Upcoming AMD CPU

miagisan

I am about to pull the trigger on a AMD HD7950, and i still have am using a AMD Phenom II x4 965 BE. I can overclock it easily to 3.8-4ghz...so i am in no rush to upgrade it. But i am looking to the upcoming AMD chips to see if they entice me at all. Not looking at the Kaveri APUs since I am gonna have a kickass vid card, but, I cant seem to find any info on their upcoming "FX" like chips (rumor is the FX is being discontinued). 

 

Any info on what the upcoming non APU chips are looking like for 2014?

Ridelynn

There probably isn't one.

I think AMD is going all APU in the consumer space. Which probably isn't a bad idea - Intel pretty much did, and AMD has much strong GPU than Intel, so they can actually compete favorably when you look at an all-in-one package.

miagisan

True...and i was afraid you would say that i been an AMD cpu user since my last intel Pentium 2 (blech!).

 

Will their upcoming APUs perhaps crossfire with the 7900 series? as of right now theb est you could crossfire with an APU is 6700 series. 

miagisan

I just hope AMD gets back into the desktop gaming chips again, I love their pricepoints and performance ratio. For now i'll stick with my Phenom, until something comes out along the lines of the FX series (would love an updated phenom series). With 8 core processors now becoming norm for consoles, games are going to start heavily gearing to those types of processors, and i feel we are going to need to move over as well at some point.

Ridelynn

---

Originally posted by miagisan
True...and i was afraid you would say that i been an AMD cpu user since my last intel Pentium 2 (blech!). Will their upcoming APUs perhaps crossfire with the 7900 series? as of right now theb est you could crossfire with an APU is 6700 series. 

---

There is talk of using the iGPU as some sort of OpenCL accelerator, which would be like a limited version of Crossfire

lafaiel

 The steamroller cpu's are supposed to be the next step in their desktop cpu's for the non apu segment.  I'm pretty much like you, been using amd since my pentium II as well lol, run my 955 at 3.8 and it works great for now.

  I'm personally working on  an FX 8350 build and going to overclock it, but waiting on water cooling parts to come in yet.

zevian

I use the 7950 with the same processor as you,   Im actually wanting to upgrade the processor, In some of my games im peaking it out, and its not the video card.    

 

Is it worth it to look at one of their new am3 chips?

 

miagisan

Originally posted by lafaiel

 The steamroller cpu's are supposed to be the next step in their desktop cpu's for the non apu segment.  I'm pretty much like you, been using amd since my pentium II as well lol, run my 955 at 3.8 and it works great for now.

  I'm personally working on  an FX 8350 build and going to overclock it, but waiting on water cooling parts to come in yet.

Any news on their performance yet or is it still hush hush until closer to their release?

Riposte.This

you can't argue with the i5 price and performance. i5 on sale for around 215 usually, out performs the other AMD chips in that price by a long shot.

miagisan

Originally posted by Riposte.This
you can't argue with the i5 price and performance. i5 on sale for around 215 usually, out performs the other AMD chips in that price by a long shot.

THis isnt a amd vs intel thread...I am strictly looking for AMD info. But thanks.

Bookah

I bought an 8350 a few months ago and love it. its AMD's top of the line CPU @ the moment (8 core 4.2 GHz turbo) and is under 200$

Because its under 200$ I was able to get a 90$ water cooler and this baby never gets over 95*F

 

I hope AMD doesn't go all in APU and stop developing there high performance CPU's!

Because I use AMD so I can peace meal upgrade components and not have to buy a new board with every new CPU generation like intel demands (Rip off lol)

In fact everyone i know who builds computers uses a similar strategy.

 

(edit: I cant imagine they would design a new CPU architecture from the ground up and abandon it, if anything they wont redesign again for a while and improve Bulldozer/Vishera like they did Phenom. Remember Phenom wasn't a great chip the first 3 years or so but by the end of the life cycle it was so tuned & refined the Phenom x6 was amazing. This makes even more sense when you consider these modules were designed to run in the APU's @ lower core counts & freq's)

Quizzical

Neither AMD nor Intel makes processors specifically targeted at desktops anymore.  Probably the last such chip was Thuban (Phenom II X6), which AMD released in 2010.  And it doesn't look likely that there will be any more such chips in the foreseeable future, either.

Intel's Sandy Bridge, Ivy Bridge, and Haswell chips, like AMD's Llano, Trinity, and Richland, are first and foremost laptop chips.  See, for example, how the primary focus of Haswell was bringing down idle power consumption--something that does not matter at all in a desktop.  Both AMD and Intel sell desktop versions of the same chips, but that's just taking the laptop chip and clocking and volting it higher.

AMD's FX chips, like Intel's Gulftown, Sandy Bridge-E, and Ivy Bridge-E, are primarily server chips.  The Intel chips are marketed as Xeon E5 series.  The problem here is that AMD's server chips are highly unprofitable, as they're far inferior to Intel's.  And if AMD kept releasing them, the problem would likely only get worse.  AMD right now has both an inferior CPU architecture, and also has to build it on an inferior process node.  While I expect AMD to narrow the gap in CPU architecture with the launch of Kaveri, the process node disadvantage is only going to widen in the near future.

