Fox2232:

Chiplet that does not meet 3950X requirements goes to 3800X or 3700X. Chiplet that does not meet 3900X requirements goes to 3600X or 3600. Considering 3600 is 3.6~4.2GHz chip, 3900 in this thread is lowest matching as it is 3.1~4.2GHz.

Chiplet that does not meet 3950X requirements, MIGHT go 3900x, 3800X or 3700X. If you think about it, they will aim to bin the best for 3950x and the started already since it was launched...trust me , this is how business goes. They binned and looking through stats understood / learned that they have many that could not land on either 3900x/3800x/3700x speed wise thus they are creating this new 3900 to get some profit. It is that simple really.. I do not think we need to argue on this, AMD has been always smart on using resources and it is fair doing so. Depending on price 3900 can be useful for some workstations.. My understanding from these circumstances is that 3950x will be really limited ...like REALLY limited...price will skyrocket due to this and already did...soo good luck getting one for a good price.

[ZoNk]Kostas007:

or you can get the 3900 and overclock it as a 3900x with less money

no you can't, maybe you haven't followed or didn't understand what was going on these past months but buying cheaper and overclocking is over, that era has ended - AMD sells you cpu that can borderline do what they should, there's no (serious) overclocking ever, not with normal or even watercooling for normal every day usage - Intel now follows by rebranding or splitting the 9900k that can or not manage 5.0Ghz on all cores I'll repeat a comment I made months ago, cars (at least europeans) are the same nowadays, years ago they over-built engines and you could easily get more horsepower out of them, now they make lighter engines that can sustain the power they sell the car for and not much more..which is...absolutely logical ! why would you sell something that does "15" for the price of something that does "10" it makes no sense buy something that does what you want, don't expect to get much more

Xdrqgol:

Chiplet that does not meet 3950X requirements, MIGHT go 3900x, 3800X or 3700X. If you think about it, they will aim to bin the best for 3950x and the started already since it was launched...trust me , this is how business goes. They binned and looking through stats understood / learned that they have many that could not land on either 3900x/3800x/3700x speed wise thus they are creating this new 3900 to get some profit. It is that simple really.. I do not think we need to argue on this, AMD has been always smart on using resources and it is fair doing so. Depending on price 3900 can be useful for some workstations.. My understanding from these circumstances is that 3950x will be really limited ...like REALLY limited...price will skyrocket due to this and already did...soo good luck getting one for a good price.

No. 3950X uses 8C/16T chiplet. Most of chiplets that get out of fabs are without need to disable 2cores. 1st step is actual disablement of bad cores. That greatly reduces opportunity to even get chiplet into 3900X/3600X/3600/3900 category. But as it gets there, 3900 is easy to make as only one chiplet needs to meet boosting requirements. Other can boost to 4GHz only and still be used. 3950X is only about clocks at some voltage target. As production improves over time, 3950X will be plenty. 3900X will become rarer than it is now. What will improve is actual spread of demand. Many of those waiting for 3900X will go for 3900 or 3950X.

