AMD Ryzen 9 3950X Overclocked To 5 GHz Across All 16 Cores On LN2

Status
Not open for further replies.
So as soon as I select "OC manually" in literally any X470/X570 BIOS, it's like TDP doesn't exist? If the goal is highest single core overclock, you'd prefer chips with lower TDP/core ratio since that usually means the cores are better binned?

Once you overclock (either manually or automatically) the TDP of the chip increases above the stock value and you need better cooling system to dissipate the heat from the chip. E.g. a chip can have a TDP of 95W on stock clocks and a TDP of 210W when overclocked. Wait to actual chips to be released to see how is AMD binning each SKU, which is silicon variability within the seam SKU, and what chip has better silicon for what you want to do. No one can answer that now.

High TDP with fewer cores doesn't imply better OC headroom. It can mean that those SKUs are using worse silicon and need higher TDP to achieve similar clocks than SKUs on better silicon.
 
I already explained my point, gave data, and answered your criticism. No need to do it all again.

Lets put it to you this way: IF the number/name in the specification windows in CPU-Z supports your argument, why do you have to have special decoding software to get such information, as you stated you have, if it is already available in CPU-Z? Where can you find the ID number of any chip, be it ES or not? (hint, it is not CPU-Z). What software can you use, and can you get it without software? All you have done is spin around in circles with me, and others trying to defend your original argument based on a number that has nothing to do with it being ES or not. The information you gave us, from a year old post (which that post was just cut and pasted off another website) was about the chips serial number/ID number, not the OPN, SKU, or code name, which was what was displayed in CPU-Z. You won't even admit that the OPN, SKU, or codename is different than the chips serial number/ID number, which is the ONLY number that will determine if it is ES or not. You have been proven wrong by me, by others, and sadly, by the very data you gave us. You can't bring yourself to admit you are wrong and don't even see that you already lost the argument a long time ago.
 
It doesn't have anything to do with TDPs. Once you overclock you are running outside of spec and the stock TDP rating is irrelevant. LN2 was needed because the voltage had to be pushed above 1.6V to achieve 5GHz clocks on that process node. Those high voltages produce excessive heat (Power ~ V³) and one has to use extreme cooling to get the system under control and running.

High TDP with fewer cores doesn't imply better OC headroom. It can mean that those SKUs are using worse silicon and need higher TDP to achieve similar clocks than SKUs on better silicon.

What keeps a motherboard manufacturer from doing some kind of TDP measurement on their own and using that for their own precision boost algorithm, like Asus's PB2? Perhaps they could do it because they know the full capability of their VRMs.

Also, leaky transistors that require higher voltages to operate can be some of the best overclockers. AdoredTV has speculated that the 3800x could be AMD's worst binned chips, but they could also be great overclockers with high voltage. The 3800x has the same TDP as the 3900x with 50% more cores.

Overclocking competition guys look for the chips that will hit high clock speeds regardless of voltage since they use exotic cooling. A highly binned notebook chip that runs 4.0 GHz at 1.0v is not as interesting to them as a chip that runs 5.2 GHz at 1.5v. That notebook chip may never go over 4.8 at any voltage.
 
Last edited:
What keeps a motherboard manufacturer from doing some kind of TDP measurement on their own and using that for their own precision boost algorithm, like Asus's PB2? Perhaps they could do it because they know the full capability of their VRMs.

Precision Boost 2 isn't ASUS' at all. ASUS does have a newer automatic overclocking feature, which basically does what your talking about. It estimates your overclocking headroom and the settings required to achieve the best results by analyzing the system's cooling, VRM temperatures, and so on.
 
Precision Boost 2 isn't ASUS' at all. ASUS does have a newer automatic overclocking feature, which basically does what your talking about. It estimates your overclocking headroom and the settings required to achieve the best results by analyzing the system's cooling, VRM temperatures, and so on.

I remember from the article on it here, but forgot the name I guess. I think you did a comparison between PB and the Asus implementation.
 
I remember from the article on it here, but forgot the name I guess. I think you did a comparison between PB and the Asus implementation.

