Speculation on Ryzen Overclocking

[antihelten]

Hey. You said they were wrong but FX-8150 is actually performing as 1,5*i7 950 in a lot of case as a whole package.

Why don't you show us some examples then.

Also note that Donaminhaber didn't just claim that the 8150 could outperform the 950 by 50% in individual apps, they claimed that it outperformed it by 50% overall:

"where the estimated results in the document processor are three different categories (media, rendering and game) under comparing the overall assessment of the 8-core Bulldozer processor Core i7 950 is 50% faster than emphasized."

By the way this is not their claim. They say that "Internal AMD tests claims this".

Which was just another lie, Donaminhaber claimed to have gotten the numbers from and internal AMD document, but JF-AMD later confirmed that no such document existed, and it didn't come from them. In other words, they made the whole thing up.

Naah... I do not think so. Just knowing these guys. They are not just random guys. If they did not tell anything I do not think WCCF will copy my translation and publish as news

WCCFTech will take literally anything and post it, with no source criticism whatsoever, just like Donanimhaber coincidently

 

[mtcn77]

Moon, I'm just the messenger, posting this where it should be, the overclocking rumors thread.

Go tell that (and I will, too), to the guys who made that video, which by the way don't seem to be randoms




I asked the guy for a source and he posted the video with the translations. It seems to be legit.

Also,



memory overclocking shouldn't be a problem either

I can verify him being a translator, lol.
One thing extra: there were a lot of mentions of Russians watching the stream... It was as if the editors were riled for no reason. They are very articulate editors, you won't see them mincing words often. Traditionally AMD has been getting a lot of flak from them for procuring weak cores. It is very out of character for them to endorse AMD after such a while. They even missed out on all deadlines to report on recent AMD events they signed up, lol. Members were teasing them for it.

 

[mtcn77]

Why don't you show us some examples then.

Also note that Donaminhaber didn't just claim that the 8150 could outperform the 950 by 50% in individual apps, they claimed that it outperformed it by 50% overall:

"where the estimated results in the document processor are three different categories (media, rendering and game) under comparing the overall assessment of the 8-core Bulldozer processor Core i7 950 is 50% faster than emphasized."



Which was just another lie, Donaminhaber claimed to have gotten the numbers from and internal AMD document, but JF-AMD later confirmed that no such document existed, and it didn't come from them. In other words, they made the whole thing up.



WCCFTech will take literally anything and post it, with no source criticism whatsoever, just like Donanimhaber coincidently

Wccf has been on my blacklist, but they have taken a turn for the 'editorials' lately. They correctly predicted, upon Intel quarterly reports, that AMD would go head on against the Intel-x99 line up. wccftech.com/intel-not-worried-ryzen-confident-kaby-lake/

 

[antihelten]

Wccf has been on my blacklist, but they have taken a turn for the 'editorials' lately. They correctly predicted, upon Intel quarterly reports, that AMD would go head on against the Intel-x99 line up. wccftech.com/intel-not-worried-ryzen-confident-kaby-lake/

Thing is WCCFTech does get things right from time to time, but they also get things wrong so often that it's pretty much a coin toss whether a given story is true or not, and that arguably makes them rather useless as a news source (or an editorial source, except as a primer for discussion).

 

[mtcn77]

Thing is WCCFTech does get things right from time to time, but they also get things wrong so often that it's pretty much a coin toss, whether or not a given story is true or not, and that arguably makes them rather useless as a news source (or an editorial source, except as a primer for discussion).

I can tell you that DH's chief editor does report on Wccf sourced articles. Sometimes you get to suck up your ego and do the work. Unfortunately that didn't get the go-ahead. Something to do with company operatives, imo.

 

[JDG1980]

Why don't you show us some examples then.

Bulldozer (FX-8150) really did beat mainstream Nehalem (i5-750) by 50% or more on various multi-threaded integer benchmarks.
http://www.anandtech.com/bench/product/109?vs=434
Nearly 50% better on Cinebench R10 MT, nearly double on 7-Zip, 66% better on POV-Ray, 67% better on x264 2nd pass.
Of course, by the time it was released, it had to go up against Sandy Bridge, where it didn't fare nearly as well.

Which was just another lie, Donaminhaber claimed to have gotten the numbers from and internal AMD document, but JF-AMD later confirmed that no such document existed, and it didn't come from them. In other words, they made the whole thing up.

JF-AMD was himself lying his ass off the whole way through the Bulldozer pre-launch. I don't know anything about Donaminhaber's credibility, but it can't possibly be lower than that of John Fruehe.