Folks, remember, AMD is a business. They are in business to make money, first and foremost. So as a business, if I do X amount of production runs and end up with a ton of 6c 12t dies, and it costs me Y$, I have two choices here. 1.) I could send a truck full of 3500/3500x (unreleased), 3600/3600x which are plentiful and not in short supply. 2.) I could (seeing as these don't quite always cut the mustard for 3900X), make a new slightly lower-power-target SKU for those wanting the "in-high-demand 12-core 24-thread 3900x chips" at 65w VS 105w 3900 12-core 24-thread chip. These 12-cores have been flying off shelves, and this will help sate the market demand a bit for the 3900x, allowing the business (AMD) to satisfy some demand for the 3900x with the 3900 *AND* give people an alternative. PLUS, these rake in TWICE the money per box in the truck, and seeing CPU's from one angle as solely cash per cubic meter, a truckload of 3900's at 400$*, the 3900's would be worth almost double what the 3600-class chips would be, on a basis of per truck shipment. (*give or take, maybe it'll knock the 3800x down a bit, that'd be logical) So as you can see, from a business perspective, if you take those 6c 12t dies and throw two of them in a CPU, you're going to make twice as much per box in sales as you would from a 3600 - and that's completely ignoring the product stock / demand situation. Taking this demand situation into account, it's only logical that AMD would desire to meet market demand *AND* increase the bottom line. Remember folks, it's not only a missed sale if someone doesn't have the ability to purchase a 12-core 24-thread chip from AMD - they may indeed grab an intel offering, which is not only a missed sale for AMD, but a sale for the competition! That's super-bad in business economics - you never want that. It's almost better to give customers things close to at-cost, in turn keeping your customer very happy and generally loyal, then to provide your competitor with a sale giving the customer an opportunity to become associated with another brand/business where you may lose them in the long-term. "AMD's contract with TSMC is suffering from poor yields, early process node etc" It's one of the only 7nm offerings they could pick up, to allow them to do what ZEN was truly designed to do. I would have done the same thing AMD did. It sure has it's risks, but it's slowly beginning to really look like the diamond in the rough in a sea of otherwise over-plentiful intel quad-cores. I'd be quite thankful for what you get, it's moving the industry forward. AMD is just doing what NVidia, Intel and all other people who make hardware in tiered product rangers do, binning and segmenting the market as demand and supply dictates. Remember, ALL manufacturers do this!!! AMD's very smart to make the 3900 - people want an upgrade, not 'just a little more' cpu than their quad-core intel chip, they want DOUBLE, or at-least a substantial noticeable bump in performance. They want to feel good with their purchase of new computer goodies when they have an upgrade itch - or just the need for more power under the hood so-to-speak. I know I was, when I replaced my 5-year-old 4790k machine with a 3700x / Asrock x570 Phantom Gaming 4 system. I can say I am overly satisfied with it, and how it runs BeamNG Drive maps, content creation, modeling, graphics and sound work, and yes even RimWorld of all things. Completely wiped the floor with the 4790k, and surpassed all expectation by leaps and bounds - the gobs of L3 cache REALLY help, too. *note / off-topic / for the record : No boost issues either, goes over rated spec on release day bios with the *stock* cooler. Scraped off (but did not scrub off) most the stock thermal paste with a credit card, and put merely a quarter grain of liquid metal (as any more would squeeze out) on the top of the CPU, and I don't ever see temps of 80c even (too little is better than too much). No windowed case, no LED's except the CPU cooler fan, no flashy disco-ball hanging in there (I'm almost 40). Just a non-descript black Fractal full-tower that's sound-insulated that does everything I need and keeps going without a fuss. Part of me wishes I had gotten the 12-core 24-thread chip, but ah-well, I'll grab the 16-core when if ever it sees a mark-down in a year or two. No unrecoverable WHEA PCI-E errors either, no blue-screens. Just for the record. That said: if you just game, even a 4790k which I upgraded from, will be just fine for a few more years unless things are really getting slow for you with a lot of units on screen in certain titles. No benchmark or review could have prepared me for the AMAZEMENT and jaw-opening performance improvement most of my programs saw. I had no idea. @Hilbert Hagerdoon (sorry if I spelled it wrong), please consider adding BeamNG Drive to your benchmark list; running lots of cars with traffic really works the CPU to max, and no-one seems to be using it currently. You can add in AI traffic via a traffic app in the app panel of the game.

kakiharaFRS:

no you can't, maybe you haven't followed or didn't understand what was going on these past months but buying cheaper and overclocking is over, that era has ended - AMD sells you cpu that can borderline do what they should, there's no (serious) overclocking ever, not with normal or even watercooling for normal every day usage - Intel now follows by rebranding or splitting the 9900k that can or not manage 5.0Ghz on all cores I'll repeat a comment I made months ago, cars (at least europeans) are the same nowadays, years ago they over-built engines and you could easily get more horsepower out of them, now they make lighter engines that can sustain the power they sell the car for and not much more..which is...absolutely logical ! why would you sell something that does "15" for the price of something that does "10" it makes no sense buy something that does what you want, don't expect to get much more

So why do I reach a 2600x while overclocking my 1600. Yea thought so. And your anology of small turbo engines for the average vs what you can do on your own is really stupid.

@Agonist i would agree although there is tweaking to be done to squeeze out every single percentage of performance you can ... the days that you would get a 3ghz chip and practically add 33% of performance or more by boosting to 4+ ghz are long gone . @TheThread maybe i am getting old but the hunt for mhz is stale for me i see people push their cpu to absolute max for example ((imaginary cpu at 3.5ghz stock)) instead of 4.2 ghz to 4.4 where they need increase voltage to get and all that for what ? get 3-4 fps on a game ... finish a render in 57 seconds instead of 58 ... i would gladly take the take the 4.2 ghz i do not think i loose almost anything and is much more sustainable in the long run tweak ... the only reason i see on it is to do it cause is a hobby and you enjoy doing it ... i even see some people changing their cpu or over paying for 100 or 200 mhz witch i find absurd how big of a difference has an r7 1700 @ 3.8 vs one that can do 3.9 or a 9900k that can do 5.1 with one that can do 4.9 at this point in my opinion the only way to get really better performance is to get a better cpu ...either more cores if your workload favor em ..or wait for the next generation/s

