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

[BrotherMichigan]

Not for AMD!

 

[Dan_D]

I haven't played with Ryzen at all yet. Does it typically allow you to set your own core count based overclocks?

You know, 5.2ghz for up to 4 cores loaded, 5.0ghz for up to 8 cores, 4.8 GHz for all core? Something like that?

I'd be happy with something like this, because generally when I need my best lightly threaded performance (gaming) my system isn't doing anything else.

I'd want to do it in the BIOS though, not in Ryzen Master or something like that. Most of the time I am not running Windows.

No. Existing Ryzen's also top out around 4.2GHz. Some can do a little more single core, but essentially, you can only hit 4.2GHz on all cores assuming you can keep them cool. With some chips like the Threadripper 2990XW, you are a bit more limited because of the insane heat and power draw the thing has. Aside from that one extreme example, Ryzen has always clocked roughly the same regardless of core count.

 

[GoodBoy]

...Desktop is a niche. The volume of sales is in mobile.

That may be true, but the majority of [H] readers are here for /desktop processor info/motherboards/ram overclocking/GPU overclocking/.. on desktops.

 

[Dan_D]

That may be true, but the majority of [H] readers are here for /desktop processor info/motherboards/ram overclocking/GPU overclocking/.. on desktops.

I hate to agree with him but it is actually true. Mobile and server sales are where the bread and butter is for Intel. That's primarily why Intel's been so focused on performance per watt for the last decade. That's where the money is and the mobile and server markets both share that same need. On the desktop, we only generally care about keeping the things cool and not so much about how much power it consumes. A CPU's power consumption isn't really a big deal unless its tied to a battery or you've got thousands of them in a datacenter. This above all else is why I think Intel was still peddling quad-core CPU's when AMD dropped the Ryzen hammer on their asses with eight cores and sixteen threads. This forced Intel to respond in the desktop market and provide us a desktop chip of sorts. That resulted in 4 core I3's, 6 core i5's and 8c/16t i9's.

 

[Dan_D]

Challenge accepted.

After saying that, I wondered if it could be done without simply adding radiators for no reason. With a couple of full cover EK water blocks for GPU's, a good CPU and motherboard block, I can see getting about half way to the $2,000 mark, but I can't really see going beyond that.

 

[BrotherMichigan]

I'm sure someone would mill me a few solid silver waterblocks if I asked kindly!

 

[Gideon]

I'm sure someone would mill me a few solid silver waterblocks if I asked kindly!

Silver is cheap enough for a project like that.

 

[GoodBoy]

Does using Silver prevent algae growth? Would it need to be a whole waterblock made of silver, or just a piece in the tank or something.

 

[BrotherMichigan]

It's more about thermal conductivity than its antimicrobial properties. At typical PC load temperatures, it's a little less than 10% more conductive than copper.

 

[Nolan7689]

Silver is cheap enough for a project like that.

I concur. At current spot prices $1000 of silver is about 5 lbs worth, or 2kg. You could easily make a full cover block for a GPU and CPU block with that amount.

 

[N4CR]

I concur. At current spot prices $1000 of silver is about 5 lbs worth, or 2kg. You could easily make a full cover block for a GPU and CPU block with that amount.

You could make maybe two of each depending on design. Especially if it was for Vega, probably more.

 

[Zarathustra[H]]

Even back in the Sandy Bridge days, the stock Intel cooler was pretty anemic. The thing was loud and the TIM really wasn't anything special. You also aren't going to get those sustained boost clocks very much using the stock cooler. Let's not be overly dramatic. You won't spend anything resembling $2,000 on an AIO of any kind. You won't even spend that on custom water cooling with multiple loops,

I find that there are hidden costs in building custom water loops that add more to the cost than one expects during the build process. Fittings and bends add a surprising amount to the bill.