It came up in a recent ASUS motherboard review. Specifically on the Maximus XI Formula. It's something I noticed in both the UEFI BIOS and in AI Suite III. It puts ASUS way ahead on the automatic overclocking front. No other company has anything like that. Its actually similar to what Intel and AMD do in their IPM and Ryzen Master utilities respectively, but those tools don't take the board into account. ASUS' overclocking does.
 
It came up in a recent ASUS motherboard review. Specifically on the Maximus XI Formula. It's something I noticed in both the UEFI BIOS and in AI Suite III. It puts ASUS way ahead on the automatic overclocking front. No other company has anything like that. Its actually similar to what Intel and AMD do in their IPM and Ryzen Master utilities respectively, but those tools don't take the board into account. ASUS' overclocking does.
OOPS!!
OUR WEBSITE IS TEMPORARILY OFFLINE
The website is down for maintenance. We apologize for any inconvenience.
Please check back again soon!

argh
hardocp-logo.png
 
Precision Boost 2 isn't ASUS' at all. ASUS does have a newer automatic overclocking feature, which basically does what your talking about. It estimates your overclocking headroom and the settings required to achieve the best results by analyzing the system's cooling, VRM temperatures, and so on.
For what it is worth, ASUS in-BIOS OC system does not analyze anything. It is simply a set of presets that it runs until it finds where the silicon fails, then moves back to the previous successful run.
 
For what it is worth, ASUS in-BIOS OC system does not analyze anything. It is simply a set of presets that it runs until it finds where the silicon fails, then moves back to the previous successful run.

I remember looking at this on the Maximus XI Formula. The regular XMP / AI Overclocking / DOCP settings are just as you describe. However, there is some sort of analysis that takes time to refresh on the bottom right of this screenshot taken from the Maximus XI APEX. You can see it has a cooler point value that it assigns based on your current temps. It shows AVX vs. Non-AVX, and various predicted clock speeds and the voltages required to get them. I talked about this in the Maximus XI Formula review, and I don't know how much analysis it does and perhaps that's the wrong word, but this is exactly what AI Suite III does now. ASUS says that these values change as it learns the CPU or circumstances change. This is from the Maximus XI APEX's very first POST. It took a few seconds for that "Prediction" pane to populate the data. This suggests its reading some sort of data, and I don't think it comes from a preset profile as this data varies by CPU. You can also force it to change by swapping to an air cooler, custom loop, AIO etc.

upload_2019-6-21_11-59-44.png
 
I remember looking at this on the Maximus XI Formula. The regular XMP / AI Overclocking / DOCP settings are just as you describe. However, there is some sort of analysis that takes time to refresh on the bottom right of this screenshot taken from the Maximus XI APEX. You can see it has a cooler point value that it assigns based on your current temps. It shows AVX vs. Non-AVX, and various predicted clock speeds and the voltages required to get them. I talked about this in the Maximus XI Formula review, and I don't know how much analysis it does and perhaps that's the wrong word, but this is exactly what AI Suite III does now. ASUS says that these values change as it learns the CPU or circumstances change. This is from the Maximus XI APEX's very first POST. It took a few seconds for that "Prediction" pane to populate the data. This suggests its reading some sort of data, and I don't think it comes from a preset profile as this data varies by CPU. You can also force it to change by swapping to an air cooler, custom loop, AIO etc.

View attachment 169137
That is what you might think....
 
It's been pretty accurate so far. I think I tried two different processors on the Maximus XI Formula, and it pretty much nailed it on both. I think I got a little higher than it suggested I could, but not by much. When we just had the presets, they over volted your CPU and weren't terribly accurate. They were way to conservative, probably to avoid warranty claims etc.
 
It's been pretty accurate so far. I think I tried two different processors on the Maximus XI Formula, and it pretty much nailed it on both. I think I got a little higher than it suggested I could, but not by much. When we just had the presets, they over volted your CPU and weren't terribly accurate. They were way to conservative, probably to avoid warranty claims etc.
I am just telling you how the system works, and I know for sure how it works, no matter how you might think it works. Of course, myself and the engineers that I have worked with on this might be fully incorrect, and you be 100% right. Was just trying to give you some information. You can do with it what you want.
 
I am just telling you how the system works, and I know for sure how it works, no matter how you might think it works. Of course, myself and the engineers that I have worked with on this might be fully incorrect, and you be 100% right. Was just trying to give you some information. You can do with it what you want.