Would be a nice surprise if this CPU is destined for the new PlayStation or Xbox

Venix:

@Agonist i would agree although there is tweaking to be done to squeeze out every single percentage of performance you can ... the days that you would get a 3ghz chip and practically add 33% of performance or more by boosting to 4+ ghz are long gone . @TheThread maybe i am getting old but the hunt for mhz is stale for me i see people push their cpu to absolute max for example ((imaginary cpu at 3.5ghz stock)) instead of 4.2 ghz to 4.4 where they need increase voltage to get and all that for what ? get 3-4 fps on a game ... finish a render in 57 seconds instead of 58 ... i would gladly take the take the 4.2 ghz i do not think i loose almost anything and is much more sustainable in the long run tweak ... the only reason i see on it is to do it cause is a hobby and you enjoy doing it ... i even see some people changing their cpu or over paying for 100 or 200 mhz witch i find absurd how big of a difference has an r7 1700 @ 3.8 vs one that can do 3.9 or a 9900k that can do 5.1 with one that can do 4.9 at this point in my opinion the only way to get really better performance is to get a better cpu ...either more cores if your workload favor em ..or wait for the next generation/s

Yea I do miss pulling some cool overclocks. I had my 2.66ghz X5650 @ 4.6ghz with just an H100. Honestly I just go for the best overclock, not max overclock. My 1600 will do 4.1 but most of the time I run it at 3.75 because it only needs a tiny bump in voltage to do so and temps stay way low. And in gpu bound games, i usually run stock 3.4 boost.

Shagula:

Would be a nice surprise if this CPU is destined for the new PlayStation or Xbox

Its more likely that the Xbox/PS chips will be like the unreleased Ryzen chips for the rumoured new Surface laptop. Low power, 8c/16t with a largish discrete GPU.

Shagula:

Would be a nice surprise if this CPU is destined for the new PlayStation or Xbox

Rather them use 8 cores and put more budget in the GPU.

Agonist:

Yea I do miss pulling some cool overclocks. I had my 2.66ghz X5650 @ 4.6ghz with just an H100. Honestly I just go for the best overclock, not max overclock. My 1600 will do 4.1 but most of the time I run it at 3.75 because it only needs a tiny bump in voltage to do so and temps stay way low. And in gpu bound games, i usually run stock 3.4 boost.

yeah it was a time that you where getting reaaaaal value out of it the best cpu i personally overclocked my self was an i7 920 from 2.6 to 3.8 for a friend with a mugen 3 on top of it ... my friend wanted me to push it farther and i really could but is where i stopped and i told him that's as far i am willing to go with it cause in case something goes wrong i will not be around to put it back in order but that was +40% ++ on clock speed ((my friend is not tech savvy at all when ever he wants to update he calls me he does not even know what he has in his pc but i do .. lol )) . My self on my 1600 i just went into bios i changed the values to the 1600x levels did a stress test to see if it is stable and let it be pushing it farther with the consumption curve of ryzen it is not worth the hustle for the gains :P i also had my i7 3770k from 3.5 @ 4.3 not huge but was with minimal voltage extra i could push 4.5 but was a lot of extra vcore and over 15 celsius jump to do that ((undelided)) so i just left it for years on 4.3

kakiharaFRS:

no you can't, maybe you haven't followed or didn't understand what was going on these past months but buying cheaper and overclocking is over, that era has ended - AMD sells you cpu that can borderline do what they should, there's no (serious) overclocking ever, not with normal or even watercooling for normal every day usage - Intel now follows by rebranding or splitting the 9900k that can or not manage 5.0Ghz on all cores I'll repeat a comment I made months ago, cars (at least europeans) are the same nowadays, years ago they over-built engines and you could easily get more horsepower out of them, now they make lighter engines that can sustain the power they sell the car for and not much more..which is...absolutely logical ! why would you sell something that does "15" for the price of something that does "10" it makes no sense buy something that does what you want, don't expect to get much more

You are totally correct. I remember the good old days where I could easily oc a chip around 50%. (And 3000+ Venice core) Heck i remember overcloking a GPU more than 100% with a stock cooler, just by adjusting a slider in the driver control panel. (Radeon 9550) Those overclock margins are now gone cos manufacturer s just clock it as high as possible and sell the item.