The parts in my system, with a single loop and fan control add up as follows:

• Top Radiator: Alphacool XT45 420mm $117
  
• Front Radiator: Alphacool Monsta 280 $127
  
• GPU Block: EK Pascal Titan full cover block $135
• GPU Backplate: EK Pascal Titan $35
  
• CPU Block: EK Supremacy EVO Elite LGA2011 $126
  
• Reservoir/Pump Combo: XSPC D5 Photon 270 Reservoir / Pump Combo $165
  
• Fittings:12x XSPC G1/4" to 3/8" ID, 1/2" OD Compression Fitting Black Chrome V2 $66
  
• Tubing:PrimoFlex Advanced LRT 3/8in. x 1/2in. Tubing (10ft pack) - Onyx Black $26
  
• Sealing Plugs: 2x Enzotech Sealing Plug for 3/8" ID Tubing, Black (for filling on top and draining on bottom) $18
• Bends: Bitspower 2x 90 degree, 6x 45 degree ~$90
  
• Fan/Pump Controller: Aquacomputer 6 XT $199
  
• Temperature Probes: 4x Aquacomputer Calitemp Aquabus based Inline Probes $112
  
• Flow Meter: Aquacomputer $58
• Flow meter cable: Aquacomputer $8
• Coolant: 4x EK Cryofuel Concentrate $60

Now, I realize full well some of these parts are not strictly necessary, and some of them could have been done more cheaply, but this build totals $1342 in parts (each rounded to closest whole dollar, excluding shipping)

Add the fact that I swapped out the pump for a better/quieter one (EK D5 PWM G2), and then we can add another $90 to that.

Could I have done it for less? Certainly, but my point is that a $2k build is certainly not out of the question if someone goes for more radiators, dual pumps or multiple loops which I didn't.

 

[KazeoHin]

I concur. At current spot prices $1000 of silver is about 5 lbs worth, or 2kg. You could easily make a full cover block for a GPU and CPU block with that amount.

The issue with silver is that it tarnishes and degrades rather easily.

 

[Mr34727]

Math shows that the 1.344 V needed to hit 4.2 GHz with Zen2 cores is fully compatible (99.5% correlation) with the 1.608 V required to hit 5.0 GHz. It is also compatible with the voltage needed to hit 4.2 GHz with A72 cores in the same process node: 1.375 V.

"compatible"

 

[IdiotInCharge]

sounds more like they're just trolling Intel which they deserve the right to do after all the abuse they received during bulldozer.

AMD deseved every bit of the 'abuse' they received for Bulldozer.

-an AMD fan that was hoping they'd one-up Core 2 instead of going in the complete opposite direction

 

[Gideon]

The issue with silver is that it tarnishes and degrades rather easily.

Silver wont tarnish in the spots where air cant touch it, so the contact part with the cpu or gpu should be fine and I doubt you would see much wear on the fins in the coolant. Copper also tarnishes yet we still use it and thats mainly because it's a ton cheaper then Silver.

 

[NWRMidnight]

AMD deseved every bit of the 'abuse' they received for Bulldozer.

-an AMD fan that was hoping they'd one-up Core 2 instead of going in the complete opposite direction

At the time, Yes. But now that we see where that "failure" has lead AMD, was it a failure or an unavoidable step in getting to Ryzen as it inherits quite a bit from Bulldozer?

Reddit discussion on that subject:

 

[IdiotInCharge]

But now that we see where that "failure" has lead AMD, was it a failure or an unavoidable step in getting to Ryzen as it inherits quite a bit from Bulldozer?

It was a failure. Revisionism won't improve it, it's done and gone.

 

[Dan_D]

At the time, Yes. But now that we see where that "failure" has lead AMD, was it a failure or an unavoidable step in getting to Ryzen as it inherits quite a bit from Bulldozer?

Reddit discussion on that subject:

That's a complicated question that probably only AMD can truly answer. Ryzen doesn't inherit that much from Bulldozer. AMD literally tried to pull the same strategy Intel did when the Athlon 64 was kicking the Pentium 4 around in almost every benchmark and test you can think of outside of video editing tests. It went for a deeply pipelined architecture that would scale well in clock speeds at the expense of IPC and general efficiency. The difference was that Intel had the potential resources and every reason to think they were right, and AMD didn't. What AMD really banked on was that software would reach a level of parallelism on the desktop that would make Bulldozer shine as it would leverage what the chips were good at. The problem is that Intel dominates the landscape when it comes to software development. This is something Intel even mentions in its internal memos about AMD. Intel has great software and tools, but it also works with developers to make sure they are up to date on the processor technologies and how to leverage them. This is something AMD has historically been very bad at.