Right now, AMD's chips are on 28 and 32 nm process nodes, while Intel is on 22 nm.  Furthermore, Intel is moving to 14 nm next year.  AMD has to rely on other fabs to build their chips (currently Global Foundries and TSMC, though they could switch to someone else if so inclined), and the other fabs aren't much interested in talking about their upcoming 20 nm process nodes.  Rather, the focus is on how great 16 nm (TSMC) and 14 nm (Global Foundries, IBM, Samsung) are going to be, once they get FinFETs.  For example, Global Foundries offers something like 6 different 28 nm process nodes.  They're only going to offer one 20 nm node, and don't seem convinced that it will be any good.  But they're working on six different 14 nm process nodes already, as they need the FinFETs and EUV lithography that will be available then, but not sooner.

Once everything gets to 14/16 nm, AMD might well not trail so perilously behind Intel in process nodes.  And once Moore's Law dies off, AMD might well have just as good of process nodes available as Intel.  But even the first of those are a ways off, and any server-oriented products that AMD launches in the meantime would probably be massive money-losers.  So AMD might well have canceled them.  And you can't release a desktop version of a server chip that you didn't make.

AMD isn't abandoning the server market entirely.  They are, however, changing their focus to APUs, for server applications that can leverage the GPU.  They've already made Kabini, and will finish Kaveri soon, so releasing server versions of both chips doesn't cost them much.  Kabini could also be a nifty server product for programs that are fine with a zillion low power, low performance cores, but need stronger CPU cores than ARM can provide.

miagisan

Originally posted by Quizzical

Neither AMD nor Intel makes processors specifically targeted at desktops anymore.  Probably the last such chip was Thuban (Phenom II X6), which AMD released in 2010.  And it doesn't look likely that there will be any more such chips in the foreseeable future, either.

Intel's Sandy Bridge, Ivy Bridge, and Haswell chips, like AMD's Llano, Trinity, and Richland, are first and foremost laptop chips.  See, for example, how the primary focus of Haswell was bringing down idle power consumption--something that does not matter at all in a desktop.  Both AMD and Intel sell desktop versions of the same chips, but that's just taking the laptop chip and clocking and volting it higher.

AMD's FX chips, like Intel's Gulftown, Sandy Bridge-E, and Ivy Bridge-E, are primarily server chips.  The Intel chips are marketed as Xeon E5 series.  The problem here is that AMD's server chips are highly unprofitable, as they're far inferior to Intel's.  And if AMD kept releasing them, the problem would likely only get worse.  AMD right now has both an inferior CPU architecture, and also has to build it on an inferior process node.  While I expect AMD to narrow the gap in CPU architecture with the launch of Kaveri, the process node disadvantage is only going to widen in the near future.

Right now, AMD's chips are on 28 and 32 nm process nodes, while Intel is on 22 nm.  Furthermore, Intel is moving to 14 nm next year.  AMD has to rely on other fabs to build their chips (currently Global Foundries and TSMC, though they could switch to someone else if so inclined), and the other fabs aren't much interested in talking about their upcoming 20 nm process nodes.  Rather, the focus is on how great 16 nm (TSMC) and 14 nm (Global Foundries, IBM, Samsung) are going to be, once they get FinFETs.  For example, Global Foundries offers something like 6 different 28 nm process nodes.  They're only going to offer one 20 nm node, and don't seem convinced that it will be any good.  But they're working on six different 14 nm process nodes already, as they need the FinFETs and EUV lithography that will be available then, but not sooner.

Once everything gets to 14/16 nm, AMD might well not trail so perilously behind Intel in process nodes.  And once Moore's Law dies off, AMD might well have just as good of process nodes available as Intel.  But even the first of those are a ways off, and any server-oriented products that AMD launches in the meantime would probably be massive money-losers.  So AMD might well have canceled them.  And you can't release a desktop version of a server chip that you didn't make.

AMD isn't abandoning the server market entirely.  They are, however, changing their focus to APUs, for server applications that can leverage the GPU.  They've already made Kabini, and will finish Kaveri soon, so releasing server versions of both chips doesn't cost them much.  Kabini could also be a nifty server product for programs that are fine with a zillion low power, low performance cores, but need stronger CPU cores than ARM can provide.

so are the APus being released (Kaveri) be significantly more powerful than the current generation APU? Cause using my Phenom 965 atm i see no reason to go to an APU yet.

Ardnut

check out bit tech, they have good reviews on hardware and might help you decide if amd's chips are worth the money:

 

http://www.bit-tech.net/hardware/cpus/

Quizzical

Originally posted by miagisan

Originally posted by Quizzical

Neither AMD nor Intel makes processors specifically targeted at desktops anymore.  Probably the last such chip was Thuban (Phenom II X6), which AMD released in 2010.  And it doesn't look likely that there will be any more such chips in the foreseeable future, either.

Intel's Sandy Bridge, Ivy Bridge, and Haswell chips, like AMD's Llano, Trinity, and Richland, are first and foremost laptop chips.  See, for example, how the primary focus of Haswell was bringing down idle power consumption--something that does not matter at all in a desktop.  Both AMD and Intel sell desktop versions of the same chips, but that's just taking the laptop chip and clocking and volting it higher.