Clawedge:

You are totally correct. I remember the good old days where I could easily oc a chip around 50%. (And 3000+ Venice core) Heck i remember overcloking a GPU more than 100% with a stock cooler, just by adjusting a slider in the driver control panel. (Radeon 9550) Those overclock margins are now gone cos manufacturer s just clock it as high as possible and sell the item.

Its the Fab PRocess thats done it. Gradually, as the Fab process has shrunk, we've got less and less margins for overclocking. I really miss the 100% OC days too 😉 Its just not going to come back. It might even get tighter and tighter, until a Major Fab change happens, ie New Fab Materials other than Silicon or something.

Evildead666:

Its the Fab PRocess thats done it. Gradually, as the Fab process has shrunk, we've got less and less margins for overclocking. I really miss the 100% OC days too 😉 Its just not going to come back. It might even get tighter and tighter, until a Major Fab change happens, ie New Fab Materials other than Silicon or something.

No, companies just became better at extracting almost every last drop of performance out of their products in order to properly compete. CPUs and GPUs are better than ever at figuring out exactly what clockspeed to run at given the workload, temperatures, power draw, and chip quality. We're not doing better now, and even if we were in some niche situation, we soon won't be. I mean, look at some overclocks mentioned up there. We're talking about 40-50% gains. That's several performance brackets above and in the current day and age, it is unthinkable for that to be left on the table with the gains being incremental from generation to generation.

yeeeeman:

You can have this CPU right now if you underclock your 3900X...

I said before that the 3900X runs hot as the sun, so I want to throw this out there: You don't lose much if you want to run a 3900X at low temps. I was running at 4.3GHz using 1.39V and that was pretty hot, I didn't want to even try going above that. Turns out 4.2GHz only takes 1.33V at most (maybe I can go lower), and even the most insane stress tests (Cinebench and Handbrake are way more stressful than traditional artificial tests) don't do much to the temps. That is using decent cooling though, an NH-D15 (SE-AM4) with their own thermal paste which is one of the better ones.

Evildead666:

Its the Fab PRocess thats done it. Gradually, as the Fab process has shrunk, we've got less and less margins for overclocking. I really miss the 100% OC days too 😉 Its just not going to come back. It might even get tighter and tighter, until a Major Fab change happens, ie New Fab Materials other than Silicon or something.

Though the process of major fab changes have slowed, it realistically doesn't have anything to do with OCing. "Back in the day" You had single-core processors, and the only realistic way to make different SKUs, were to have different default clocks on a variety of prices and possibly some cache changes. Now, we have cores, and minute differences relatively speaking on the clock. Take the Althon XP , where you had Athlon XP 2000+ @ 1.6Ghz all the way to the Athlon XP 3200+ @ 2.3Ghz from 2003 to 2004. You could if you wanted to get the 1.6Ghz and hope you can overclock it to at least 2.3Ghz, but buying the Athlon XP 3200+ @ 2.3Ghz you didn't have a lot of OC headroom left, 2.4Ghz was pretty typical, that is if the processors were unlocked, etc. In between the 2000+ and 3200+ were 19 different processors, many of which overlapped eachother due to release dates in 2004 from Thorton to Barton. But the point i am making here is, these were all single-core CPUs with mostly the difference of default frequencies, when it came to real-world performance. Now, we have multi-core processors, where the race is mostly the cores and over-all CPU performance. Now what is the major point in clocking many models low, other then to over-saturate their own portfolio? Could you imagine there being even 6 different 4-core, 6 different 6-core, and 6 different 8-core processors, among whatever cores you have (2/12/16/32/etc.) with a varying degree of different core speeds? You'd effectively have to either have a lot of similar if not exactly the same cost (example: Have a 6-core 2.0Ghz processor for $200, or a 4-core 3.0Ghz for the same $200) or spread out the cost between chips so far you'd have 30-40 different CPUs from $100-3000 That'd be crazy, which is why we get not hugely different frequencies within the same core-count, and very near their max frequency in the first place. Instead of getting significantly lowered CPU frequencies, we get less core counts.

K.S.:

Been reading on that recently; lot of rumors pointing to some existing variant of a 3000 AMD SOC revision. Personally - I'd be more interested with AMD setting aside series 4000 for the next consoles as their road-map stated it was done. Leaving one to assume it'd be a matter of manufacturing left for that series?