Actually, I just looked into it and it's mostly based on preset profiles. it says as much in the AI Overclocking guide within the BIOS itself. It has predefined profile presets in a table which it references and makes it"s "prediction" based on your cooler score. The cooling score and system temps are all that it takes into account.

What you see in the UEFI BIOS appears to be a lot like what AI Suite III supposedly does by ASUS' description but this is obviously not true. So I was wrong.
 
Actually, I just looked into it and it's mostly based on preset profiles. it says as much in the AI Overclocking guide within the BIOS itself. It has predefined profile presets in a table which it references and makes it"s "prediction" based on your cooler score. The cooling score and system temps are all that it takes into account.

What you see in the UEFI BIOS appears to be a lot like what AI Suite III supposedly does by ASUS' description but this is obviously not true. So I was wrong.

Well, the cooler score is something good I suppose.
 
Lets put it to you this way: IF the number/name in the specification windows in CPU-Z supports your argument, why do you have to have special decoding software to get such information, as you stated you have

No. I said that I have a decoder to decodify the strings of engineering samples such as 1D2801A2M88E4_32/28_N or 2S1404E2VJUG5_20/14_N and the strings of qualification samples such as ZD3601BAM88F4_40/36_Y or ZS1406E2VJUG5_22/14_N.

The decoder isn't a "special decoding software". You are making up things again.
 
No. I said that I have a decoder to decodify the strings of engineering samples such as 1D2801A2M88E4_32/28_N or 2S1404E2VJUG5_20/14_N and the strings of qualification samples such as ZD3601BAM88F4_40/36_Y or ZS1406E2VJUG5_22/14_N.

The decoder isn't a "special decoding software". You are making up things again.

Oh, So you have a "key" that tells you what each part of the id number number or batch code means. Well why didn't you say so, because the way you said decoders with no explanation, i iook it as though you needed decoding software to optain that number, not decifer it. Big difference.

What is so funny about all this, is you keep arguing about the id number/batch codes and have yet to realize that CPU-Z doesn't displays that ID number.

BTW... I don't make stuff up! I do misunderstand things at times.. remember that next time before making accusations that are not true.
 
Last edited:
What is so funny about all this, is you keep arguing about the id number/batch codes and have yet to realize that CPU-Z doesn't displays that ID number.

BTW... I don't make stuff up! I do misunderstand things at times.. remember that next time before making accusations that are not true.

s-l600.jpg


So, again, what you say isn't true. You continue making up stuff.

I am not going to decode the whole ID string, because CPU-Z is already reporting information as the number of cores, silicon-grade level, TDP, and so. I will only mention the info that CPU-Z isn't giving, but is enconded in the ID string 1S1601A4VIHF4_29/16_N like that this is a first generation engineering sample with 1.6GHz base and 2.9GHz boost clocks.
 
Last edited:
(1) Is TDP a limiting factor or am I a moron? (I'm not seeing TDP being mentioned)

(2) How can we overcome any TDP limitations present on Zen/2 when overclocking manually?

(3) Which X570/X470 boards have/will have the features required for (2)?

(4) Is it possible to use PB to overclock higher by reducing the impact of TDP via (2) on boards in (3)?

Considering these boards will need to handle everything from 4 cores to 16, at a point power delivery is going to become an issue if you're overclocking the 16 core part on a mid-tier board.

We'll have to wait and see considering the 3950X is so far out, but it doesn't bode well, price wise, for anyone expecting great value when choosing a motherboard to OC the 16 core. =\
 
View attachment 169344

So, again, what you say isn't true. You continue making up stuff.

I am not going to decode the whole ID string, because CPU-Z is already reporting information as the number of cores, silicon-grade level, TDP, and so. I will only mention the info that CPU-Z isn't giving, but is enconded in the ID string 1S1601A4VIHF4_29/16_N like that this is a first generation engineering sample with 1.6GHz base and 2.9GHz boost clocks.

Nice screenshot. It shows your arrogance, and how much you really don't know. In that arrogance, you once again felt it gives you the right to accuse me of making stuff up (another false accusation). I was hoping that you would finally admit you are not as smart as you think you are, but apparently not.

