News Silicon Lottery to Bin and Sell Ryzen 3000 CPUs

Math Geek

it's not just about clock speed to me. what would an extra 1-200 mhz get me day to day? (assuming i can oc it myself to only slightly slower speeds)

too much is put into pure speed. as we saw with the FX series, 5 ghz is not some magical thing that propels the chip into crazy performance. what price will they be asking for this extra bit of boost and what real world benefit will it have for me?? that to me is the question to be asking.

can't wait for reviews so we can find out the answer to that one :)


Oct 8, 2007
A selected Ryzen 3xxx chip is pretty interesting.

However, it seems like many folks are still thinking traditional OC. From what I've read, you let Ryzen 2xxx or 3xxx overclock / boost itself, which it will do about as well as a manual OC. The major benefit is that it will downclock when the speed isn't needed, saving a lot of power and heat. If you manually OC, it shuts off the boost circuits and the CPU will perform the same but run hot all the time.

I'm willing to be corrected but AFAIK a manual OC on Ryzen 2 or 3 is pointless, even silly, except for a benchmark run.

A selected chip will, presumably, just clock itself higher on the same cooling.
it's not just about clock speed to me. what would an extra 1-200 mhz get me day to day? (assuming i can oc it myself to only slightly slower speeds)
Yeah, it's kind of silly how far people will go for marginally better performance. When you get near 5Ghz, a 100Mhz increase works out to only a 2% difference, and real-world performance gains will be even lower at most tasks.

I'm willing to be corrected but AFAIK a manual OC on Ryzen 2 or 3 is pointless, even silly, except for a benchmark run.
That's not exactly how it works. There are still limitations to how high the processors can boost. On the X parts, multi-core boost clocks do tend to remain up near the overclocking limits of the processors though, so performance gains from manual overclocking will be minimal. The non-X parts leave a lot more room for overclocking though, since their stock multi-core boost clocks are lower. A 2700X can boost up around 4.1 to 4.2GHz on all cores provided it has adequate cooling, and a 2600X's boost clocks remain within about 100MHz of that, but a 2700 will drop to around 3.5GHz when most of its cores are loaded. The 2600 doesn't drop off as much, but it also doesn't boost quite as high as the 2700 does for lightly-threaded tasks.

This is likely in order to maintain the lower TDPs of these processors, as well as to differentiate the product stack. So overclocking can definitely help performance on the non-X processors, and bring their performance more in-line with their higher-priced counterparts. In the case of the 2700, it's possible to get nearly 20% higher clock rates over stock when running heavily threaded workloads.

As for the 3000-series, we'll have to wait for reviews to know for sure, but I suspect the situation will be similar. The multi-core boost clocks of the lower-priced variants at a given core count will likely drop off more significantly than their higher-priced counterparts. Being a new process node, it's possible that there could be more variance in overclocking headroom as well though.
@cryoburner You are contradicting yourself by saying 2% difference due to overclocking in one sentence and 20% higher clocks in multi threaded workloads.

I agree it's mostly a mute point to OC for single threaded clocks as the boost does such a nice job. However if you have the cooling and are looking for multi threaded performance than there are huge gains to be had from overclocking. If one is able to say OC the 3800x to 4.9Ghz we are talking a 20% performance gain for multithreaded workloads versus the all core clock fo 3.9Ghz.

Some people just like tweaking and playing with stuff more than the logical numbers prove out as well.

Personally I have OC'ed more than enough for fun if AMD would come out with a highly binned 3850x that was clocked much higher than the 3800x I would pick it up over trying to OC.
@cryoburner You are contradicting yourself by saying 2% difference due to overclocking in one sentence and 20% higher clocks in multi threaded workloads.
No, I said that at around 5 GHz, a 100 MHz (0.1 GHz) increase in clock rates can only equal a 2% increase in performance, at best. So, paying an extra $100 or so for a binned CPU guaranteed to get an extra 100 MHz is arguably a bit silly, since the performance difference should be more or less imperceptible.

Overclocking a Ryzen 2700 is quite a different story, since the CPU only hovers around 3.5 GHz at stock with most of its cores loaded. Compared to those lower clocks, each 100MHz increase works out to be closer to a 3% increase compared to stock, and the processor provides a decent amount of room for overclocking. So, if one manages to overclock that 2700 up to around 4.1 to 4.2GHz on all cores, you could be looking at a 600+MHz overclock, or around 17-20% higher multi-core clocks compared to stock, and it shouldn't cost all that much to get there. However, the difference between a 4.2 and 4.1 GHz overclock is again only about 2.4%, so it's questionable whether there would be much point in paying significantly more to get that extra 100 MHz.

