I agree, it will be just like x58, you will really need a beefy MB to handle the high OC's, just like the x58 boards alot of them could just not handle 200Mhz FSB you had to be real careful when selecting a board.
the problem with Zen OC looks to be VRM cooling.
I can reach a fairly nice overclock with one click. You don't need to bother with manual OC
Great info, thxOk so we got a bit of clarification on the Turkish OC info. These guys are totally legit.. they are from donanimhaber.com one of the oldest review sites in Turkey. Those who followed the Bulldozer launch will remember they had info which ended up being true back then as well.
Original translation: http://i.imgur.com/MwqFkRX.png
Some clarification:
On "auto overclock" feature:
This could be just like sliding the power limit to the max, and XFR could take over or similar.
He also goes on to say that Single Threaded performance it's just slightly behind the 7700k even when OCed but not in all tests, in one particular test it's faster than 7700k in single threaded benchmarks.
He also says with just the one click "auto overclock", the CPU has simply no competition and that not even 10 core 6950x can keep up with the Ryzen when OC'ed.
He explicitly states that "You are OC'ing your Ryzen 1800x Via Auto Overclock, and then there is no Intel chips that can keep up with it."
Summerized from reddit, and a few Turkish speakers there: https://www.reddit.com/r/Amd/comments/5w0at9/ryzen_oc_potential_read_this_translation_to_fuel/
They are using Crosshair Hero, don't they? Because ASUS boards do have that stuff.This could be just like sliding the power limit to the max, and XFR could take over or similar.
Strong statement, Cotton, let's see if it works out in OC vs OC tests.He explicitly states that "You are OC'ing your Ryzen 1800x Via Auto Overclock, and then there is no Intel chips that can keep up with it."
The Stilt has already commented on this. RyZEN does not need a beefy board like X58 thanks to its power-thrifty nature. (X58 was a furnace) A board will stand out or fall into obscurity depending on how well its BIOS is programmed to ensure compatibility with memory and extract performance out of the CPU.I agree, it will be just like x58, you will really need a beefy MB to handle the high OC's, just like the x58 boards alot of them could just not handle 200Mhz FSB you had to be real careful when selecting a board.
He is saying good things about auto overclock 2 times. In first he says "It can reach to great frequencies via Auto Overclock with that motherboard" In the second he says "I do not think I would bother to manually OC it." So I had completed the puzzle while he is looking at a single piece. I am a simultaneous interpreter. So translation is my profession.
In 7700k case; He says that "In a specific test it outperforms stock 7700k in single threaded benchmark. But in general single threaded workloadwise it is just a little bit behind 7700k" He just does not mention whether the Ryzen is OC'ed or not. But seems like it is definitely OC'ed. I do not think they'll beat Skylake IPC by +10%. It is impossible.
For the 6950x the exact thing they said; "You are OC'ing your Ryzen 1800x Via Auto Overclock, and then there is no Intel chips that can keep up with it."
He explicitly states that "You are OC'ing your Ryzen 1800x Via Auto Overclock, and then there is no Intel chips that can keep up with it."
Now that I think about it... That approach works pretty well in practice. When all cores are used the clock speed is not that important. If it really can boost single or two cores much higher then the results could be rather interesting.Yup. That is what I think might happen for 1700/1700X/1800X.
Single-threaded performance = Equal or better than competing Intel solutions because of XFR
Multi-threaded performance = Equal or better than competing Intel solutions because of better SMT and/or more cores
This way, even if single-threaded IPC is slightly lower, it will outperform 7700K/6850K/6900K in majority of tasks. Overclocked ST performance crown will still belong to 7700K.
I think you have just nailed it, if the tech press treat XFR like they do turbo boost 3 on the intel chips then at least 1800x would be capable of practically matching 7700k ST, whilsh destroying practically every intel chip im most MT scenarios.Yup. That is what I think might happen for 1700/1700X/1800X.
Single-threaded performance = Equal or better than competing Intel solutions because of XFR
Multi-threaded performance = Equal or better than competing Intel solutions because of better SMT and/or more cores
This way, even if single-threaded IPC is slightly lower, it will outperform 7700K/6850K/6900K in majority of tasks. Overclocked ST performance crown will still belong to 7700K.
This is not for sure... We don't know what is the Ryzen clock limit. 5GHz on air with one core on an early ES on developement board with frying VRMs is encouraging: take a MB with beefed up VRMs, a good air cooler, put all the XFR knobs to maximum and let the clock skyrocket... You can also reach over 4.5GHz...Overclocked ST performance crown will still belong to 7700K.
I think you have just nailed it, if the tech press treat XFR like they do turbo boost 3 on the intel chips then at least 1800x would be capable of practically matching 7700k ST, whilsh destroying practically every intel chip im most MT scenarios.
What this means for the 1700 i dont know.
Yes, it does look like the launch bios is broken, it seems to be very complex (xfr?) it will be months until we see best zen.It's even better than TC 3.0: it does not need an OS driver and it scales with the cooler and VRM quality... AFAIK TC 3.0 has fixed turbo clock...
You still have faith in XFR, don't you.put all the XFR knobs to maximum
XFR is not really a technical thing, as CPC and looncraz pointed out.it seems to be very complex (xfr?)
Okay, calm down. It is my job to be all doom and gloom about everything.The memory latency is worrying, that might be more hardware rooted and we might have to wait until zen+
Prepare for some mild disappointment if those are your expectations.If the 1700 can reach 4.5Ghz on good air like D14 then we are talking. I am still sticking with X79 and 2670 due to ram prices right now, registered ecc ddr3 is DIRT cheap.
It's not faith. It's knowledge. I have seen technical papers on it, in the Bristol ridge version. My organization has access to all papers on ieeexplore and i have read all relevant papers on Bristol ridge. I am monitoring ieeexplorer for new papers on Ryzen, but for now the newest paper is of january 2017 and is the full description, up to the lowest level, of the technologies in bristol ridge. Shadow p-states is very similar to XFR. XFR should be better. Boot time calibration already works as i described on bristol ridge: test at boot to infer dc offset, vrm quality, motherboard quality, silicon quality and estimate tdp, current and voltage limits, for using during normal operation. Ryzen should have all those technologies, probabily improved.You still have faith in XFR, don't you.
Prepare for some mild disappointment if those are your expectations.
Rumors from gibbo that 1700 clocks to 4050mhz on the best crosshair mobo and on water, 3.8ghz every other mobo.
1700x/1800x rumored to clock higher, of course new bios may change this position.
Are you sure your papers are not on Pure Power/Precision boost, however? CPC did point out that there do not seem to be any technical docs on XFR, only marketing. And looncraz version of it is extremely compelling. Plus the recap highly suggests their value is that of "Pure Precision Power Boost^TM".I have seen technical papers on it, in the Bristol ridge version.
That's your speculation, isn't it? Besides, isn't Bristol Ridge die basically identical to Carrizo.Shadow p-states is very similar to XFR. XFR should be better