I was comparing Carrizo vs the older AMD cpus, not Intel.
really? sweet.
Quote: Carrizo will be 20nm
really? sweet.
Will AMD Carrizo offer a considerable performance increase over Kaveri?
|
09-18-2014, 04:18 AM
(This post was last modified: 09-18-2014, 04:26 AM by omega_rugal.)
I was comparing Carrizo vs the older AMD cpus, not Intel.
Quote: Carrizo will be 20nm really? sweet. 09-18-2014, 01:03 PM
Oh. Well that's kind of confusing then since you did say:
omega_rugal Wrote:unlocked and cheaper than intel`s counterparts,and: omega_rugal Wrote:the user can offset the lower IPC per core with higher clock speed if he needs to... Which don't really make any sense in that context. It's going to be faster than kaveri. That's for sure. Having a new architecture on a new lithography be slower than its predecessor requires multiple royal fuckups at every level of a company. I have no doubt that both clock rates and IPC will be boosted (or maybe just IPC) slightly. But it will likely only be as fast as a sandy bridge era cpu from Intel running at stock speeds. Most of the improvements will likely be on the GPU side. We've seen a clear pattern with llano, trinity/richland, and kabini/kaveri of increasing the portion of the power and die size budget that the IGP gets with each generation. Everything that they've said so far seems to point in the direction of continuing that trend. They will likely continue to have the high end models unlocked and the low end models locked.
"Normally if given a choice between doing something and nothing, I’d choose to do nothing. But I would do something if it helps someone else do nothing. I’d work all night if it meant nothing got done."
-Ron Swanson "I shall be a good politician, even if it kills me. Or if it kills anyone else for that matter. " -Mark Antony 09-18-2014, 01:06 PM
I wish Nvidia had a x86 license. :/ AMD's igpus are always slower than even low end Dedicated GPUs. And they try to market them to gamers.
09-18-2014, 02:04 PM
Nvidia doesn't have the capability or the want to develop x86 cpus even if they did have a license for it. Nobody it stupid enough at this point to invest tens of billions of dollars to develop cpus that are doomed to fail against Intel because of their massive R&D advantage. The only reason AMD can do it is because they've been in this industry for decades now and already have a lot of data from R&D. But a complete startup attempt against Intel in the x86 market would get crushed in a heartbeat. Even if it had a decent IGP nobody would buy an x86 cpu from nvidia if it had the IPC of a pentium 4. Besides the same limitations that apply to AMD IGPs would apply to nvidia is well. Low die size, power budget, and memory bandwidth. Nvidia can't overcome those things just because they're nvidia. So I really doubt an nvidia on die IGP would be any better.
"Normally if given a choice between doing something and nothing, I’d choose to do nothing. But I would do something if it helps someone else do nothing. I’d work all night if it meant nothing got done."
-Ron Swanson "I shall be a good politician, even if it kills me. Or if it kills anyone else for that matter. " -Mark Antony 09-19-2014, 05:49 AM
(09-18-2014, 02:04 PM)NaturalViolence Wrote: Nvidia doesn't have the capability or the want to develop x86 cpus even if they did have a license for it. Nobody it stupid enough at this point to invest tens of billions of dollars to develop cpus that are doomed to fail against Intel because of their massive R&D advantage. The only reason AMD can do it is because they've been in this industry for decades now and already have a lot of data from R&D. But a complete startup attempt against Intel in the x86 market would get crushed in a heartbeat. Even if it had a decent IGP nobody would buy an x86 cpu from nvidia if it had the IPC of a pentium 4. Besides the same limitations that apply to AMD IGPs would apply to nvidia is well. Low die size, power budget, and memory bandwidth. Nvidia can't overcome those things just because they're nvidia. So I really doubt an nvidia on die IGP would be any better. An x86 license doesn't mean you have to make x86 CPUs. You could make ARM CPUs compatible with x86 code. Like Nvidia wanted to do before but Natively instead of Emulated. 09-19-2014, 07:50 AM
Isn't the definition of an x86 cpu a cpu that can execute x86 code natively? I don't know of any cpus that can run multiple ISAs at the physical level.
"Normally if given a choice between doing something and nothing, I’d choose to do nothing. But I would do something if it helps someone else do nothing. I’d work all night if it meant nothing got done."
-Ron Swanson "I shall be a good politician, even if it kills me. Or if it kills anyone else for that matter. " -Mark Antony 09-19-2014, 08:23 AM
09-19-2014, 09:09 AM
So you're saying that nvidia should acquire an x86 license and make an arm cpu that can run both arm and x86 ISAs? That sounds like a really bad idea on so many levels. Extreme complexity and design challenges, lack of a market, the fact that nobody that has tried to do it has succeeded, performance on both ISAs would be hurt, x86 IPC wouldn't be very good to begin with due to lack of R&D, etc.
"Normally if given a choice between doing something and nothing, I’d choose to do nothing. But I would do something if it helps someone else do nothing. I’d work all night if it meant nothing got done."
-Ron Swanson "I shall be a good politician, even if it kills me. Or if it kills anyone else for that matter. " -Mark Antony 09-19-2014, 09:51 AM
If anyone would want to do it, it would be nVidia, as their Shield devices currently only natively support a very small number of games that have been ported specially. Having something just fast enough at x86 that some non-CPU-centric, or just older games can run without needing to be ported would be great for them. They could also get help in the form of not-quite-state-of-the-art-R+D-data from AMD/Intel by paying for it, as they have money. To a point it doesn't even have to be too complicated if they translate x86 into bytecode in a similar way to ARM to bytecode, as each ALU doesn't have to know it's doing anything weird, just that it's throwing two integers at each other and seeing what happens.
NOTE: Some of this might not be particularly applicable to real life, as companies generally do things to make profit, not because it would be cool.
OS: Windows 10 64 bit Professional
CPU: AMD Ryzen 5900X RAM: 48GB GPU: Radeon 7800 XT 09-20-2014, 02:01 AM
A much better solution would be for them to just use an existing x86 cpu from Intel or AMD. This would be much cheaper and more effective.
"Normally if given a choice between doing something and nothing, I’d choose to do nothing. But I would do something if it helps someone else do nothing. I’d work all night if it meant nothing got done."
-Ron Swanson "I shall be a good politician, even if it kills me. Or if it kills anyone else for that matter. " -Mark Antony |
« Next Oldest | Next Newest »
|