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

Status
Not open for further replies.
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...

No chance in hell of those prices hitting before Christmas, unless Intel drops their prices by like 30+% across the board or these bomb super hard. There is no way they will knock $250 off the 3950x or $100 off the 3900x. If they do a sale like that I could see, at most, $20-$30 cheaper.
 
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.
Agreed. I still think late mid to high 4's with a good chip on a good water rig
5Ghz single core? hard to call at this point.
 
Agreed. I still think late mid to high 4's with a good chip on a good water rig
5Ghz single core? hard to call at this point.

Probably the absolute ceiling for now. Maybe possible on mild subambient cooling?

Right now I'd like to see what a good 240mm AIO will do.
 
Probably the absolute ceiling for now. Maybe possible on mild subambient cooling?

Right now I'd like to see what a good 240mm AIO will do.

Agreed - chilled water would be interesting to see if helps much, might be e.g. IO die needs more cooling.. bit like the 'hot spot' temp on Vegas. Who knows.
Don't feel it'll go any more. TSMC estimations were around 4.5-4.6 if extrapolating (so were mine) and it appears AMD reached there. Extreme end I think won't be over 5.
240mm would be neat to see especially on the 8 core parts.
 
Agreed - chilled water would be interesting to see if helps much, might be e.g. IO die needs more cooling.. bit like the 'hot spot' temp on Vegas. Who knows.
Don't feel it'll go any more. TSMC estimations were around 4.5-4.6 if extrapolating (so were mine) and it appears AMD reached there. Extreme end I think won't be over 5.
240mm would be neat to see especially on the 8 core parts.

Looking at a new build with the 12 core and a 240mm AIO, probably on an ASUS board with performance boost. I would love to see low to mid 4s on all cores.
 
More leaked benchmarks on the usual grain of salt rumor sites, 9900k still beats the 6 core one on single threaded performance.

14+++++++++++++++++ is legendary.
 
Looking at a new build with the 12 core and a 240mm AIO, probably on an ASUS board with performance boost. I would love to see low to mid 4s on all cores.
Me too! As an 'entry wish' it's pretty realistic I feel. More of a big air guy though haha, love air cooling. N/A of the computer world.
Also interested to see how it scales e.g. 4 cores, 2 cores with clocks. Very curious and can't wait to play with the new memory bus.
 
Elchapuzasinformatico review of the R5 shows 1.344 V needed to hit 4.2 GHz. This is fully compatible with the 1.608 V required to OC the R9 to 5.0 GHz

AMD-Ryzen-9-3950X-5-GHz-Overclock-1030x576.png
 
Last edited:
Elchapuzasinformatico review of the R5 shows 1.344 V needed to hit 4.2 GHz. This is fully compatible with the 1.608 V required to OC the R9 to 5.0 GHz

View attachment 169966

His review has zero overclocking in it (it is all done stock), it is all done on a x470 board, not a x570, with the board controlling the voltages (set on auto) and is all done on reference cooler (air). Auto voltage on Ryzen has always ran the CPU at higher voltages than needed. Hence, why most users take control of their voltages and under volt (the 2700x can hit as high as 1.55v when set to auto, yet, when the user takes control of the voltage, that drops down to 1.3 to 1.38 on a 2700x under full load). So, no, they are not comparable, not even close. If the R5 can do 4.2 ghz with only 1.344 volts on air with volts with auto voltage, imagine what it can do on LN2 with the voltages manually controlled). (I already linked the video where gamer nexus states that they are hitting near 6 Ghz with the lower core count parts). However, the 4.2 Ghz is only it's boost clocks. For the record, my 2700x can do 4.15 Ghz all cores stresstesting with no manual overclocking (all boosting controlled by the motherboard/auto) and with 1.35v with it under volted by .100v. It does 4.35Ghz on 3 cores for single core boost. If I set my ram to 2100 from 3200, it can hit as high as 4.6 Ghz single core boost with just the under volt, but that is only for a split second and settles around 4.45 Ghz. So ram speeds play a part in clock speeds. This using water cooling.


So again, no his review is not comparable.
 
Last edited:
  • Like
Reactions: N4CR
like this
1.6V has to be very high for 7nm...

