News Silicon Lottery Reveals AMD Ryzen 3000 Binning Stats

[Admin]

Silicon Lottery has updated its binning database with statistics for Ryzen 3000-series processors.

Silicon Lottery Reveals AMD Ryzen 3000 Binning Stats : Read more

 

[MrN1ce9uy]

What's keeping AMD from hitting higher clock speeds?

 

[redgarl]

OC is now a thing of the past. SL themselves admit it. On the contrary, the fact that those chips are so close in term of frequency from each others demonstrate standard in the manufacturing process.

https://www.extremetech.com/computing/295693-the-end-of-high-performance-overclocking-may-be-nigh

SL then writes:

AMD has done a fantastic job here overall, and we’re very aware this is the start to the end of our company in general. As both AMD and Intel optimize their binning process more and more, overclocking will not be possible as CPUs will boost themselves on their own to the highest clocks possible.

 

[Aspiring techie]

What's keeping AMD from hitting higher clock speeds?

Likely density and architecture.
Their IPC increase likely came at the expense of a little clockspeed.
If they had simply done a die shrink of Zen+, I guess they would have hit slightly higher (~200MHz) clocks.
But that's just my 2 cents.

 

[MrN1ce9uy]

I will give it to AMD for making very efficient chips. Intel's i9-9900K will heat up even the beefiest air coolers past comfortable temps even at stock clocks. So I will admit high clock speeds aren't everything.

 

[hannibal]

7nm... Intel has big difficulties to get high frequences at 10nm. So no wonder than amd is Also hitting the wall! It was surprice to even amd that They could increase the speed from zen+ so much!
More cores less gigaherts in the future!

 

[TCA_ChinChin]

I might be completely misinterpreting something or just didn't catch something from the article, but is there a reason why voltages were kept so low? Surely they could have gained another 100 or 200 MHz simply by raising voltages to 1.35-ish volts without much detriment to the health of the processor? If it is heat related, then does SL mention that? Or is it simply because that number is the highest clock they will reach regardless of the voltage used, so they simply lowered the voltage to the lowest they could while keeping that max clock? That and the voltages used to achieve each result for each chip is quite different?

Obviously SL isn't here to provide benchmark data for processors, they are just tuning these chips to their own standard, then going the extra mile to publish data on the results, but I'm just curious.

 

[AlistairAB]

I might be completely misinterpreting something or just didn't catch something from the article, but is there a reason why voltages were kept so low? Surely they could have gained another 100 or 200 MHz simply by raising voltages to 1.35-ish volts without much detriment to the health of the processor? If it is heat related, then does SL mention that? Or is it simply because that number is the highest clock they will reach regardless of the voltage used, so they simply lowered the voltage to the lowest they could while keeping that max clock? That and the voltages used to achieve each result for each chip is quite different?

Obviously SL isn't here to provide benchmark data for processors, they are just tuning these chips to their own standard, then going the extra mile to publish data on the results, but I'm just curious.

Because it is 12 cores. That makes a lot of heat. If you try to run 12 cores at 1.35V too much heat would be produced, it isn't really a problem with voltage safety. Silicon Lottery tests with Prime95 for hours, so you can imagine how much heat Prime95 makes. If you're just gaming you can push up your voltage more.

 

[MrN1ce9uy]

Or is it simply because that number is the highest clock they will reach regardless of the voltage used, so they simply lowered the voltage to the lowest they could while keeping that max clock?

That's what I assumed. But...

I know my board cranks my voltage to 1.45-1.5V and I consitently hit 4.275GHz on my Ryzen 7 3700X with everything on Auto. Temps are great too.

Check it out.

*Edit: I suppose they were using synthetic benchmarks. This is just a game using 40-50% of the CPU. Soo go figure.

View: https://www.youtube.com/watch?v=du97S72KT3A

 

[derekullo]

On the bright side the difference in performance for the lowest versus the highest frequency is only
4.7 % for the 3900X
2.3% for the 3800X
2.4% for the 3700X

At least when compared to the i7-9700K with a 7.8% difference between the best and worse chips with AVX2.

https://siliconlottery.com/pages/statistics

Some may look at this as a lack of overclocking room ... on the flipside it ironically means more guaranteed performance if you aren't buying from silicon lottery.

 

[tennis2]

Basically you can expect to eke out an extra 10% all-core over stock. Certainly not exciting for overclockers, but good news for the uninitiated.

 

[alextheblue]

Surprisingly, the Ryzen 7 3700X shows lower overclocking potential. Only 21% of Ryzen 7 3700X parts got to 4.15 GHz on 1.262V.

I suspect (due to binning) that a lower percentage of the 3700X models will reach the speeds of the 3800X. However, that's just a hypothesis, and your assertion is not grounded in hard data. The numbers presented only prove that at a lower voltage, you achieve lower overclocks.

 

[mamasan2000]

At the same time, this can be seen as a good thing for all the non-overclockers, the majority. Very little performance left on the table. Everyone gets great perf.
3000-series is nothing like 1000- or 2000-series. Those two generations benefited from overclocking.

I've been thinking, about RAM. What if someone made a Deep Learning app for RAM that calculated the best timings for a kit, any kit. Automated, tests every reasonable, safe setting. The test would take time but eventually you would be hardpressed to come up with better settings yourself. The DL test would have to have a bulletproof stability test incorporated.

 

[Finstar]

Damn, i finally won silicon lottery. My 3700X does 4.3GHz on 1.3v and 4.4GHz on 1.43v

 

[Blytz]

Anyone seen (or have they done) numbers on the 3600's ?

 

[Olle P]

Anyone seen (or have they done) numbers on the 3600's ?

They're not selling Ryzen 5, which makes sense since the price they'd charge would give you an untested Ryzen 7 that would most likely perform better at stock...

 

[SethNW]

AMD had to change what boost clock stands for, like to improve yids on new 7nm pricess, since that is fairly new with room to mature. So they changed what boost clock means, instead of all core boost it now stands for what best core can achieve, while other cores are maxing out on lower frequency, due to silicon quality. But they also worked with Microsoft to address that in task scheduler, which was optimized to send most sensitive tasks to the fastest core. While other less limiting threads go to slower cores. In a way it is bit like memory hierarchy where you have very small amount of fastest cache and biggest amount of slowest virtual memory, but due to how hierarchy works, it has big size and fast speed. I guess they had similar idea and it does allow them to sell CPUs for cheaper due to bigger yields.

But if you are looking for good and fun chip for overclocking, Intel always was it, since Ryzen first came out. And I feel like even Intel will have to become more aggressive in boost, so huge overclocking improvements we used to get will be less and less of a thing. Eve on Intel it used to be bigger gap between stock and over clicked, than it is today. So my guess is that it will just slowly fade outside doing it for fun or extreme overclocking.