 

[antihelten]

Bulldozer (FX-8150) really did beat mainstream Nehalem (i5-750) by 50% or more on various multi-threaded integer benchmarks.
http://www.anandtech.com/bench/product/109?vs=434
Nearly 50% better on Cinebench R10 MT, nearly double on 7-Zip, 66% better on POV-Ray, 67% better on x264 2nd pass.
Of course, by the time it was released, it had to go up against Sandy Bridge, where it didn't fare nearly as well.

Since when is "various multi-threaded integer benchmarks" the same as the overall performance across "media, rendering and game", and since when is an i5 750 the same as an i7 950?

Cinebench R10 MT: 11% faster than i7 950
POV-Ray: 13% faster than i7 950
x264 2nd pass: 24% faster than i7 950

No where near 50% faster (unfortunately Anandtech didn't have any 7-zip numbers for the i7 950, but the 8150 would probably be 10-20% faster)

JF-AMD was himself lying his ass off the whole way through the Bulldozer pre-launch. I don't know anything about Donaminhaber's credibility, but it can't possibly be lower than that of John Fruehe.

Yes, JF-AMD was doing everything he could to make Bulldozer look like less of disaster than it was, so clearly it is not in his interest to shoot down someone who claims that Bulldozer will be great, and yet he did just that. So either the leak was legit and JF-AMD did something that doesn't make any sense whatsoever, or the leak was fake, your call.

And again Donaminhaber was also the site who happily passed of fake benches from OBR, so even without the 50% claim, their credibility is still in the gutter (or as I said earlier WCCFTech tier).

Seriously, just go back and read some of the bulldozer discussion from back in the day either here or over at hardforum (where JF-AMD was also quite active), and you will quickly see just how much respect people had for Donaminhaber back in those days. Now it's certainly possible that they have cleaned up their act since then, but I wouldn't bet on it.

 

[bjt2]

Nah, i was worrying for a second i would lose the overclocking bet, nothing more.

Die looks exact same either way (well, color scheme is different). If changes were there, they were on microcode level.

Long ago i heard that the energy saving on carrizo was partly disabled due to malfunctioning. Probabily you are right...

 

[PhonakV30]

hmm so ryzen can go up 4.4ghz on all cores or just one core.I'm not so sure.

guide from MSI's X370 Titanium.

Edit : like MSI Z270 with Intel 7700K

http://www.guru3d.com/news-story/ms...ame-boost-knob-ocing-with-liquid-cooling.html

 

[antihelten]

hmm so ryzen can go up 4.4ghz on all cores or just one core.I'm not so sure.

guide from MSI's X370 Titanium.

Edit : like MSI Z270 with Intel 7700K

http://www.guru3d.com/news-story/ms...ame-boost-knob-ocing-with-liquid-cooling.html

post #239

 

[IEC]

Not sure if anyone posted it, but here is the manual of MSI X370 motherboard
Second page is of the most interest (and nice Spinal Tap reference):

So it seems that at least MSI is confident that 1800X can indeed clock to 4.4Ghz

Dialing it to 11, eh?

What a coincidence that it was released on March 2nd, 1984. Hmm...

 

[inf64]

Has this been posted yet?
https://www.scan.co.uk/shop/compute...-professionally-overclocked-amd-ryzen-bundles

We now know what is attainable with good air and AIO WC setups. Scan had to test their bundles before offering them up for sale. Note there are some errors in listings, some are labeled as 1800X while specs say it is 1700(no X, or X) OCed to either 3.8Ghz, 4Ghz or 4.2Ghz. They likely put a wrong spec sheet on some of them, read the description for details. All bundles are using 3000Mhz DDR4. 4.2Ghz OCed ones use Crosshair Hero VI Asus boards while 3.8Ghz (1700) use ASUS Prime B350-Plus. There are two cooling options, be quiet air cooler or H100i WC.

Ryzen 1700(3Ghz base,65W stock): 3.8Ghz on good air cooling (be quiet! Dark Rock) or Corsair Hydro H100i

This Overclocked bundle features the latest AMD Ryzen R7 1700 Processor. The R7 1700 is a 8 Core 16 thread CPU which we overclock from 3GHz to 3.8GHz for great performance in all games and applications. It is kept cool by a Corsair Hydro H100i water cooler and installed in an ASUS Prime B350 motherboard and accompanied by 16GB of 3000MHz Corsair Vengeance LPX DDR4.

Ryzen 1700X (3.4Ghz base, 95W) : 4Ghz on Corsair Hydro H100i

This Overclocked bundle features the latest AMD Ryzen R7 1700X Processor. The R7 1700X is a 8 Core 16 thread CPU which we overclock from 3.4GHz to 4GHz for great performance in all games and applications. It is kept cool by a Corsair Hydro H100i cooler and installed in an ASUS X370-Pro motherboard and accompanied by 16GB of 3000MHz Corsair Vengeance LPX DDR4.