So it never made sense for AMD to think that software would change within a window that would have shown Bulldozer in a better light. The thing is, AMD wasn't entirely wrong as Bulldozer has aged surprisingly well and doesn't struggle as much on some newer software as its Intel counterparts of the era would. AMD was about 10 years too early for that architecture and I'm not even convinced some version of it would be the right call. Clearly, what AMD is doing with Ryzen is what makes sense now and for the foreseeable future. In contrast, I think Intel learned valuable lessons from the Pentium 4 and what the architecture was good at would make a return in later chips. The basic cache design and Hyperthreading being the two things that the P4 had going for it. Where Intel has made a mistake has been its sheer arrogance. It never thought AMD could rise to challenge them again. Intel also never considered building a desktop part on par with what we saw out of Ryzen and that's obvious by its reaction to Ryzen. Intel had in recent years concentrated on the mobile market and simply adapted its mobile parts for desktop use. HEDT parts were simply faster clocked server chips with some features disabled for segmentation.

The enthusiast market, and by extention the desktop market is something Intel doesn't and probably never has understood. The old desktop market, Intel understood fine but its transitioned into a niche market where all people use desktops for is content creation and gaming. Intel's own internal research shows this. Intel's hubris allowed AMD a window to make Ryzen a success and Intel's monolithic nature makes it hard for them to respond. Couple that with Intel's process problems and here we are.

 

[NWRMidnight]

It was a failure. Revisionism won't improve it, it's done and gone.

Success is built from failures.

 

[Derangel]

Success is built from failures.

Dude. Bulldozer was a failure. You don't need to pretend it wasn't or try to sugar coat it. AMD fucked up, big time. There is nothing wrong with admitting it. They're doing great right now, but don't pretend their failures didn't happen.

 

[NWRMidnight]

Dude. Bulldozer was a failure. You don't need to pretend it wasn't or try to sugar coat it. AMD fucked up, big time. There is nothing wrong with admitting it. They're doing great right now, but don't pretend their failures didn't happen.

Excuse you? Please show me where I pretended it never happened, or sugar coated it. All I am saying is that failure played a part in the creation of Ryzen.
My apologies however, I didn't mean to push your buttons, as I didn't realize you where so sensitive about the subject.

 

[Derangel]

Excuse you? Please show me where I pretended it never happened, or sugar coated it. All I am saying is that failure played a part in the creation of Ryzen.

But now that we see where that "failure" has lead AMD, was it a failure or an unavoidable step in getting to Ryzen as it inherits quite a bit from Bulldozer?

You trying to deny that it was a failure. That what it means when you put something in quotes like that. Not to mention you trying to deny it was a failure and some kind of planned/required step.

Success is built from failures.

You sugar coating it.

 

[funkydmunky]

[QUOTE="Derangel, post: 1044245145, member: 175884"You sugar coating it.[/QUOTE]
You dusting it off?

 

[funkydmunky]

[QUOTE="Derangel, post: 1044245145, member: 175884"You sugar coating it.

You dusting it off?[/QUOTE]
Apparently something is F'ed here. WT [H]?

 

[NWRMidnight]

You trying to deny that it was a failure. That what it means when you put something in quotes like that. Not to mention you trying to deny it was a failure and some kind of planned/required step.

You sugar coating it.

Well, you can interpret it any way you like, but I am telling you, that you are wrong, I quoted "failure" to emphasize in a Ryzen discussion, that it's failure possiblely lead us to Ryzen. Hence why I asked if it was an unavoidable step. I didn't say it was planned or required. You are trying to define what I said to mean something it's not.

Let me word it differently, or ask this question:. Would we have Ryzen if it wasn't for the failure called Bulldozer? Hence why I said success is created from failures, because I don't think we would have Ryzen without bulldozer. We would have something else in it's place possibly better, or possibly worse. It all has to do with ones perspective. Has nothing to do with denying anything or sugar coating it.