So, now It's time to school you and show everyone how wrong you really are. There is NO software in the world that can give you the ID number-batch code of a cpu, which is the ONLY number that can show if it is an ES or not. Not CPU-Z, not Sandra, not aida64, non of them can display that number. That is because that ID number/batch code is only located on the heat spreader of the cpu and is not obtainable thru software. There are the rare occasions that the manufacture will include that number in the specification label (your screenshot) but that is only because AMD chose to include it on that particular ES. There are hundreds of ES of every cpu, that do not include the CPU ID/batch codes in the specification label (hence why you don't understand why you lost this argument before it you ever decided to start it). Now, unless you can give us a screen shot or a video of CPU-Z removing a heat sink, wiping off the thermal past, reading the CPU-ID/batch code and displaying it on the screen, you have nothing to support your position. Having it included in the specification label by AMD is not the same thing, as CPU-Z cannot display such information on every CPU be it an ES or not. It has to be included in the specification label by AMD, which is rarely done.

The rest of your comment about the ID/batch code number is just to try and make yourself look smart to those that don't really understand that all of your wisdom is just thru google.

Example:

http://www.cpu-world.com/info/id/AMD-ES-identification.html

and

http://www.cpu-world.com/info/id/AMD-Bulldozer-identification.html
 
Last edited:

Whats worse. It seems he sourced this image from here

https://www.ebay.com/itm/AMD-EPYC-7...Socket-SP3-For-Dell-R6415-R7415-/163572219855

WHICH IS WHY IT IS IMPORTANT TO SOURCE ALL MATERIALS!!!

We don't know what this dude in japan did to get the string in there, but I seriously doubt its legitimacy.

CPUID Processor Brand String has a limited amount of characters. Many of them reserved. I highly doubt AMD would take the time to bake in such specifics

cpu-z does a lot to fill in the info.
 
As an eBay Associate, HardForum may earn from qualifying purchases.
View attachment 169344

So, again, what you say isn't true. You continue making up stuff.

I am not going to decode the whole ID string, because CPU-Z is already reporting information as the number of cores, silicon-grade level, TDP, and so. I will only mention the info that CPU-Z isn't giving, but is enconded in the ID string 1S1601A4VIHF4_29/16_N like that this is a first generation engineering sample with 1.6GHz base and 2.9GHz boost clocks.

Since you generally can't overclock EPYC. How come the multiplier is 34x and it's running at 3365.13, if its a 1.6ghz processor with 2.9ghz boost?
 
Nice screenshot. It shows your arrogance, and how much you really don't know. In that arrogance, you once again felt it gives you the right to accuse me of making stuff up (another false accusation). I was hoping that you would finally admit you are not as smart as you think you are, but apparently not.

So, now It's time to school you and show everyone how wrong you really are. There is NO software in the world that can give you the ID number-batch code of a cpu, which is the ONLY number that can show if it is an ES or not. Not CPU-Z, not Sandra, not aida64, non of them can display that number. That is because that ID number/batch code is only located on the heat spreader of the cpu and is not obtainable thru software. There are the rare occasions that the manufacture will include that number in the specification label (your screenshot) but that is only because AMD chose to include it on that particular ES. There are hundreds of ES of every cpu, that do not include the CPU ID/batch codes in the specification label (hence why you don't understand why you lost this argument before it you ever decided to start it). Now, unless you can give us a screen shot or a video of CPU-Z removing a heat sink, wiping off the thermal past, reading the CPU-ID/batch code and displaying it on the screen, you have nothing to support your position. Having it included in the specification label by AMD is not the same thing, as CPU-Z cannot display such information on every CPU be it an ES or not. It has to be included in the specification label by AMD, which is rarely done.

The rest of your comment about the ID/batch code number is just to try and make yourself look smart to those that don't really understand that all of your wisdom is just thru google.

Example:

http://www.cpu-world.com/info/id/AMD-ES-identification.html

Thar CPU-world links for samples is incomplete and/or wrong. For instance it says for the first character in the ID

Sample type
1 ???
2 Prototype
Z ???

The decoding is:
1 = ES0 = first gen engineering sample.
2 = ES1 = second gen engineering sample.
3 = ES2 = third gen engineering sample.
...
Z = Qualification Sample.


