AMD's Future Chips & SoC's: News, Info & Rumours.

Page 98 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

InvalidError

Titan
Moderator
https://www.pcgamesn.com/amd/ryzen-9-3950x-best-gaming-cpu

"AMD’s “special binning” the 16-core Ryzen 3950X to make it “your best gaming part”

"“We were special binning the ones to get the 16 cores,” AMD’s Sasa Marinkovic tells me, “and they are more exclusive and harder to bin, andharder to get.”
Given the size and cost of wafer-level test machines with hundreds of test channels along with the fact that every chiplet has to pass through one of those anyway to determine which die are viable and worth sawing out of the wafer, I seriously doubt the "special binning" is separate from the normal binning process. Most likely, the "special binning" is nothing more than a chip bin in the normal binning process for chiplets that meet whatever "special" criteria AMD is looking for on the 3950X.
 
  • Like
Reactions: goldstone77
I will also call that big BS coming* from AMD's marketing dept.

It's in TSMC's best interest having a reliable and homogeneous process as they can possibly get, so this is just piggybacking the "silicon lottery" phrasing.

And as for the OC talk. One silver lining about Overclocking that you all must never forget: the whole point of it was to tweak the CPUs in an off-spec way to make them over-perform. What we have now is a kiddy sandbox; not that I'm being ungrateful as it's easy to squeeze as much performance as you can nowadays, but please keep in mind today's OC is not the same as previous decade's OC. Granted that CPUs are now 100x more complex beasts, but the whole idea of OC has been to tweak beyond what the sandbox allows you to. I don't know if there's room still for that "old school" OC spirit, but I'm sure we've just grown lazy, in general, when talking about enthusiasts and overclocking. I'm now old and lazy, so I'm fine with that, lel.

Cheers!
 
With features like xfr and gpu turbo states, some high end products will automatically run at close to overlclocked speeds if cooling and power delivery allows.
The best overclocking comes from lower end chips.
You could take an i3 8350k and push it close to 5ghz, nearly 1ghz over its max factory frequency.
Try bumping a high end 9900k to 1 ghz above its factory turbo, and it wont even get close.
 
  • Like
Reactions: rigg42

goldstone77

Distinguished
Aug 22, 2012
2,245
14
19,965
Can you give me a summary, please? I can't watch the video now XD

Cheers! :p
He basically talks about his 2700X and benchmarks in relation to all core overclocking vs. stock. I believe he is saying that XFR allowing one or two cores to jump higher than using an all core overclock, is better for him when testing in many cases.
 

InvalidError

Titan
Moderator
I believe he is saying that XFR allowing one or two cores to jump higher than using an all core overclock, is better for him when testing in many cases.
What he says in the video is basically that XFR already does a good enough job that an all-cores overclock that barely hits XFR boost isn't worth the trouble in most cases and he's shown at least two games where the higher single/dual-core XFR boost is actually better than a lesser all-cores overclock.
 
  • Like
Reactions: goldstone77
Those are also close to my findings but he missed to say which MB and BIOS as well as RAM frequency and timings of which faster ones have some good influence on single core scores more than on multi core. That and the lousy sample CPU he got. makes his tests "ah not so valid". I got much better scores than he did and somewhat larger spread between OC and PBO scores. That's of course all synthetic and largely unnoticeable in real world.
 
  • Like
Reactions: rigg42
Those are also close to my findings but he missed to say which MB and BIOS as well as RAM frequency and timings of which faster ones have some good influence on single core scores more than on multi core. That and the lousy sample CPU he got. makes his tests "ah not so valid". I got much better scores than he did and somewhat larger spread between OC and PBO scores. That's of course all synthetic and largely unnoticeable in real world.
I think I shared my own scores when I got the CPU to a few fellas in the forum and I beat every single review on both pure OC and just toying with PB2. I tried different things, but I never tested for full stability, so it's even trickier.

The important take-away is that, as it stands and out of the box, the net difference between a good/mediocre OC and what PB2 is capable of is not far away and, in some scenarios, it even performs better.

Cheers!
 
  • Like
Reactions: NightHawkRMX

rigg42

Respectable
Oct 17, 2018
639
233
2,390
I watched that video. I thought it was kind of dumb and obvious. There is not much point in buying an X SKU and manually overclocking it. This has been well established since these CPU's were released. If you are going to manually overclock why wouldn't you save the money on the non x and buy a good cooler to overclock it? The entire crux of the argument is meaningless because value isn't a part of the equation. I don't know how you make a video on this and not mention the 2700 or the cost of accompanying components required to achieve the boost speeds shown in the charts. You won't be hitting 4.35 on a prism cooler. The entire point of the video seemed to be he was implying that the auto boost on the new chips is better than manually overclocking without violating NDA. I generally like this guys content but this was a BS video.

