Ryzen2 may well hit 5GHz on 7nm says GF CTO

Just a taste of 2700X IPC improvment (tested locked @ 4Ghz)

source:

no, I think he means IF they add a chiller it may do 4.9-5.0.

They use a custom wc, nothing super special I guess, good block big rad and done

no, I think he means IF they add a chiller it may do 4.9-5.0.

 

They use a custom wc, nothing super special I guess, good block big rad and done

Well then 4.8 on a $500 custom loop...that's still short of what the 8700K is getting. If you drop the extra $1200 on the chiller then maybe get it up to what the 8700K is doing. Again I'm only saying this in the DCS world where the only thing that matters is a single core speed correct? I'm certainly not an expert on the subject but I also like to have the fastest available speeds. Now as said before Vulkan will likely change all that but how far out is that? I would guess it's still a couple of years away.

I'm also not trying to bash Ryzen. The faster they go then the faster Intel goes. Competition can only help us all!

Skatezilla's results are definitely impressive but still not a practical application for all but a very rare few(an extra $1700 of worth of hardware to cool it). I would still think DCS will run faster on Intel. Mining maybe not but that is a small relatively unimportant segment of the market. As I'm sure you know Bitmaster every 8700K will hit 5Ghz pretty easily.

I need to redo the title ! Ryzen+ may hit 5G on good dies..how bout that ?

There you go !!!!!!!!!!!!!!!!!!!!!!!!!

 

Goodbye Intel

SkateZilla, just because I'm curious, what core voltage was used ? (I bet pretty higher than 1.4...?)

Our studio overclocked a 2700x to 4.8 Ghz on Custom Loop liquid cooling,

 

We can prolly get 4.9/5.0 if we deep cool the liquid in the loop via cooler unit.

 

So Price of Custom Loop ($500~), and Cost of a Chiller to cool water to pre-specified temp, $1200

I should Prolly elaborate some.

4.8 Ghz on a Custom Liquid Cooling w/ 16 Threads Full Tilt Stress Test for over 6 hours and Temps Never exceeded 60°C, Temperature on the CPU Cores nor the Package was not the problem.

The Main Issue we encountered when going above 4.8 GHZ was a result of Voltage at 1500mv+,

The VRM Array couldn't handle it properly, w/ temp Spikes and vdroops as a result, could be a mainboard model issue. Chilling the Liquid would better control said spikes, but thus far, Friends in the OC Community have not seen such issues on their Mainboards.

In a Controlled temp Environment of 24°C, a Decent Water Cooling Loop w/ a 320mm Radiator properly mounted or even externally mounted to keep liquid at Ambient temp +~5°c would keep it cool at 5 GHz.

I have friends that are at 5.1 GHz on All cores on Custom Water Loops that range from $249-$600),

5.8 GHz on the other hand, required a significant voltage bump over 5.0 GHz, and thus the Exotic Cooling of both the CPU and VRM Array.

I should Prolly elaborate some.

 

4.8 Ghz on a Custom Liquid Cooling w/ 16 Threads Full Tilt Stress Test for over 6 hours and Temps Never exceeded 60°C, Temperature on the CPU Cores nor the Package was not the problem.

 

The Main Issue we encountered when going above 4.8 GHZ was a result of Voltage at 1500mv+,

 

The VRM Array couldn't handle it properly, w/ temp Spikes and vdroops as a result, could be a mainboard model issue. Chilling the Liquid would better control said spikes, but thus far, Friends in the OC Community have not seen such issues on their Mainboards.

 

In a Controlled temp Environment of 24°C, a Decent Water Cooling Loop w/ a 320mm Radiator properly mounted or even externally mounted to keep liquid at Ambient temp +~5°c would keep it cool at 5 GHz.

 

I have friends that are at 5.1 GHz on All cores on Custom Water Loops that range from $249-$600),

 

5.8 GHz on the other hand, required a significant voltage bump over 5.0 GHz, and thus the Exotic Cooling of both the CPU and VRM Array.

Thank you for the clarifications. So with all that said and using a normal water cooling setup in DCS terms(lol that's all I use my rig for) Are we talking superior performance over a 8700K? Relatively the same?

That's 2 cores more at said speed, wow..I mean I have not expected to read about a Ryzen+ that hits 5.1 on 8c/16t full tilt on a good DIY Loop in reasonable price range. That is awesome if it turns out to be repeatable by many.

