Ryzen 3000 listed online early on russian site.

[SL2]

Isn't there a 10-15% IPC bump on top of that though? If so then it will be quite a bit faster.

Yeah, probably, but I was just talking about the specs in the link, that some call "hyping things out of proportion "..

 

[Wavetrex]

The only thing that happened here is Intel being lazy and greedy and allowing AMD to catch up. Minimum effort, maximum profit.
The last significant architecture change was from Nehalem to Sandy Bridge, after that all of them are "optimizations" of the same shiet. A bit more cache, a bit lower latency, same bus, same weak iGPU, same everything.

How about the reverse of "Too good to be true" ?
We had 10 years of "Too bad to be false", yet it happened... Intel gave us 5% there, 3% there, and until Ryzen, a bloody QUAD CORE with 100Mhz more turbo than the last gen was everything we got.

Seriously, AMD deserves to win this round, and win HARD.
9900K is a total failure in my opinion, a CPU that needs to run at 100 degrees to actually sustain that promised 5Ghz on anything less then highest end water cooling... I pity those who bought it end of last year, KNOWING 7nm is coming, and soon !

 

[kastriot]

Well, we will see what is true and what not @CES

 

[robot zombie]

This is the exact reason I went with Intel for the gaming rig....

Going the 1700 > 2700x > 3700x route would cost a ton and limit my ram options. If this is true then I can expect price drops on the 10 core x299 chips, if it isn't then the 7820x @ 5ghz will last me some time anyways. The AMD chip releases are amazing but it would send my upgrade itch into full meltdown mode. I'm bad enough as it is...

Dude... tell me about it. It is killing me. But, I mean, you don't *have* to do it I like my 2600 and realistically could happily hold onto it for years, and if the price is wrong on the 3000 series maybe I actually will keep it, but goddamnit if the improvements are what people say they are the 3700x is gonna be calling my name.

I have my ways of justifying it. I am a parts hoarder though, sometimes you come into things or have tie-over's you know? I've got a refurb B350 I could toss the 2600 in... ...or maybe like I have in the past I'll pass it down to someone who could use the upgrade. I did a super-budget AMD build for a friend who would love to have what I have now. At least then I can say to myself that I'm doing something good for somebody... if not myself lol.

I dunno, it's neither here nor there. One of Intel's most noted detriments is only giving us marginal improvements year after year. And then here comes AMD, seemingly making some real leaps with each release. Personally I think that is a good thing for all of us. At any given time doing a build, you're going to have something better than last time for your cash. But then, if you wanna save a little, last gen is still great compared to the one before it, even if you're compromising on having the latest and greatest *shrug*

I guess with things this way, you never really need to upgrade, but if you want to, at least it's actually an upgrade. Even if it does test your self control a little lol

 

[Asviar]

Still can't imagine why do I would need 16 cores. IMHO, 12 is almost that average PC user may need within next few years.

 

[HD64G]

People need to understand that 7nm allow twice the size for the same power comsumption. So if AMD wants it, they can release a 16-core for the same clocks as 2700X that will have the IPC-cache-memory latency improvements (10-15%) and consume the same as 2700X. Or, they can choose to get the clock increase allowed by the better process (>25%) and end having a 5.5GHz (single core boost that is) 8-core CPU in the same power envelope. Since the first products cannot be so well binned, they will have some CPUs boosting up to 5GHz, which is easily achievable since 5.5GHz is the limit, that will be of different core number and power envelopes, keeping the best for TR and Rome ones that have much higher profit margin. We will surely have big competition this time for gaming CPUs imho. Zen and Zen+ were already competitive in almost every other aspect of compute power. As customers and PC enthusiasts, let's hope for the best tech in good price.

 

[Wavetrex]

12 is almost that average PC user may need within next few years.

The human eye cannot see more than 12 cores

 

[Asviar]

The human eye cannot see more than 12 cores

Yeah, average human eye

 

[robot zombie]

Still can't imagine why do I would need 16 cores. IMHO, 12 is almost that average PC user may need within next few years.

I said the same thing when Zen+ hit, and I'd still agree. I think 6 physical cores is solid ground to be on now and in the future.

Not much use for more, save for special applications. But I think enthusiasts are still rightly excited that higher core counts are more accessible. There was a time not too long ago when you couldn't pay any amount of money to easily get the core counts AMD brings, that or you spent a few thou. I know I'd love to have a TR to play around with, even if I don't really need it for my actual usage.

Another way to look at it is that if 12 cores (or 6+6) is the most your average user would need these days, then all is really as it should be, with that level in the midrange, and higher counts being in the top-end, for the smaller number of people who need that. What's available at the top should always be more than people need in a healthy market. Absolute best of best as a point of entry to meeting one's needs sucks.

 

[SL2]

The only thing that happened here is Intel being lazy and greedy and allowing AMD to catch up.

Plus Intel having a real hard time with their 10 nm process. After the first 10 nm shrink, Intel planned to release the next big thing since Nehalem, but 10 nm didn't happen.

I'm sure some people would disagree, but to me it seems plausible. From the first Core i CPU it all made sense, until after Skylake. After that it would have been a 10 nm shrink, something we're still waiting for. What we've seen since then is basically added cores/cache and more clock speed, nothing else. IPC backs this up (right?), it pretty much stalled after Skylake.