Getting the most out of either CPU requires a good cooler, a good motherboard (more important for the X), significant configuration in bios, and stress testing. I'd bet money I could get more performance out of 2700 system for the same or less money than a 2700x. You can spend less on the CPU and motherboard and more on cooling and RAM with a 2700. I'd argue it takes more knowledge, skill, and significantly more money to get a 2700x to maintain 4.2 all core at full load and consistently boost to 4.35 than it does to manually overclock a 2700 to 4.2 all core. Not really worth it for 5-10 fps in certain games. Especially if you took the $50 or so you saved on the CPU and put it toward a B-Die RAM kit which would probably nullify the clock speed advantage and give better performance across the board.
 
  • Like
Reactions: CountMike
Ryzen 2700X can be bought for 200 dollars now, Imagine the price drop during season sale. I hope Ryzen 3600 justifies it's price being a 6 core cpu :p.
I think the 3600X will be able to justify itself having the same price as the 2700X on the account of having higher clocks and reaching parity in multi-threaded scores; oh, this is assuming the single/low cores performance is higher than the 2700X as well. Maybe that's why the 2700X is $200 now, as it will be a hard sell when the 3600X launches?

Cheers!
 

InvalidError

Titan
Moderator
I think the 3600X will be able to justify itself having the same price as the 2700X on the account of having higher clocks and reaching parity in multi-threaded scores; oh, this is assuming the single/low cores performance is higher than the 2700X as well. Maybe that's why the 2700X is $200 now, as it will be a hard sell when the 3600X launches?
Based on the "leaked" benchmarks from the past week or so, it does look like the 2700(X) will be in a tough spot a few days from now.
 
  • Like
Reactions: goldstone77
Based on the "leaked" benchmarks from the past week or so, it does look like the 2700(X) will be in a tough spot a few days from now.
Single thread 3600x is already higher than 2700x and MT is damn near. Single thread is bound to stay pretty level on all 3rd gen with some ups and downs according to frequency. What is unusual for AMD is that higher core count, higher frequencies.
 
  • Like
Reactions: NightHawkRMX

jdwii

Splendid
Yes the reason why i showed that video was that i firmly believe he was implying that Ryzen 3000 series chips their X versions have been pushed close to their limits out of box. Though i'm again guessing 4.3-4.5ghz all core stable overclocks will be possible on the 8 core models and below. With just a AIO or high-end heatsink.
 

InvalidError

Titan
Moderator
Yes the reason why i showed that video was that i firmly believe he was implying that Ryzen 3000 series chips their X versions have been pushed close to their limits out of box.
With both first-gen and second-gen not overclocking past XFR boost on all cores without disproportionate effort, I'd say AMD appears to have done a good job getting nearly all the practical performance it can out of X-chips. I see no reason why AMD wouldn't aim to do the same with Zen 2 products. I'm not expecting Zen 2 to overclock much past 4.5GHz all-day all-cores on air-based cooling either. If a large enough number of AMD's chips could achieve sustained higher clocks, AMD would have bumped the specs and likely launched additional more finely-binned products.
 

JaSoN_cRuZe

Honorable
Mar 5, 2017
457
41
10,890
I believe that buying an x570 motherboard for 3700X and below would not be budget friendly especially with an X570, Waiting for the same features except Gen 4 in B550 would be a much better combo for 3700X & below. Gen 4 is for enthusiasts for now. Who knows B550 boards may launch closely with Ryzen 4000 series on its tail, similar to B450 & Zen+.
 
I believe that buying an x570 motherboard for 3700X and below would not be budget friendly especially with an X570, Waiting for the same features except Gen 4 in B550 would be a much better combo for 3700X & below. Gen 4 is for enthusiasts for now. Who knows B550 boards may launch closely with Ryzen 4000 series on its tail, similar to B450 & Zen+.
That's my thinking too unless one wants PCIe 4 for any price. Less power draw, potentially lower TDP do not require anything new. With such compatibility, MB manufacturers are not likely to stop 400 series chipset MBs very soon and 500 series seem to be more expensive to manufacturers.