Ryzen processor. Core voltages up to 1.45 V are also sustainable, but our models suggest that processor longevity may be affected. Regardless of your voltage, make sure you’re using capable cooling to keep temperatures as low as possible.
I have it in PDFs that accompanied the CPUs

Bluefalcon13 said:

I can't, but ATM asked them for that info post review. They gave him that info. It's posted in the review thread if you'd like to see it. I believe it's page 2 or 3 with 10 posts per page on the various tapatalk apps

Well damn: I was wrong, that was for temperatures. ATM has it posted in his review on the frontpage, unless you are saying that ATM lied to us..... and the world.

Another edit: yes, they didn't say maximum voltage, they said max recommended voltage for overclocking. They also stated 1.45V with 'reduced lifespan' which could be weeks or 10 years vs normal. Either way, 1.35V is a good, solid, safe number.

Click to expand...

I asked because the specs AMD provides for Ryzen are a joke and nowhere do they mention any limitation on core voltage. Given it is a 14nm FinFET, 1.45V seems more likely high limit to me, 1.35V is unreasonably low. The ASRock B350 I had, with the core voltage on auto, juggled the core voltage around as high as 1.43V.

And don't hyperventilate about what ATM said and am I questioning whether he lied, that's just a dumb thing to say. ATM repeated what he was told by an AMD rep, but there's no way to know if what he was told is truly a specification value or just some AMD employee's opinion. Again, AMD has very little on the website to go by.

Johan45 said:

Information from AMD Regarding Overclocking

As a general guideline: a CPU voltage of up to 1.35 V is acceptable for driving everyday overclocks of the AMD Ryzen processor. Core voltages up to 1.45 V are also sustainable, but our models suggest that processor longevity may be affected. Regardless of your voltage, make sure you’re using capable cooling to keep temperatures as low as possible.
I have it in PDFs that accompanied the CPUs

Click to expand...

It would be nice if AMD would put that PDF out on their website.

Based on my investigation into voltage scaling on my R7 1700 (posted here), I wouldn't suggest going over 1.35v either, maybe not even over 1.3v. Beyond that point the heat and power start shooting up a lot, for pretty minimal gains. In addition to being energy inefficient, I don't feel the way the chip responds at that point is promising. It's basically telling us that it really doesn't want to go higher, and although you can force it to accept whatever voltage you want, I'm not sure it's wise from the standpoint of longevity.

I'm running on a GAMAXX 400 as well right now, since I'm waiting for my AM4 brackets to arrive, and it seems halfway decent for a budget heatsink, at least if you replace the fans. I have a Panaflo Medium Output (86 CFM) and a Scythe Glide Stream 120 (109 CFM) in push-pull, and at least with Brute Force™ it's capable of keeping the CPU cool even at 3.9 GHz (sounds like a jet engine though). Overall I'm not that impressed with other aspects of the heatsink though. It's very difficult to unhook the clip to remove the heatsink (a couple of times I resorted to just unscrewing the plastic mounting brackets), and the stock fan that comes with it is pretty wimpy.

- - - Updated - - -

DaveB said:

I asked because the specs AMD provides for Ryzen are a joke and nowhere do they mention any limitation on core voltage. Given it is a 14nm FinFET, 1.45V seems more likely high limit to me, 1.35V is unreasonably low. The ASRock B350 I had, with the core voltage on auto, juggled the core voltage around as high as 1.43V.

And don't hyperventilate about what ATM said and am I questioning whether he lied, that's just a dumb thing to say. ATM repeated what he was told by an AMD rep, but there's no way to know if what he was told is truly a specification value or just some AMD employee's opinion. Again, AMD has very little on the website to go by.

Click to expand...

It just doesn't seem like you get a lot for that last 0.1 volts. At least in my experience, it only buys you about 100 MHz. So, I just don't know whether it's worth it, especially if AMD is suggesting to stay lower. Also, keep in mind that this is Samsung LPP+ 14 nm process. It may not be the same as Intel. Obviously it doesn't seem to like clocking as high. It could be that since it's designed for low power consumption, it simply can't handle higher voltage as well.

However, to be fair, 1.45v will also overheat on a Wraith Spire, so that could also be why they're saying to stay around 1.35v (for people with inadequate coolers).