Not being able to do 10 nm isn't laziness, Intel WANTS to get there, and it hurts them big time for not being able to.

 

[OneCool]

3700X looks nice

 

[Patriot]

If that's true, then leaks about 8 cores per CCX were also accurate. We'll probably see some bad-ass 220-250W Threadreaper chips w/ 32C/64T!

That's not leaks, that's actual real chips which AMD demoed in the Rome chip.
64 cores, 8 chiplets x 8 cores.

The question is... will they use the same for Ryzen 3000, in combination with an I/O hub ? Or a more integrated design in a single larger die.
My gut tells me it will be a small I/O hub with 2 chiplets, that makes the most sense as they can fab all the 8-core chiplets with the same process and then just package differently for various markets.

CCX != chiplets FFS

a chiplet is 2 CCX's so 4c CCX.

That said after the adored single source, not his main source, take this with truck-fulls of salt video... AMD unequivocally said, Desktop Ryzen 2 (3000) will not be like rome, it will be completely 7nm.

 

[phanbuey]

Plus Intel having a real hard time with their 10 nm process. After the first 10 nm shrink, Intel planned to release the next big thing since Nehalem, but 10 nm didn't happen.

I'm sure some people would disagree, but to me it seems plausible. From the first Core i CPU it all made sense, until after Skylake. After that it would have been a 10 nm shrink, something we're still waiting for. What we've seen since then is basically added cores/cache and more clock speed, nothing else. IPC backs this up (right?), it pretty much stalled after Skylake.

Not being able to do 10 nm isn't laziness, Intel WANTS to get there, and it hurts them big time for not being able to.

They might just abandon and go straight 7nm.

 

[Vario]

Interesting article comparing Intel 10nm node to GF 7nm node
https://www.eejournal.com/article/life-at-10nm-or-is-it-7nm-and-3nm/

 

[xkm1948]

Interesting article comparing Intel 10nm node to GF 7nm node
https://www.eejournal.com/article/life-at-10nm-or-is-it-7nm-and-3nm/

GloFo canned their 7nm a while ago just FYI.

GloFo always made bold claims but failed to deliver, thus why AMD try to move away from GloFo to TSMC

 

[cdawall]

The bigger the chip (or the more cores that are enabled) the lesser that chances that it'll clock higher. Chances are those 12 core CPUs (6+6) or 4+4 will clock higher on average than anything else , including TR and EPYC.

That heavily depends on binning. The best binned parts go to TR hence why even the 2990wx can turbo almost as high as a 2700x. If any of the older rumors are believed true TR4 will have 5ghz turbo chips as well.

 

[Vya Domus]

That heavily depends on binning.

Better binning doesn't change the fact that the die itself is significantly less likely to clock very high. That's why you bin them in the first place.

 

[Zyll Goliat]

"Apparently"this should be prices for 3000 series!!!

 

[xkm1948]

Hype train full speed ahead

CHO CHO all aboard!

 

[Hellfire]

I have this ready for CES...

 

[Deleted member 158293]

Last time I was this eager for a CPU launch was the Nehalem Bloomfield processor launch... it's been a while... and honestly didn't think it would happen again.

 

[cdawall]

Better binning doesn't change the fact that the die itself is significantly less likely to clock very high. That's why you bin them in the first place.

Which die? The one that is identical to the ones you are talking about... That is the reason the current chips can clock the same regardless of core count and why an mcm can be superior to a monolithic design.

 

[Vya Domus]

and why an mcm can be superior to a monolithic design.

It's superior but this isn't the be all and end all.

That is the reason the current chips can clock the same regardless of core count

The point isn't that they can but that they are less likely to do so. A 64 core monolithic die is less likely to clock as high as a 16 core one in the same way that it's less likely to come by 8x8 core dies that they all clock to 5Ghz as opposed to just finding 2x8 ones. AMD made their job easier but they still have to deal with the same physical limitations and variance of the manufacturing process, no amount of binning can ever change that.

 

[cdawall]

It's superior but this isn't the be all and end all.



The point isn't that they can but that they are less likely to do so. A 64 core monolithic die is less likely to clock as high as a 16 core one in the same way that it's less likely to come by 8x8 core dies that they all clock to 5Ghz as opposed to just finding 2x8 ones. AMD made their job easier but they still have to deal with the same physical limitations and variance of the manufacturing process, no amount of binning can ever change that.

They have to put the absolute best binned cores on the high core count parts, otherwise they do not fit the tdp envelope. That's why you have people like me who actually have a high core count part and have it clocked as high on all 32 cores as most are able to hit on the 8 core parts.

AMD binning is pretty top tier and has been this whole time. That's one thing they have done well cpu and gpu.

 

[mastershake575]

I thought my 2700X would suffice for the next 5 years I guess not. Thank you AMD for hurting my wallet more, I just can't resist if it ends up real.

Whose to say your 2700x won't last 5 years ? Video game requirements have been stagnant the last 2-3 years and it will probably be at least 2 years before we see new consoles/game engines.

Even when the new game engines come out it will take a a year or two for them to get polished + your CPU will still easily be faster than whatever CPU is in the next gen consoles.

CPU demand/importance is extremely overrated. People where worried that the ivybridge/sandybridge would be only decent since they came out inbetween console generation and look at them now, 8 years later and they can still play games on very high settings.......