The cache will be spread over 3 dies and the average clock will be lower.
I want to see it first before I call an 24 core a gaming CPU.
The cache will be spread over 3 dies and the average clock will be lower.
I want to see it first before I call an 24 core a gaming CPU.
Because I'm not planning to make it run at 280W TDPIt will be across 4 dies and why would the average clock be lower? It has twice (if not more) the power allotment.
We, the answer is no. AMD isn’t targeting TR to enthusiasts anymore. It’s going to be for professionals in content creation, engineering, etc.
Well I'm only going to buy one system...the 24 core 4 channel is overkill not that good for games and the 16 core 2 channel is low for simulations.I don't see the mystery here. If you want a 16 core, get the 3950x, if you need more than that, get the TR3 platform with 24 cores or more.
I would wait until benchmarks come out. The 16 core is no pig in simulations, and the 24 core might actually be about the same in games, with the IO die, and the cores spread out more on that big HSF, it may even clock higher.Well I'm only going to buy one system...the 24 core 4 channel is overkill not that good for games and the 16 core 2 channel is low for simulations.
Where were you able to obtain this information? The NDAs for review reports haven't expired yet.I'm only going to buy one system...the 24 core 4 channel is overkill not that good for games
But isn't that a you decision not an actual outlook on the CPU itself. But even if you capped it lower you are looking at a 3600/3700x situation. A game loads even lets say 8 cores. Those will clock up pretty high even under a limited power allotment. Limiting the power usuage will only really hurt under heavy core usage like your simulation work.Because I'm not planning to make it run at 280W TDP
The 16 core AM4 is 105 W TDP
Yeah, the irony is that Intel X299 cost less per core than TR3, so AMD kinda left mid range systems to Intel with current pricing. Intel's 14 and 18 core offerings are priced real well for amount of performance they offer. AMD needs 16 core TR offering, but 4 chiplets, IO die, chipset on a pricey mb is a boatload of silicon to sell in a range of prices between desktop Zen2 16C and 24C TR3.
AMD are getting around any such limitations well enough in the server market, where they offer 8-core (8-ccx, 4-ccx, 2-ccx), 12-core (4-ccx), and 16-core (8-ccx, 4-ccx) Epyc 7002 SKUs. They'll have their reasons why they did not announce similar Threadripper 3000 models at this time.being able to have configuration smaller then AMD can get (because all of this revolves around AMD's limitations with its Die configurations)
AMD are getting around any such limitations well enough in the server market, where they offer 8-core (8-ccx, 4-ccx, 2-ccx), 12-core (4-ccx), and 16-core (8-ccx, 4-ccx) Epyc 7002 SKUs. They'll have their reasons why they did not announce similar Threadripper 3000 models at this time.
I would wait until benchmarks come out. The 16 core is no pig in simulations, and the 24 core might actually be about the same in games, with the IO die, and the cores spread out more on that big HSF, it may even clock higher.
I don't see the mystery here. If you want a 16 core, get the 3950x, if you need more than that, get the TR3 platform with 24 cores or more.
Dual-socket 2011-3, or single- or dual-socket SP3, or single- or dual-socket 3674 would be more appropriate choices for CFD.Say if you are wanting this for CFD
If the CFD problem sizes are too large for a 4-channel memory machine, run the solver on a dedicated machine.EYPC doesn't have the high ST speeds for interactive work (i.e. meshing or post-processing) for a workstation.
AMD say it is aimed at the so-called content creators; TR40 board makers say the same and video games.If TR3 is aimed at professionals
This includes only professionals who are ready to bite the bullet and use video game hardware for work.the Intel alternative may end up being the more attractive one due to better balance of price vs. capability
You just listed products that AMD already has competitive products in. The point was that AMD really can't do something that fits closer to the 1k-1200 without pushing the 24 down really low. I mean that would be awesome. But in the long run for many reasons it wouldn't make sense for AMD to have half their lineup overlap other well selling products with lower manufacturing cost. They have been careful not to price the CPU to much because of the platform when there is overlapp and selling an 8 core cpu with 4 ccds and a huge Io die for $450 bucks doesn't seem like a smart move. At least with Intel's hdet the when core size is mostly similar the die size isn't that much larger.AMD are getting around any such limitations well enough in the server market, where they offer 8-core (8-ccx, 4-ccx, 2-ccx), 12-core (4-ccx), and 16-core (8-ccx, 4-ccx) Epyc 7002 SKUs. They'll have their reasons why they did not announce similar Threadripper 3000 models at this time.
The single part among the three 8-core SKUs of Epyc 7002 which has 8 CCXs, and thus needs 4 partly functional CCDs, is priced higher than that (yet still fairly reasonably for the more exotic server CPU which it is). And this one may be a low volume part anyway.They have been careful not to price the CPU to much because of the platform when there is overlapp and selling an 8 core cpu with 4 ccds and a huge Io die for $450 bucks doesn't seem like a smart move.
Sales of 1950X and 1920X started earlier than 1900X indeed (August 10 vs. August 31, 2017). But model numbers and specs were announced for the three together at Siggraph, end of July.The first 8 core threadripper 1900X also was not released until much later. (if I remember correctly)
Server it makes sense because the chips fill certain needs like NAS setups where the IO matters more then the CPU, or certain licenses that are charged per core where memory size makes the biggest difference.The single part among the three 8-core SKUs of Epyc 7002 which has 8 CCXs, and thus needs 4 partly functional CCDs, is priced higher than that (yet still fairly reasonably for the more exotic server CPU which it is). And this one may be a low volume part anyway.
If the other two SKUs weren't priced smartly, they would have vanished from the market by now or moved up in price; neither one happened. But it is of course a different market.
As for workstation CPUs, a sizeable overlap WRT core count and computing throughput between desktop CPUs and workstation CPUs has always existed and has never been a problem, except maybe for a casual observer. There is still differentiation by memory capacity, amount of I/O, and reliability/ availability features.