Skatezilla's results are definitely impressive but still not a practical application for all but a very rare few(an extra $1700 of worth of hardware to cool it). I would still think DCS will run faster on Intel. Mining maybe not but that is a small relatively unimportant segment of the market. As I'm sure you know Bitmaster every 8700K will hit 5Ghz pretty easily.

According to Silicon Lottery (which sells guaranteed OC's)...

As of 3/22/18, the top 88% of tested 8700Ks (delidded which drops temps 5-20*F) were able to hit 5.0GHz or greater.

There are a lot of unhappy 8700K owners that would dispute your "5Ghz pretty easily" assertion. If you Google around, you will find that there is strong evidence to suggest popular review sites were given cherry picked 8700K's at launch. The only way to guarantee 5Ghz and up on all cores is with a pre-tested chip.

Speaking for myself, I have always had bad karma when it comes to CPUs. The last time I got what people claim is the "easy overclock" it was on a Slot 1 Celeron 300A that ran at 450 many moons ago.

The 8700K is a fantastic chip, and beyond a doubt the gaming king at the moment, but it isn't the sure thing people are lead to believe.

Thank you for the clarifications. So with all that said and using a normal water cooling setup in DCS terms(lol that's all I use my rig for) Are we talking superior performance over a 8700K? Relatively the same?

the AMD chips probably not going to match a 8700K in single thread still, however:

DCS used to be pegged at the CPU in single thread but this is no longer the case as the games objects have been offloaded to the GPU.

Between the R5 2600 and the Intel 8700K I would go for the former if that allows me to to buy a bigger GPU with the leftover cash. The AMD barebones of MOBO+CPU is MUCH cheaper.

Willing to bet R5+1080Ti > 8700K+1080.

If you can afford both the Intel 8700K and the 1080Ti thats a no brainer. Go ahead.

According to Silicon Lottery (which sells guaranteed OC's)...

As of 3/22/18, the top 88% of tested 8700Ks (delidded which drops temps 5-20*F) were able to hit 5.0GHz or greater.

 

There are a lot of unhappy 8700K owners that would dispute your "5Ghz pretty easily" assertion. If you Google around, you will find that there is strong evidence to suggest popular review sites were given cherry picked 8700K's at launch. The only way to guarantee 5Ghz and up on all cores is with a pre-tested chip.

 

Speaking for myself, I have always had bad karma when it comes to CPUs. The last time I got what people claim is the "easy overclock" it was on a Slot 1 Celeron 300A that ran at 450 many moons ago.

 

The 8700K is a fantastic chip, and beyond a doubt the gaming king at the moment, but it isn't the sure thing people are lead to believe.

Well maybe not every...but the vast majority will with really nothing special to cool them. I'm sure there are exceptions and those will probably be the loudest people reporting things as is normally the case.

the AMD chips probably not going to match a 8700K in single thread still, however:

 

DCS used to be pegged at the CPU in single thread but this is no longer the case as the games objects have been offloaded to the GPU.

 

 

Between the R5 2600 and the Intel 8700K I would go for the former if that allows me to to buy a bigger GPU with the leftover cash. The AMD barebones of MOBO+CPU is MUCH cheaper.

 

Willing to bet R5+1080Ti > 8700K+1080.

 

If you can afford both the Intel 8700K and the 1080Ti thats a no brainer. Go ahead.

That might be the case. DCS has always been a GPU arms race. Another reason I'll likely be going to Volta or Turing...not because my current setup needs it but because I'm thinking Pimax 8K-X will.

ooph, 8K... LOL :eek:

Maybe when multi-GPU (not SLI) in vulkan and DX12 are a thing and adopted by the industry. I dont see any single card handle that at an acceptable performance level. The 11 series Geforces are not going to double in performance (so to speak because the increase in pixels is quadratic). They will be more akin of a refresh with a modest improvement. AMD GPU's have not kicked NVIDIA's pants like their CPU's division did to Intel, so NVIDIA is milking the most out of us for the smallest investment possible.

8k VR ?

6 fps ??

The Pimax 8K-X will be 200 Deg FOV and native 4K per eye. They've said it will run smooth on the next gen of cards. The Pimax 8K which will be 200 Deg FOV and 1440p upscaled to 4K per eye will run fine on a GTX 1070 and 1080. Already been tested in DCS. So not really 8K that's just the name...

Can we see a screenshot Skate ??

People are not believing those numbers w/o a proof, which I do understand.

Thank you for the clarifications. So with all that said and using a normal water cooling setup in DCS terms(lol that's all I use my rig for) Are we talking superior performance over a 8700K? Relatively the same?