 

[GoodBoy]

....Everything Dennard scaling slow down is referring to involves transistor count increases. Ryzen won't have any transistor count increases until Ryzen 3 using 7nm + node....

No, you are misunderstanding Dennard scaling, which has to do with process shrinks/transistor sizes and therefore power density, not inherently the amount of transistors. So whether it has 1 billion transistors or 10 billion, the issues are the same.

Also you are misunderstanding Moore's Law, which states transistor density per area X doubles ~ every 2 years, which is unrelated to the total transistor count any particular chip may have.

Moore's law ended ~2008.

We still are managing to shrink things, but at a much slower rate, and will stop at some point.

 

[NWRMidnight]

No, you are misunderstanding Dennard scaling, which has to do with process shrinks/transistor sizes and therefore power density, not the amount of transistors. So whether it has 1 billion transistors or 10 billion, the issues are the same.

No, I understand Dennard scaling, but we are talking about the breakdown. and the cause is the part I don't think you understand. It isn't that the die shrinkage (transistor shrinkage) is getting to small, it is the increasing number of transistors that is being done along with the die shrinkage that is causing the breakdown. As the die's are being shrunk, transistor count are also being increased which basically nullifies Dennard scaling, hence the breakdown. If we just did a die shrinkage with no increase in transistors, which is what is going on with Ryzen, Dennard scaling isn't being broken down. Basically, Moore's law is causing the breakdown, which states that transistor count is doubling every 2 years. In the case with Ryzen from Generation 1 thru Generation 3, Moore's law isn't applied because there is zero increase in transistor count. An increase won't happen till Generation 4.

And no, Moores law does not state that the transistor density doubles per area X every 2 years.. from your Dennar Scaling wiki:

Moore's law says that the number of transistors doubles about every two years. Combined with Dennard scaling, this means that performance per watt grows at this same rate, doubling about every two years. This trend is referred to as Koomey's law. The rate of doubling was originally suggested by Koomey to be 1.57 years[4] (somewhat faster than the doubling period of Moore's law), but more recent estimates suggest this is slowing.

It specifically states that the number of transistors doubles. It actually doesn't say anything about density.

Even the Moore's Law wiki states:

Moore's law is the observation that the number of transistors in a dense integrated circuit doubles about every two years

Example: lets take a circuit, and lets say it has 1 billion transistors. 2 years later, per Moore's law, that same circuit now has 2 billion transistors. 2 years later it is 4 billion, then 8. It increases exponentially.

In a sense, I think we are saying the same thing but in a different way about Moore law. But with Ryzen, the same circuit has the same number of transistors, just smaller. There is no increase in transistor count.

Power draw has to do with Koomey's law, not Dennard scaling. Koomey's law is the combination of Moores Law and Dennar scaling (per the quote above)

 

[GoodBoy]

There are a lot of bad definitions out there, but it is a doubling per area. Density is a factor:

(n.) Moore's Law is the observation made in 1965 by Gordon Moore, co-founder of Intel, that the number of transistors per square inch on integrated circuits had doubled every year since the integrated circuit was invented.

"per square inch" is density.

The doubling per area at a rate of one doubling every 2 years, ended ~ 2008.

It's dead, Jim.

****

Edit: after looking around and reading varying definitions of Moore's Law, I can see why you think it's transistor count.

(Moore observed) One component (1959), 32 (1964), 64 (1965)-Moore put these numbers on a graph and connected the dots with a line. “The complexity [of cheap integrated circuits] has increased at a rate of roughly a factor of two per year,” he wrote.

Note that the above rate is a doubling every single year! Kinda crazy.

Then in 1975 "Moore slightly revised his “law.” From that point on, he said, the number of devices on a chip would double every two years."

[I note that here, area was not specifically mentioned]

But it isn't just transistor count. Other more loose definitions state it as a "doubling of computing power every 2 years".