Developers are going to need finalized products to developer their games on for initial and immediate releases. If they were to wait for 4000 series, the consoles will have to wait until late 2021 release date

Neo Cyrus:

I said before that the 3900X runs hot as the sun, so I want to throw this out there: You don't lose much if you want to run a 3900X at low temps. I was running at 4.3GHz using 1.39V and that was pretty hot, I didn't want to even try going above that. Turns out 4.2GHz only takes 1.33V at most (maybe I can go lower), and even the most insane stress tests (Cinebench and Handbrake are way more stressful than traditional artificial tests) don't do much to the temps. That is using decent cooling though, an NH-D15 (SE-AM4) with their own thermal paste which is one of the better ones.

This is false. 3900X has 2 chiplets that have same dimensions as 3800X while stock power draw is only bit higher than 3800X. 3900X may produce more heat under full load, but it has easier time to dissipate it. I am using NT-H1 as you are. But in case I'll ever get 3900X, I already have CM Master Gel Maker in shopping cart. (Just not to forget.)

Aura89:

Developers are going to need finalized products to developer their games on for initial and immediate releases. If they were to wait for 4000 series, the consoles will have to wait until late 2021 release date

I think that they need CPU with same feature set and similar to target performance. (Which can be tuned.) And then they'll need GPU capable to do same. If not, then they need those instructions emulated in API to enable development of visual effects. Then do final performance tuning as proper device gets to them. I am pretty sure that CPU wise even 1700 would do. GPU wise, that's hard part as there is real development of new instructions with magnitude higher performance on given workloads.

Fox2232:

This is false. 3900X has 2 chiplets that have same dimensions as 3800X while stock power draw is only bit higher than 3800X. 3900X may produce more heat under full load, but it has easier time to dissipate it. I am using NT-H1 as you are. But in case I'll ever get 3900X, I already have CM Master Gel Maker in shopping cart. (Just not to forget.) I think that they need CPU with same feature set and similar to target performance. (Which can be tuned.) And then they'll need GPU capable to do same. If not, then they need those instructions emulated in API to enable development of visual effects. Then do final performance tuning as proper device gets to them. I am pretty sure that CPU wise even 1700 would do. GPU wise, that's hard part as there is real development of new instructions with magnitude higher performance on given workloads.

They have had the Devkits for some time now. Probably based on Navi, but with the RDNA2 drivers. iirc Navi was a bit of a hybrid between GCN and RDNA ? I agree, the CPU could be almost anything Ryzen based.

Aura89:

Though the process of major fab changes have slowed, it realistically doesn't have anything to do with OCing. "Back in the day" You had single-core processors, and the only realistic way to make different SKUs, were to have different default clocks on a variety of prices and possibly some cache changes. Now, we have cores, and minute differences relatively speaking on the clock. Take the Althon XP , where you had Athlon XP 2000+ @ 1.6Ghz all the way to the Athlon XP 3200+ @ 2.3Ghz from 2003 to 2004. You could if you wanted to get the 1.6Ghz and hope you can overclock it to at least 2.3Ghz, but buying the Athlon XP 3200+ @ 2.3Ghz you didn't have a lot of OC headroom left, 2.4Ghz was pretty typical, that is if the processors were unlocked, etc. In between the 2000+ and 3200+ were 19 different processors, many of which overlapped eachother due to release dates in 2004 from Thorton to Barton. But the point i am making here is, these were all single-core CPUs with mostly the difference of default frequencies, when it came to real-world performance. Now, we have multi-core processors, where the race is mostly the cores and over-all CPU performance. Now what is the major point in clocking many models low, other then to over-saturate their own portfolio? Could you imagine there being even 6 different 4-core, 6 different 6-core, and 6 different 8-core processors, among whatever cores you have (2/12/16/32/etc.) with a varying degree of different core speeds? You'd effectively have to either have a lot of similar if not exactly the same cost (example: Have a 6-core 2.0Ghz processor for $200, or a 4-core 3.0Ghz for the same $200) or spread out the cost between chips so far you'd have 30-40 different CPUs from $100-3000 That'd be crazy, which is why we get not hugely different frequencies within the same core-count, and very near their max frequency in the first place. Instead of getting significantly lowered CPU frequencies, we get less core counts.

Basically, what i'm saying is that when we overclocked, the Wall would come progressively, back in the day. As time goes on, and the fab process tech has evolved, the wall comes faster and faster. You just get a little OC room, and the the voltages just have to go straight up a cliff, whereas back in the day, it was more like a light hill, to moderate incline, until you hit the cliff. (Member Phase change coolers ? 🙂) I remember drooling over those Athlon systems running at insane speeds....;) Now its like a hockey stick, and the manufacturers are clocking all the way up to those speeds, because they can reliably. I suppose they've just got really good at fabbing now, and have dialled it in that way for profits, because it certainly makes the chips a lot more equal.