Look at the ram speed, they are using a early beta bios which had issues, Id wait to see some actual reviews with a solid platform before drawing a conclusion on voltage needed. But expect anything at 5GHz to take some voltage, just a matter of how safe that voltage is for 24/7 use.
 
burn a whole through the socket if not cooled by LN2 or dry ice

:D

This is [H] so why would that be a issue, hell I have a phase change cooler in the closet if it was needed. Likely water needed for those clocks anyway or even chilled water but at least we know it's possible but 1 sample is hardly anything to to draw conclusions on yet is my point.
 
His review has zero overclocking in it (it is all done stock), it is all done on a x470 board, not a x570, with the board controlling the voltages (set on auto) and is all done on reference cooler (air). Auto voltage on Ryzen has always ran the CPU at higher voltages than needed. Hence, why most users take control of their voltages and under volt (the 2700x can hit as high as 1.55v when set to auto, yet, when the user takes control of the voltage, that drops down to 1.3 to 1.38 on a 2700x under full load). So, no, they are not comparable, not even close. If the R5 can do 4.2 ghz with only 1.344 volts on air with volts with auto voltage, imagine what it can do on LN2 with the voltages manually controlled). (I already linked the video where gamer nexus states that they are hitting near 6 Ghz with the lower core count parts). However, the 4.2 Ghz is only it's boost clocks. For the record, my 2700x can do 4.15 Ghz all cores stresstesting with no manual overclocking (all boosting controlled by the motherboard/auto) and with 1.35v with it under volted by .100v. It does 4.35Ghz on 3 cores for single core boost. If I set my ram to 2100 from 3200, it can hit as high as 4.6 Ghz single core boost with just the under volt, but that is only for a split second and settles around 4.45 Ghz. So ram speeds play a part in clock speeds. This using water cooling.

So again, no his review is not comparable.

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.
 
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.

Do you really believe taking 1.344v and dividing it by 4.2, then mulipling it by 5 which equals 1.6v, is how you determine what voltage it will take for 16 cores to hit 5 Ghz? Uh.. No! Even if that was an accurate way to figure what voltage was needed to hit 5 Ghz (it isn't), it would only be comparable to another 6 core cpu, not a 16 core cpu, as the extra cores changes the equation. Considering it takes form 1.25v to 1.35v to hit 4.2 Ghz on all cores on Ryzen 2000 (2700x) using AIO cooling, the Die shrink alone means it will take less voltage on Ryzne 3000 to hit 4.2, So it is fair to say that your 1.344v staring voltage is incorrect. The test was done on a x470 with a early bios, which adds to the core voltage of 1.344 being inaccurate , even if your math formula was correct, which it isn't. You also have to figure in the motherboards power delivery system, VRM capabilities etc. of the motherboard being used. Considering that the bios was unstable and they couldn't attempt to overclock (it would crash), I would take the advice of Gideon and wait before coming to that conclusion no matter what math equation you want to use.
 
Last edited:
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.

If you believe voltage requirements scale linearly with clockspeed, you've never overclocked before
 
Do you really believe taking 1.344v and dividing it by 4.2, then mulipling it by 5 which equals 1.6v, is how you determine what voltage it will take for 16 cores to hit 5 Ghz? Uh.. No! Even if that was an accurate way to figure what voltage was needed to hit 5 Ghz (it isn't), it would only be comparable to another 6 core cpu, not a 16 core cpu, as the extra cores changes the equation.

The function relating voltage to frequency doesn't depend on the number of cores.

Considering it takes form 1.25v to 1.35v to hit 4.2 Ghz on all cores on Ryzen 2000 (2700x) using AIO cooling, the Die shrink alone means it will take less voltage on Ryzne 3000 to hit 4.2, So it is fair to say that your 1.344v staring voltage is incorrect. The test was done on a x470 with a early bios, which adds to the core voltage of 1.344 being inaccurate , even if your math formula was correct, which it isn't.

4.2GHz is far beyond the optimal range for 12LP,

The scaling and power efficiency is excellent until 3.5GHz and reasonable scaling takes place until 4.0GHz. Beyond this point the voltage scaling is anything but linear, and the CPU is requiring higher voltage in increasingly larger steps to scale further. For example, 4.1GHz requires > 112mV higher voltage than 4.0GHz and the difference in voltage results in 41.5W (or 34.3%) increase in power consumption.