And you continue making stuff up. I already demonstrated you are plain wrong when I gave you an example of CPU-Z showing the ID for an engineering sample. Of course CPU-Z isn't the only that does this.

Geeekbench also shows the ID of engineering samples

AMD-Zen-Naples-32-Core-Benchmark-740x489.png


Sandra also shows the ID of engineering samples

AMD-Ryzen-3000-Zen-2-Quad-Core-ES-Processor-1024x959.png


Userbench also show the ID of engineering samples

AMD-Ryzen-3000-Series-12-Core-CPU.png


It is evident some of you simply don't want to admit that the 5GHz LN2 overclock was made on a retail chip. So some of you will negate everything that points that it wasn't an engineering sample.

I stop here. No need to waste more time on this.
 
Thar CPU-world links for samples is incomplete and/or wrong. For instance it says for the first character in the ID



The decoding is:
1 = ES0 = first gen engineering sample.
2 = ES1 = second gen engineering sample.
3 = ES2 = third gen engineering sample.
...
Z = Qualification Sample.


And you continue making stuff up. I already demonstrated you are plain wrong when I gave you an example of CPU-Z showing the ID for an engineering sample. Of course CPU-Z isn't the only that does this.

Geeekbench also shows the ID of engineering samples

View attachment 169526

Sandra also shows the ID of engineering samples

View attachment 169527

Userbench also show the ID of engineering samples

View attachment 169528

It is evident some of you simply don't want to admit that the LN2 overclock was made on a retail chip. So you will negate everything that points that it wasn't an engineering sample.

I stop here. No need to waste more time.

I don't know if is is accurate, missing data, or not, as that wasn't my point, because I don't really care to decipher CPU ID/Batch codes, and I don't need to do it in an open forum to try and persuade people into believing I am smart. I was just showing examples of how easy it is to find such information using google, I wasn't debating it's accuracy, hence why it says "Example". Even your post from a year ago about CPU ID numbers was cut and paste off the internet. All you are proving is you know how to use google/copy/paste. Bravo! you are one step above a monkey.

Very Nice! More screenshots to show your incompetence and ability to use google. And more false accusations, you are the man.

All you are showing us, is the ability of CPU-Z, Sandra, etc to be able to display the specification Label which that label is and it's contents are controlled by the manufacture. They are not reading the CPU ID/Batch codes off the chip, as that is impossible. The only way they can get that number is if AMD includes it in the Specification label, which AMD only does for a handful of ES chips, not all of them. (The Chip used in the LN2 is a prime example)

To prove my point, please give us a screenshot of CPU-Z, sandra, Aida64, or what ever software you want, showing us the CPU ID/Batch code for the CPU you currently have in your computer, with this post in the background to verify it is a real screenshot. go ahead, we will wait...... no stealing screenshots off the internet using google.... no photo chopping allowed either.. sorry.......... Still waiting..... what? You can't? that's odd....... Oh wait, you can't without faking it, because software is not able to obtain a physical number that is only stamped onto the heat spreader. What that means is CPU-Z, Sandra, etc are only capable of displaying the specification Label, which on rare occasions includes the CPU ID number when it is a ES chip. NO amount of googled screenshots, or benchmark results is going to change that.

It has nothing to do with admitting anything. We simply do not believe you when you claim that the CPU used for the LN2 is not a ES chip (all based off a screenshot), because there is no way of knowing by the information in that screenshot. As well as the fact the person who has the chip already verified it was a ES chip, and during your sorry attempt to look "smart" you proved yourself wrong when you claimed it shows the OPN number.

The only verifiable way you can determine if the CPU used for the LN2 is a ES chip, is by the CPUID/Batch code that is stamped on it's heat spreader, not by the number you claim is the OPN number, that was included in the specification label of that particular cpu in that screenshot.

But please.. keep accusing me of making stuff up, and keep pounding those keys in your google searches.. it's comical to watch a desperate person fumble around like this.

In the end, we are only going in circles, as we won't really know until the retail chips are released, and we see the tests done on those chips, not on ES chips, or what we believe is an ES chip (excluding you of course)
 
Last edited:
What difference does it make if the overclock was done on a retail CPU or an engineering sample anyway? Engineering samples this close to release are pretty much retail silicon. Engineering samples are generally cherry picked, but they aren't any more capable than the retail CPU's in most cases. In fact, earlier CPU's are generally worse than retail chips.
 