By any of these measures, it is dead. As you stated, Ryzen had a die shrink with essentially the same transistor count. If area(density) is not a part of the law (and I've found it stated both with and without the area specified as pointed out above), then it just further proves Moore's Law is dead. We haven't had a new cpu come out that was "double the computing power" in some time.. and to try to counter it, chipmakers are adding more cores, but that doesn't help all programs. The number of transistors per chip isn't doubling either, and really, without a defined area to measure against the metric isn't that useful. we aren't going to have GPU's with 500 billion transistors on a 10 inch square wafer of silicon as its just not feasible or affordable, and I don't think that was what Moore meant by the law (hence the area really does matter moreso than just a total overall transistor count, imho).

Transistors per area - that stopped doubling ~ 2008
Transistors per chip, allowing the die to get bigger - this isn't happening every 2 years either. 5 billion in 2012 (61 core xeon), the first time we see 10 billion is a 2016 sparc chip, then 19 in 2017 (Epyc). But that isn't doubling every 2 years or we would have >45 billion transistor chips now.
Computing power - this definitely has slowed after the first i7's come out (which were the last, nice huge performance jump, someone correct me if I am wrong). And we get maybe 10 or 20% performance increase every 2 years (CPU's anyway). GPU's are doing better but have also slowed..

I mighta got sidetracked there.. but I will concede to you that the definition of Moore's Law is somewhat ambiguous, so I retract my statement that "you misunderstand it"

 

[NWRMidnight]

There are a lot of bad definitions out there, but it is a doubling per area. Density is a factor:


"per square inch" is density.

The doubling per area at a rate of one doubling every 2 years, ended ~ 2008.

It's dead, Jim.

https://en.wikipedia.org/wiki/Moore's_law It's alive Jim! It all depends on interpretation, and we really have no way of proving which interpretation is correct. Science, scientific laws, etc, are proven and disproved constantly due to peoples different interpretation as well as advancements in technology.

Is density a factor or is it a single circuit that is a factor? If the transistor count does not increase in a given circuit, or group of circuits, but the transistors of the circuit are shrunk, reducing the amount of power needed, what does that mean? If the number of transistors are increased in a circuit, what does that mean? The reason that speed increases are usually present with a die shrink is because transistor count are increases within the same circuit or group of circuits to achieve the higher clocks. If there is a performance increase with a die shrinkage, but no transistor count increase (which reduces power draw) no change in number of circuits, what does that mean? Where is AMD getting their speed increases, lower temperatures, etc if your perspective is true with no transistor increases in their cpu. how are they able to give us higher clocks with less heat, using less TDP then the 12 nm with no change in transistor count, if Dennar scaling is broken and Moores law is dead? Some things just don't make sense if such laws are written in stone.

Edit: I just saw your additions to your comment. I also understand your take on it, I just happen to think outside the box or view things differently than most people, hence why I end up having discussions such as this. At the end of the day, this type of discussions, be it agreeing or disagreeing is what Science and technology is all about and makes it exciting. Where we are just waiting for someone to apply a theory in a different way, resulting in a break thru and/or new technologies.

Here is a questions that just popped into my head: Is the transistor count on the Ryzen based off the whole cpu chip, per core, or per chiplet? Wouldn't additional cores/chiplets increase transistor count? So how is it stated that Ryzen, all 3 generations so far, have the same transistor count on the wiki?

 

[Zarathustra[H]]

At the time, Yes. But now that we see where that "failure" has lead AMD, was it a failure or an unavoidable step in getting to Ryzen as it inherits quite a bit from Bulldozer?

Reddit discussion on that subject:

I wasn't aware that the Zen architecture was related to Bulldozer. I was under the impression they essentially threw out the entire thing and used Jim Keller to create a new, mostly from scratch design with SMT.

What are you basing that on?

 

[NWRMidnight]

I wasn't aware that the Zen architecture was related to Bulldozer. I was under the impression they essentially threw out the entire thing and used Jim Keller to create a new, mostly from scratch design with SMT.

What are you basing that on?

I didn't mean to imply it was related to Bulldozer., but we really don't know, as AMD doesn't discuss such things. But there sure are a lot of similarities that makes a person wonder if Bulldozer never happened, would we have the Ryzen architecture as we know it today, or would Ryzen even exist? Did any part of the Bulldozer architecture influence Ryzen's architecture? Was it's failure a stepping stone to get to Ryzen? I was just asking a question, that caused some peoples panties to get all bunched up.