My point is that overclocking is fine, and depending on the stock clocks of a given processor relative to its typical overclocking limits, one might be able to get a decent amount of additional performance out of it. However, you run into diminishing returns when paying a premium to improve your odds of getting a slightly higher overclock.

Looking at SiliconLottery's current offerings, an i7-9700K capable of overclocking to 5.2 GHz is priced at $590, or $470 for one that can hit 5.1 GHz, with shipping starting at $13. Meanwhile, according to their own data, 89% of 9700Ks are capable of overclocking to at least 5.0 GHz at 1.35 volts. So, going by that data, just by picking up a random 9700K off Amazon for $365 shipped, you will be almost guaranteed to get a 5.0 GHz overclock out of it, and it should have about a 36% chance of hitting 5.1 GHz, or a 10% chance of hitting 5.2GHz under their test conditions. That works out to paying nearly $120 to guarantee up to 2% more performance in CPU-limited scenarios compared to what is typical, or nearly $240 for up to 4% more performance. There's probably better components that one could spend that money on, unless their system is already more or less as high-end as it gets.

As for overclocking a 3800X to 4.9 GHz, for one thing, we still don't know whether they will even overclock that high, and for another, we don't know whether a 3700X will overclock to a similar level at a lower price. Also, we don't even know exactly what these processors all-core boost clocks will be until reviews come out. 3.9 GHz is the base clock of the 3800X, which you will likely only ever encounter if the processor is overheating under load. If it's anything like the 2700X, expect all-core boost clocks to be within a couple-hundred megahertz or so of the single-core boost. The 2700X has a base clock of 3.7 GHz and a single-core boost of 4.3 GHz, but it will happily boost all cores up near 4.1 GHz at stock, quite a bit higher than the base clock. That's why the benefits of overclocking a 2700X can be considered a bit questionable, while the benefits of overclocking a 2700 tend to be much more substantial, since you are starting with lower clocks to begin with, yet both processors can overclock to a similar level. It may or may not be a similar situation with the 3700X and the 3800X.

Also, that would be a "moot" point, not a mute point. : 3


Dec 5, 2016
So, the article gets some things wrong. Silicon Lottery did not always delid and use Liquid Metal TIM, then start binning. They have ALWAYS binned the chips for frequency at a specific voltage. They even post a methodology of which stress tests are run for an hour, usually the cooling used, ram, MB, etc.

What they are doing is binning for frequency with a specific voltage. When they go up 100MHz (or 50MHz as they did with Ryzen 1000 series and Threadripper 1000 series), they then only increase the voltage in a fixed amount, which then weeds out any chips that cannot perform at that frequency with a specific voltage.

I paid $1250 for my 1950X at launch from them, which due to their limited cooling, was only rated to like 4GHz at a specific voltage. I currently can run that chip at 4.2GHz@1.375V on all 16-cores. It is truly a golden sample among 1950Xs (I also have monstrous cooling with 3x480mm rads). I also lucked out on a decently strong IMC, which is not binned for, being able to run 3466-3600 with CL14 since the beginning. If I wanted to run lower speeds, I can drastically drop my voltage and heat output.

With Intel chips, you normally see a 300-500MHz range on CPUs from SL. With Ryzen gen 1, you only saw about a 200MHz range, which were binned into 50MHz steppings. Because so little OCing headroom was left, many opted to get them from off the shelf. But, with Ryzen 3000, the IPC is drastically increased. That means that each 50MHz to 100MHz stepping has more impact on performance than prior generations. We still need to see what is left in the tank, though, as far as speeds that can be hit and at what voltages.

Why does this matter? People are complaining about a premium. Let's put that in perspective, though. How much will that performance benefit you over the lifetime of your product? With a workstation CPU (or any production machine where the owner is willing to overclock to varying degrees, or to allow it to boost higher under load with PBO, which now allows for 200MHz over the rated speeds on the spec, or removing the limiter for the boost and setting the limits according to the three metrics of PPT/TDC/EDC), getting a little more performance out of the chip CAN be worth it (although validation WILL take longer and needs to be extremely rigorous while leaving headroom on temps and voltages for unforeseen workloads; this is even more so with RAM overclocking (I average at least two weeks of validation followed by regular monitoring on this point); also, note, this is NOT to be done on mission critical machines!!!).

For other binning companies, if in Europe, look at Caseking and the affiliated OverclockersUK, which are affiliated with Der8auer and 8Pack, respectively (two extreme overclockers).

Note: Overclocking may void your warranty as you are running the product out of spec.