Prove him wrong.

How? I've quoted him earlier in the thread saying different things at different times.

When he was trying to show that there was no room for improvement from QS to retail, he insisted, repeatedly, that clocks on QS are final.

Then we have what he first claimed was a QS chip, but it was much lower clocks than the announced retail.

So it changed to "clocks can be raised and lowered as needed for testing. This is a retail chip, so it's different"

Ignoring the fact that clocks can be changed during testing for ES, QS or retail.

Then there was the choice of Xeon for the TACC super computer.

"TACC already chose Cascade lake and rejected Rome"

Then when Zen 2 was picked for Frontier, which is planned to be substantially faster (1500 petaflop vs 40).

It changes to this...

"The goverment has a strategic policy of using different competitors for their computers. They already chose Intel before, so they had to chose AMD for Frontier. It is the same reason why they also order IBM supercomputers, not because Power is a beter CPU than Xeon."

His position and argument change constantly, he keeps moving the goalposts whenever he's wrong.

But you wouldn't know anything about that would you...


Room heat is real, and another 100w is felt. That's just the reality.

"Oh no, Intel is letting its CPUs run as fast as possible! Someone stop them!"

This behavior is desirable. Particularly if the cooling solution and power delivery can handle it, which most can; hell, what if it's running in a cooler environment?

The key here is stability, and that has not presented itself to be an issue. You can quibble about the marketing but the performance stands.
 
No, prove him right.

Not my fight. I just find it bordering on mini-modding if you're going to go from 'you're wrong' to 'you shouldn't post here'. Take it to the mods if you feel that his posts violate the rules. And be specific.

Or make your case. Or, be ridiculed for complaining.

How? I've quoted him earlier in the thread saying different things at different times.

Then keep quoting him and pointing out inconsistencies if it bothers you? Attack the argument, not the poster.

But you wouldn't know anything about that would you...

I know that there are a variety of perspectives. I run different systems each with different requirements, so if you take exception to something I say, be sure to present your argument in context.

You're pretending that computing in general and enthusiast computing specifically is one-dimensional. It isn't.

And again, be very specific to address the argument.
 
What a hilariously impressive absolutely fruitless and completely pointless argument we've spent the last x amount of days on. Thanks Juan. Hopefully later revision/steppings from whatever the hell the OP chip is gets us higher clocks. Whoop de doo.

........
What do you guys think about Jayztwocents predicting a price slide after launch giving us the 3900x for $379-400 between back to school sales and christmas and the 3950x for $499 by christmas time? I'm torn between holding on to my 4790k for a few more months and waiting for reviews, steppings, sales OR the value of my 4790k dipping hard and nullifying the effect of any future sales.... At launch it'll be a 3800x tops for me, but if he's right about $400 or less 3900x then I could do that in a couple months...
 
That's the point. There may be a little tiny bit more headroom come launch after binning but juan can't ever be wrong nor leave the dead, buried bone alone.

Given the reported clocks for that ES were slightly lower than retail, it might have a little more headroom on LN2 at launch. However, those of you hoping for 5GHz on water or an AIO are likely to be disappointed. I don't think that's realistic.
 
  • Like
Reactions: N4CR
like this
What a hilariously impressive absolutely fruitless and completely pointless argument we've spent the last x amount of days on. Thanks Juan. Hopefully later revision/steppings from whatever the hell the OP chip is gets us higher clocks. Whoop de doo.

........
What do you guys think about Jayztwocents predicting a price slide after launch giving us the 3900x for $379-400 between back to school sales and christmas and the 3950x for $499 by christmas time? I'm torn between holding on to my 4790k for a few more months and waiting for reviews, steppings, sales OR the value of my 4790k dipping hard and nullifying the effect of any future sales.... At launch it'll be a 3800x tops for me, but if he's right about $400 or less 3900x then I could do that in a couple months...
I don't think it will drop that fast. The 3950x is a shit load of cpu for the 750 they'll ask (750 right?)
 
Status
Not open for further replies.
Back
Top