Dennard scaling ceased to work around 2006. No longer a die shrink reduces the voltage.

I have given voltages for X470 boards, for X570 boards, and for non-AM4 boards.
 
Last edited:
burn a whole through the socket if not cooled by LN2 or dry ice

:D

Volts Dont burn, Amperage burns and 1.6v is barely enough to tickle your tongue, or should I be worried that my 230v plug output is going to burn my house down?

The only possible way to burn is if a capacitor is leaking and that ignites mostly due to heat or an arc but motherboards are just not that crap anymore
 
Dennard scaling ceased to work around 2006. No longer a die shrink reduces the voltage.

While it may be true that a die shrink doesn't necessarily reduce the voltage, it has certainly happened since 2006. As recently as 2011 for Intel, Vdd was 0.75V for their 22nm process (it was the same for 32nm), which dropped to 0.70V for 14nm and 10nm.
 
Commence the 3 page debate on how dangerous 1.5v is in modern electronics. Or maybe like TPD, AMD and Intel have different measures and tolerances.

*Popcorn*
All I was going to say is that Volts Amperage and Wattage are all tied in to each other. Sure 1.5 isn’t enough to tickle but go ahead and use cheese for your thermal paste and check them temps. :ROFLMAO:
 
Its got its merits....but dam just running the division 2 with that cpu and nothing else running probably keeps that cpu pegged at 100% usage lol forget multitasking?
Im running a 5820K and 980 Ti and have 3 monitors and usually have chrome open and firefox for youtube while playing Div2. I don't have an performance issues that I have seen.
 
Volts Dont burn, Amperage burns and 1.6v is barely enough to tickle your tongue, or should I be worried that my 230v plug output is going to burn my house down?

Nolan7689 beat me to it, but this is still worth highlighting out of bare ignorance. The handles on a CPU are Volts and Frequency, which when raised raise Amps and then Watts. Meaning that if you raise the voltage so that you can raise frequency, you're also increasing the amperage significantly.

And given that it took LN2 to get the CPU to 5.0GHz across all cores, well, the amperage involved was quite likely enough to set something on fire if improperly routed :D
 
Nolan7689 beat me to it, but this is still worth highlighting out of bare ignorance. The handles on a CPU are Volts and Frequency, which when raised raise Amps and then Watts. Meaning that if you raise the voltage so that you can raise frequency, you're also increasing the amperage significantly.

And given that it took LN2 to get the CPU to 5.0GHz across all cores, well, the amperage involved was quite likely enough to set something on fire if improperly routed :D
There are so many caveats like Intel have higher current (A) ratings, and how a 1.1 vCore turns out +100 degrees. It all comes down to how AMD and Intel don't see voltage et Al nor TDP the same yet AMD CPUs run cooler and less power as well as come with really good stock fan which is a massive factor.

Sure it is more likely to have an all core around 4.5ghz but that alone is 700mhz higher than a 2700 so I'm cool with that and the performance from Tre Mexicans seems to show fantastic stock clock performance
 
There are so many caveats like Intel have higher current (A) ratings, and how a 1.1 vCore turns out +100 degrees. It all comes down to how AMD and Intel don't see voltage et Al nor TDP the same yet AMD CPUs run cooler and less power

This depends entirely on the CPUs compared and the settings and testing environment. Should be specific to the high-end consumer parts, and should not be discussing temperature as that is only useful for comparing CPUs of the same configuration, at best, and really should only be used to compare performance of a single individual part in terms of 'more' and 'less'.

Sure it is more likely to have an all core around 4.5ghz but that alone is 700mhz higher than a 2700 so I'm cool with that and the performance from Tre Mexicans seems to show fantastic stock clock performance

I agree that the potential is there, and an all-core of 4.5GHz is nothing to sneeze at, but we do need to wait for more detailed reviews.
 
This depends entirely on the CPUs compared and the settings and testing environment. Should be specific to the high-end consumer parts, and should not be discussing temperature as that is only useful for comparing CPUs of the same configuration, at best, and really should only be used to compare performance of a single individual part in terms of 'more' and 'less'.



I agree that the potential is there, and an all-core of 4.5GHz is nothing to sneeze at, but we do need to wait for more detailed reviews.