Ryzen 1800X (3.6Ghz base, 95W): 4.2Ghz on Hydro H100i

This Overclocked bundle features the latest AMD Ryzen R7 1800X Processor. The R7 1800X is a 8 Core 16 thread CPU which we overclock from 3.6GHz to 4.2GHz for great performance in all games and applications. It is kept cool by a Corsair Hydro H100i cooler and installed in an ASUS X390-Pro motherboard and accompanied by 16GB of 3000MHz Corsair Vengeance LPX DDR4.

So as expected 1700 has lower OC potential than 1700X, 1700X in turn a bit lower than 1800X. 1800X doing full stability on all 8 cores @ 4.2Ghz matches earlier leaks of ~4.2-4.3Ghz OCing headroom on release SKUs. This matches (roughly) what BDW-E can do, maybe a smidge lower. Given compterbase showed us 6-8cores are must haves for 2017 gaming, Ryzen is going to be a killer CPU.

 

[lightmanek]

If Scan.co.uk is selling pre-overclocked 4.2GHz bundles, this means that most 1800X should do a bit more than that. I wouldn't be surprised if 4.4GHz will be achieved on quite a few samples under good water.
Good times ahead! I will be doubling my current rig performance in less than 3 weeks time! (i7 4790k)

 

[lolfail9001]

I wouldn't be surprised if 4.4GHz will be achieved on quite a few samples under good water.

Unless of course 4.2ghz is thermals limited, h100 is close enough to decent water, last time i checked.

 

[Ryzenenthusiast]

Unless of course 4.2ghz is thermals limited, h100 is close enough to decent water, last time i checked.

Preoverclocked builds never tries to push the boundries. So 4,4 under good AIO and 4,6 under best custom water loops 7/24 OC won't be a big surprise.

 

[inf64]

Preoverclocked builds never tries to push the boundries. So 4,4 under good AIO and 4,6 under best custom water loops 7/24 OC won't be a big surprise.

I also think these pre-OCed builds are more on the "safe" side.

 

[lightmanek]

Unless of course 4.2ghz is thermals limited, h100 is close enough to decent water, last time i checked.

True, but looking at other 3XS overclocked bundles, they offer 6900X and 6950X with 4.2GHz OC same as Ryzen and looking at 7700K they best bundle is 4.8GHz OC.
Shop cannot afford to sell OC bundles on the limit as cost of after-sale support and warranties would outweigh any extra profit they make. What they sell is supposed to just work. There is untapped potential, hopefully similar to what 6900X can do in a system tuned for max. clocks, or even better.

 

[majord]

Has this been posted yet?
https://www.scan.co.uk/shop/compute...-professionally-overclocked-amd-ryzen-bundles

We now know what is attainable with good air and AIO WC setups. Scan had to test their bundles before offering them up for sale. Note there are some errors in listings, some are labeled as 1800X while specs say it is 1700(no X, or X) OCed to either 3.8Ghz, 4Ghz or 4.2Ghz. They likely put a wrong spec sheet on some of them, read the description for details. All bundles are using 3000Mhz DDR4. 4.2Ghz OCed ones use Crosshair Hero VI Asus boards while 3.8Ghz (1700) use ASUS Prime B350-Plus. There are two cooling options, be quiet air cooler or H100i WC.

Ryzen 1700(3Ghz base,65W stock): 3.8Ghz on good air cooling (be quiet! Dark Rock) or Corsair Hydro H100i


Ryzen 1700X (3.4Ghz base, 95W) : 4Ghz on Corsair Hydro H100i


Ryzen 1800X (3.6Ghz base, 95W): 4.2Ghz on Hydro H100i


So as expected 1700 has lower OC potential than 1700X, 1700X in turn a bit lower than 1800X. 1800X doing full stability on all 8 cores @ 4.2Ghz matches earlier leaks of ~4.2-4.3Ghz OCing headroom on release SKUs. This matches (roughly) what BDW-E can do, maybe a smidge lower. Given compterbase showed us 6-8cores are must haves for 2017 gaming, Ryzen is going to be a killer CPU.

Well The fact they can get a 1700 to 3.8Ghz on a board with such a weedy 4+3 phase VRM is a good sign there's more in the 1700. It also achieves the same 3.8Ghz on the air cooled bundle.

All that aside.. interesting point of comparison is the BW-E intel systems.. also 4.2Ghz.

This should really mean 4.3-4.4 should be doable in a DIY environment , just like BW-E

 

[lolfail9001]

Preoverclocked builds never tries to push the boundries. So 4,4 under good AIO and 4,6 under best custom water loops 7/24 OC won't be a big surprise.

We do not know how tough the voltage wall is still. 4.2Ghz is not pushing it on Broadwell-E but 4.3Ghz already requires voltages that cause degradation.