I cant run DCS on my Studio Systems. We merely upgraded to Ryzen+ for the Increased IPC and Memory Support.

That being said, my Home Rig still runs a FX8350 Overclocked to Oblivion.

Can we see a screenshot Skate ??

 

People are not believing those numbers w/o a proof, which I do understand.

There will be plenty of Articles in a week or so from Overclocking friends.

5.0 Ghz w/ Custom Liquid Cooling (With Emphasis on Radiator, Fluid Type and Ambient Temp) can do 5 GHZ on a Quality Board w/ a Quality VRM Array.

AS with any Chip Design, There's points in the Overclocks that vary with each chip.

Max Overclock on Default Voltage,

Max Overclock on Air Cooling,

Max Overclock on Liquid Cooling,

Max Overclock on Exotic Cooling,

Max Overclock without Diminishing Returns.

on Ryzen 2700X, Most people can Overclock from 3.7Ghz Base to 4Ghz-4.2 GHz Easily, and some might be able to do it without increasing the voltage, while others will have to bump the Default up 15-20%, and up to 1400mV

However after 4.3, Careful Tuning to both the CPU Voltage and the Memory timings is needed to maintain stability.

And a Decently Configured Liquid Cooling Loop in a Controlled Ambient Temp Environment, as when you get to 4.5 Ghz and Above the TDP Rises to well above 150w.

4.8Ghz seems to be what I'm seeing from friends to be the Point of Diminishing Return, after 4.8, the Bump in Voltage to maintain stable OC is not worth the amount of Frequency Increase you get w/ each bump of 100mv.

I cant run DCS on my Studio Systems. We merely upgraded to Ryzen+ for the Increased IPC and Memory Support.

 

 

That being said, my Home Rig still runs a FX8350 Overclocked to Oblivion.

Well good to know. Either way it looks like the new Ryzen is a big step forward and can only help us all by pushing more tech our way.

Skate, you deserve a better DCS rig !

Skate, you deserve a better DCS rig !

I will upgrade my home system when Ryzen II and the Next AMD GPU Architecture hits in a year or so,

GPU And Memory Prices needed to DROP, and I Mean Drop A lot.

I hear ya ! I couldnt as well if I had to buy again, RAM + GPU just make it obsolete, wife would kill me, really .

the Ram package I want is $1199, So....

128er way to go ;)

I do have to wonder if anyone has successfully built a liquid nitrogen loop for 24/7 use and what kind of stable overclock they were able to achieve.

 

In any case, that 4.8 GHz I mentioned previously seems to be unconfirmed and more rumor than fact. I expect 4.3 GHz will be the brick wall we hit with most Ryzen 2xxx chips. That's a decent bump by itself. I think the more interesting part is in stock settings though. My Ryzen 5 1600x system does 4.1 GHz on 1 core when completely idle, but the instant I have so much as a browser window open on a blank page, it drops to 3.7 GHz. The reason is it has only 2 boost states, 4.1 single core and 3.7 all core. There is nothing in between.

 

The Ryzen 2xxx's use a new system which, from AMD charts, suggests it will scale more like the current Intel chips. So on a 2600x, it will hold that top boost speed with 2-3 cores active instead of only 1, and it will drop only slightly as core usage goes up. This isn't important to those of us who overclock the snot out of everything, but for the other 99% of the world, that's a big change.

 

Regardless, the 8700k and 8600k look to remain the top DCS chips for the short term. When DCS moves from DX11 to Vulkan, that may change but it will depend on how ED codes it.

It doesn't have to be LN2. You could do a 24/7 phase change system. I had one years ago and it was awesome. Not quite as cold as LN2, but its pretty damn cold, none the less.

Regardless of how you would cool it, anything above 1.50v is for sure a mid-term CPU Killer.

I am very sure such a stressed CPU wont last 3 years, the VRM's even less.

Regardless of how you would cool it, anything above 1.50v is for sure a mid-term CPU Killer.

 

I am very sure such a stressed CPU wont last 3 years, the VRM's even less.

well if you're not gaming or anything, you can program the AMD Tool to Drop Voltage and re-enable core parking.

That way it's not a 24/7 overclock and you're not burning up the pathways in the chip with 1500mv 24/7, but only when you really need it.

Since DCS isnt multi-Threaded past 1 Thread for Simulation 1 for Sound, then you could prolly get away w/ Overclocking and turning off SMT.

or boosting XFR2 to 5+ Ghz on a Single core and lock DCS to that core.