Clock speed fixation is lame, Sandy bridge ran 3.3-3.7ghz and was effectively faster than a 5ghz FX, AMD is squeezing massive performance out of lower clockspeed again that is more efficient.

The 8700/9900 only exists because Intel is hellbent on trying to stay ahead but in the process through efficiency out the window. When I buy a CPU I don't want to pop the lid off and put some third party ice cream on, nor do I want to spend 2k for a closed loop 240 rad cooler, I want to pop it on the board with cooler in the box and Bazinga get to making my music.
 
Clock speed fixation is lame

Clock speeds are real, and will continue to be a major part of CPU performance.

When I buy a CPU I don't want to pop the lid off and put some third party ice cream on, nor do I want to spend 2k for a closed loop 240 rad cooler, I want to pop it on the board with cooler in the box and Bazinga get to making my music.

Well, that's you. I still wouldn't use a cooler from a CPU vendor because I know that it's built to a price, outside of a function test. I have plenty of old coolers for all sockets laying around to do that today if needed.

And if I were buying an enthusiast part, which for Intel is anything with a 'K' or 'X', I'd be getting some form of improved cooling.

I like that AMD is putting out some above average coolers, but the reality is that if you're going to actually overclock them to the point where they'd even get close to the single-core performance of an Intel CPU, you're going to need quite a bit more cooling and you'll be pulling quite a bit more power.
 
Clock speed fixation is lame, Sandy bridge ran 3.3-3.7ghz and was effectively faster than a 5ghz FX, AMD is squeezing massive performance out of lower clockspeed again that is more efficient.

Back then people, including me were overclocking Sandy Bridge between 4.8GHz - 5.0GHz. Clock speed "fixation" isn't the issue. Clock speed, is a measurement of performance. It isn't the only one and all things being equal, architectural efficiency matters more, but when architectures are equal, clock speed is king. Higher clock speeds with any given architecture are faster. Period. That's why overclocking has pretty much always existed.

The 8700/9900 only exists because Intel is hellbent on trying to stay ahead but in the process through efficiency out the window.

You seem to love hyperbole. For 14nm, the 9900K is actually quite amazing. Until Ryzen 3000, it was absolutely the fastest CPU on the market in terms of IPC and clock speeds. Barring applications that can scale past 8c/16t, its still the fastest processor you can buy today.

When I buy a CPU I don't want to pop the lid off and put some third party ice cream on, nor do I want to spend 2k for a closed loop 240 rad cooler, I want to pop it on the board with cooler in the box and Bazinga get to making my music.

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,
 
Clock speeds are real, and will continue to be a major part of CPU performance.



Well, that's you. I still wouldn't use a cooler from a CPU vendor because I know that it's built to a price, outside of a function test. I have plenty of old coolers for all sockets laying around to do that today if needed.

And if I were buying an enthusiast part, which for Intel is anything with a 'K' or 'X', I'd be getting some form of improved cooling.

I like that AMD is putting out some above average coolers, but the reality is that if you're going to actually overclock them to the point where they'd even get close to the single-core performance of an Intel CPU, you're going to need quite a bit more cooling and you'll be pulling quite a bit more power.


Clockspeed is a means to achieve performance much like how 5ghz FX were needed just to try keep up to Sandy bridge running lower clocks, now it seems AMD have gone with efficient uArch and Intel is bleeding efficiency for clockspeed to try convince you that a 9900K is better than a 6700K. If it so happens to be true that a 4.2ghz AMD CPU runs with Intel in gaming, I am sure that will give some people the fanny wobbles
 
Clockspeed is a means to achieve performance much like how 5ghz FX were needed just to try keep up to Sandy bridge running lower clocks, now it seems AMD have gone with efficient uArch and Intel is bleeding efficiency for clockspeed to try convince you that a 9900K is better than a 6700K. If it so happens to be true that a 4.2ghz AMD CPU runs with Intel in gaming, I am sure that will give some people the fanny wobbles

Umm...a 9900K is better than a 6700K. While the changes aren't huge, it is an improved version of Skylake. It has more cores, more threads, and it clocks higher. How would it not be better? Architectural efficiency and clock speed aren't separate. Take any architecture and increase its clock speeds and you will get more performance. It's that simple.
 