True, but looking at other 3XS overclocked bundles, they offer 6900X and 6950X with 4.2GHz OC same as Ryzen and looking at 7700K they best bundle is 4.8GHz OC.

Yes, these are the safe OCs. What it really means is that 4.2Ghz with good AIO is the safe bet on 1800X. That's about as well as i have expected, tbh.

There is untapped potential, hopefully similar to what 6900X can do in a system tuned for max. clocks, or even better.

So... 4.2Ghz ?

 

[lightmanek]

....

So... 4.2Ghz ?

I was thinking more in the lines of 4.25GHz 0.25x multi will come handy!

 

[A///]

The dial boards were available with Intels too. It's like an auto OC for you, a simplistic approach. It also applies stupid amounts of voltage with lack of ROI with certain badly binned chips. OC the old fashioned way.

 

[guskline]

Ryzenenthusiast: If I get a rock solid 1800x at 4.2 on all cores it should be faster than my 5960x at 4.4 Ghz. I did a ton of OCing at first with my 5960x and I could get it to 4.5 solid but the temps were a bit high for my liking even under custom water. If I had one of the professional tuners have a go at it I'm sure they could eek out more. However, I have it at 4.4 and no matter what benchmark I throw at it the machine is rock solid AND fast. Heck I remember upping the voltage etc and I could boot into windows at 4.8 BUT.

BTW I always run a long ASUS RealBench test, at least 6 hours. It does a nice job of stressing the overall setup. I also use my licensed Aida64 stress test and run at least as long.

I am going to build a 1800x/Asus CF6/Gskill DDR4-3200(2x8) combo under an EK Supremacy EVO cpu block (plenty of rad cooling 2-360 slims) and let it run stock and then try to get it stable at 4.1-4.2. Should be fun to compare to my 5960x.

 

[IEC]

Ryzenenthusiast: If I get a rock solid 1800x at 4.2 on all cores it should be faster than my 5960x at 4.4 Ghz. I did a ton of OCing at first with my 5960x and I could get it to 4.5 solid but the temps were a bit high for my liking even under custom water. If I had one of the professional tuners have a go at it I'm sure they could eek out more. However, I have it at 4.4 and no matter what benchmark I throw at it the machine is rock solid AND fast. Heck I remember upping the voltage etc and I could boot into windows at 4.8 BUT.

BTW I always run a long ASUS RealBench test, at least 6 hours. It does a nice job of stressing the overall setup. I also use my licensed Aida64 stress test and run at least as long.

I am going to build a 1800x/Asus CF6/Gskill DDR4-3200(2x8) combo under an EK Supremacy EVO cpu block (plenty of rad cooling 2-360 slims) and let it run stock and then try to get it stable at 4.1-4.2. Should be fun to compare to my 5960x.

I'll be using an almost identical setup, so best of luck clocking to >4.2GHz for us both!
1800X
Asus CF6H
G.Skill DDR4-3400 (2x8)
EK Supremacy EVO MX
420mm + 360mm rads
Dual pump system

 

[guskline]

IEC, what case will you be using? I was originally going with a Fractal Define S but it was too tight with both Magicool rads even though they were slim. I was afraid of there not being enough room for the Asus CH6 so I opted for a Thermaltake V51. I thought it had enough room in the roof for the EK CE420 but again too tight for my tasts so I'm going with dual Magicool G2 slim 360 rads with 6 XSPC 1600 rpm 120mm fans controlled by a fan controller. Since I'm only cooling the cpu it is really overkill but put in both rads if I eventually WC a single Vega.

I'm going to pair the WC'd 1800x with 2 air cooled reference Sapphire RX480s-8G in CF. Should be fun!

 

[Majcric]

Ryzenenthusiast: If I get a rock solid 1800x at 4.2 on all cores it should be faster than my 5960x at 4.4 Ghz. I did a ton of OCing at first with my 5960x and I could get it to 4.5 solid but the temps were a bit high for my liking even under custom water. If I had one of the professional tuners have a go at it I'm sure they could eek out more. However, I have it at 4.4 and no matter what benchmark I throw at it the machine is rock solid AND fast. Heck I remember upping the voltage etc and I could boot into windows at 4.8 BUT.

BTW I always run a long ASUS RealBench test, at least 6 hours. It does a nice job of stressing the overall setup. I also use my licensed Aida64 stress test and run at least as long.

I am going to build a 1800x/Asus CF6/Gskill DDR4-3200(2x8) combo under an EK Supremacy EVO cpu block (plenty of rad cooling 2-360 slims) and let it run stock and then try to get it stable at 4.1-4.2. Should be fun to compare to my 5960x.

I have to ask how does your 5960x do under prime 95? If Ryzen can do 4.4/4.5 on air I may pick one up.