 

[IdiotInCharge]

I didn't mean to imply it was related to Bulldozer., but we really don't know, as AMD doesn't discuss such things. But there sure are a lot of similarities that makes a person wonder if Bulldozer never happened, would we have the Ryzen architecture as we know it today, or would Ryzen even exist? Did any part of the Bulldozer architecture influence Ryzen's architecture? Was it's failure a stepping stone to get to Ryzen? I was just asking a question, that caused some peoples panties to get all bunched up.

You implied a position which is wholly untrue, so yeah, you were called out.

 

[NWRMidnight]

You implied a position which is wholly untrue, so yeah, you were called out.

False, and already clarified. Just because you and two others chose to read into it that way doesn't make it true. In fact, if you look below your comment, you will see another poster who understood what I meant, and actually answered the question. he also quoted my whole comment, and did not just cherry pick part of it and ignoring the rest to make it appear I implied something I didn't. So, rather than answering the question, you and a couple other's chose to latch onto one word in quotes, without taking in the full context of the comment, or taking a look at the Reddit discussion, and got all butt hurt over it. Besides, even if I did have the opinion that Bulldozer wasn't a failure , I have that right without being attacked or called out for having such opinion. Just as you have the right to disagree.

 

[BrotherMichigan]

Don't you know? Having a different opinion is the same as commiting physical violence these days!

 

[IdiotInCharge]

Besides, even if I did have the opinion that Bulldozer wasn't a failure , I have that right without being attacked or called out for having such opinion.

You're on a public forum, so no you don't. Posts fall within the bounds of the rules or they don't. I recommend refraining from trolling or otherwise trying to trigger a response.

To answer the question (if there was one) more directly, Bulldozer is why we almost didn't have Ryzen, why it took so long to get here, and part of the reason why its release was so rocky vs. say Core 2 and following architectures.

 

[NWRMidnight]

You're on a public forum, so no you don't. Posts fall within the bounds of the rules or they don't. I recommend refraining from trolling or otherwise trying to trigger a response.

To answer the question (if there was one) more directly, Bulldozer is why we almost didn't have Ryzen, why it took so long to get here, and part of the reason why its release was so rocky vs. say Core 2 and following architectures.

A public forum governed by rules that says otherwise. I suggest you go read the rules again, as well as the warning already placed in this topic. I also suggest you go read my comment again in it's entirety, as the question is plain as day. Correcting false implications, cherry picking, and manipulation of what I said is not trolling, nor is calling someone out for making personal attacks. I am a blunt, straight to the point person. Sorry that it rubs you the wrong way.

If Bulldozer is why we almost didn't have Ryzen, why did Ryzen's development start in August of 2012, 10 months after the release of Bulldozer? If you said the lack of R&D budget delayed the start of Ryzen's development, that would be accurate. But I haven't seen any evidence other than your opinion above, that shows Bulldozer caused us to almost not have Ryzen. CPU development cycle's are around 4 years on average. Ryzen's development cycle was 4.5 years from day one of development to release, so it didn't take any longer than normal, with a very small R&D budget to boot. Some of Intel's development are 10 years long, with a huge R&D budget. A rocky release is expected after releasing something that many believe was a failure.

 

[IdiotInCharge]

If Bulldozer is why we almost didn't have Ryzen, why did Ryzen's development start in August of 2012, 10 months after the release of Bulldozer?

Because AMD almost went out of business on the way.

 

[NWRMidnight]

Because AMD almost went out of business on the way.

That wasn't due to Bulldozer, that was actually caused by Intel dirty business tactics that hurt competitors, the antitrust and patent lawsuits with Intel (settled in 2009), and the purchase of ATI in 2006, which left them strapped for cash and little money for R&D the following years.

 

[IdiotInCharge]

That wasn't due to Bulldozer, that was actually caused by Intel dirty business tactics that hurt competitors, the antitrust and patent lawsuits with Intel, and the purchase of ATI during the same time frame.

Making a terrible product that is less competitive than its predecessor absolutely contributed to their near-bankruptcy experience.