Back then people, including me were overclocking Sandy Bridge between 4.8GHz - 5.0GHz. Clock speed "fixation" isn't the issue. Clock speed, is a measurement of performance. It isn't the only one and all things being equal, architectural efficiency matters more, but when architectures are equal, clock speed is king. Higher clock speeds with any given architecture are faster. Period. That's why overclocking has pretty much always existed.



You seem to love hyperbole. For 14nm, the 9900K is actually quite amazing. Until Ryzen 3000, it was absolutely the fastest CPU on the market in terms of IPC and clock speeds. Barring applications that can scale past 8c/16t, its still the fastest processor you can buy today.



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,


Back then Sandy benchmarks were having stock 2500K's beating FX8150's OC'd to 5ghz, and people were unhappy at how AMD needed to sacrifice efficiency for performance, we can even go to the Pentium 4 days in how 3ghz P4's were being destroyed by much lower clocked Hammers on performance and efficiency. Clockspeed is a PC persons version of a sports car

The 9900K is the fastest, because it is clocked to 5ghz, comparing non K or X SKU's to AMD's with the more normal clocks similar to AMD's they are not really that impressive, so clockspeed actually is the way Intel are staying ahead and it has resulted in efficiency being sacrificed as now it is Intel that are the "it runs hot" joke.

I did not say 2K is US dollars, I am not from the US and yes it costs 2200 bucks in my country for a Corsair H115i, pretty much the same price as a Ryzen 5 2600 or a kit of 16GB trident RGB 3200mhz.
 
now it seems AMD have gone with efficient uArch and Intel is bleeding efficiency for clockspeed to try convince you that a 9900K is better than a 6700K. If it so happens to be true that a 4.2ghz AMD CPU runs with Intel in gaming

Not really.

And your claim of not just IPC parity, but of a significant IPC advantage for Ryzen 2 is quite laughable.

I am sure that will give some people the fanny wobbles

I am sure that you think your emotional quips are cute, but they're not. They're trolling.
 
I did not say 2K is US dollars

You didn't list any currency while posting on a US forum, so yes, you did say US dollars by omission. If you mean some other currency, of which there are hundreds in the world, then specify that. We're not going to guess, we're just going to laugh.
 
  • Like
Reactions: Dan_D
like this
Back then Sandy benchmarks were having stock 2500K's beating FX8150's OC'd to 5ghz, and people were unhappy at how AMD needed to sacrifice efficiency for performance, we can even go to the Pentium 4 days in how 3ghz P4's were being destroyed by much lower clocked Hammers on performance and efficiency. Clockspeed is a PC persons version of a sports car

The 9900K is the fastest, because it is clocked to 5ghz, comparing non K or X SKU's to AMD's with the more normal clocks similar to AMD's they are not really that impressive, so clockspeed actually is the way Intel are staying ahead and it has resulted in efficiency being sacrificed as now it is Intel that are the "it runs hot" joke.

I did not say 2K is US dollars, I am not from the US and yes it costs 2200 bucks in my country for a Corsair H115i, pretty much the same price as a Ryzen 5 2600 or a kit of 16GB trident RGB 3200mhz.

Are you kidding me? Prior to the security issues, Intel still had a substantial lead on AMD in terms of IPC. That means clock for clock its faster. Period. It also clocks higher. The only reason why Ryzen holds its own in some applications is because it had more cores. When Ryzen launched, Intel was still selling us quad core CPU's. AMD often had twice the cores in a given price point and it was close enough in clocks and IPC that it would perform better in some tests. AMD has found success with Ryzen for two reasons. 1.) Price vs. performance. 2.) Due to offering more cores for the same or less money, AMD has found performance advantages in some cases. Intel isn't sacrificing efficiency to get faster. Its architectures have continued to improve on a clock for clock basis even if those improvements are small. Yet, clock speeds are going up. That's due to process maturity.

I've said it before, you have to look at the reasons why a given CPU wins a given benchmark to understand what's going on and to evaluate things in the correct context. AMD wins when and where more cores are a benefit. I do the benchmarks and work with the data all the time. That may certainly change with Zen 2. I don't have one on hand yet and haven't done the testing. Early indications are that AMD is going to close the gap significantly. Its also worth pointing out that AMD has increases Ryzen 3000's clocks a fair amount over the preceding generations of Ryzen CPU's. That's part of closing that gap with Intel.
 
Status
Not open for further replies.
Back
Top