AMD's FX chips, like Intel's Gulftown, Sandy Bridge-E, and Ivy Bridge-E, are primarily server chips.  The Intel chips are marketed as Xeon E5 series.  The problem here is that AMD's server chips are highly unprofitable, as they're far inferior to Intel's.  And if AMD kept releasing them, the problem would likely only get worse.  AMD right now has both an inferior CPU architecture, and also has to build it on an inferior process node.  While I expect AMD to narrow the gap in CPU architecture with the launch of Kaveri, the process node disadvantage is only going to widen in the near future.

Right now, AMD's chips are on 28 and 32 nm process nodes, while Intel is on 22 nm.  Furthermore, Intel is moving to 14 nm next year.  AMD has to rely on other fabs to build their chips (currently Global Foundries and TSMC, though they could switch to someone else if so inclined), and the other fabs aren't much interested in talking about their upcoming 20 nm process nodes.  Rather, the focus is on how great 16 nm (TSMC) and 14 nm (Global Foundries, IBM, Samsung) are going to be, once they get FinFETs.  For example, Global Foundries offers something like 6 different 28 nm process nodes.  They're only going to offer one 20 nm node, and don't seem convinced that it will be any good.  But they're working on six different 14 nm process nodes already, as they need the FinFETs and EUV lithography that will be available then, but not sooner.

Once everything gets to 14/16 nm, AMD might well not trail so perilously behind Intel in process nodes.  And once Moore's Law dies off, AMD might well have just as good of process nodes available as Intel.  But even the first of those are a ways off, and any server-oriented products that AMD launches in the meantime would probably be massive money-losers.  So AMD might well have canceled them.  And you can't release a desktop version of a server chip that you didn't make.

AMD isn't abandoning the server market entirely.  They are, however, changing their focus to APUs, for server applications that can leverage the GPU.  They've already made Kabini, and will finish Kaveri soon, so releasing server versions of both chips doesn't cost them much.  Kabini could also be a nifty server product for programs that are fine with a zillion low power, low performance cores, but need stronger CPU cores than ARM can provide.

so are the APus being released (Kaveri) be significantly more powerful than the current generation APU? Cause using my Phenom 965 atm i see no reason to go to an APU yet.

I'm going to go out on a limb and say that the top bin desktop Kaveri (A10-7800K?) will be roughly competitive on the CPU side with a Core i5-4440, which is a lower clocked Haswell quad core (turbo up to 3.3 GHz).  AMD will narrow the gap with Intel substantially, but not catch up.  Kaveri will launch early next year.

-----

I should probably back up and observe that AMD has announced a new 12/16 core server chip (two dies, so the desktop version will be 6 or 8 cores) for next year with Piledriver cores.  It will probably mean that they replace the current Vishera chip with the tweaked Piledriver cores in Richland.  So that will presumably come to desktops as well, likely early next year.  But that will only mean an extra ~5% performance or so as compared to current generation FX processors.

miagisan

Originally posted by Quizzical

Originally posted by miagisan

Originally posted by Quizzical

Neither AMD nor Intel makes processors specifically targeted at desktops anymore.  Probably the last such chip was Thuban (Phenom II X6), which AMD released in 2010.  And it doesn't look likely that there will be any more such chips in the foreseeable future, either.

Intel's Sandy Bridge, Ivy Bridge, and Haswell chips, like AMD's Llano, Trinity, and Richland, are first and foremost laptop chips.  See, for example, how the primary focus of Haswell was bringing down idle power consumption--something that does not matter at all in a desktop.  Both AMD and Intel sell desktop versions of the same chips, but that's just taking the laptop chip and clocking and volting it higher.

AMD's FX chips, like Intel's Gulftown, Sandy Bridge-E, and Ivy Bridge-E, are primarily server chips.  The Intel chips are marketed as Xeon E5 series.  The problem here is that AMD's server chips are highly unprofitable, as they're far inferior to Intel's.  And if AMD kept releasing them, the problem would likely only get worse.  AMD right now has both an inferior CPU architecture, and also has to build it on an inferior process node.  While I expect AMD to narrow the gap in CPU architecture with the launch of Kaveri, the process node disadvantage is only going to widen in the near future.

Right now, AMD's chips are on 28 and 32 nm process nodes, while Intel is on 22 nm.  Furthermore, Intel is moving to 14 nm next year.  AMD has to rely on other fabs to build their chips (currently Global Foundries and TSMC, though they could switch to someone else if so inclined), and the other fabs aren't much interested in talking about their upcoming 20 nm process nodes.  Rather, the focus is on how great 16 nm (TSMC) and 14 nm (Global Foundries, IBM, Samsung) are going to be, once they get FinFETs.  For example, Global Foundries offers something like 6 different 28 nm process nodes.  They're only going to offer one 20 nm node, and don't seem convinced that it will be any good.  But they're working on six different 14 nm process nodes already, as they need the FinFETs and EUV lithography that will be available then, but not sooner.

Once everything gets to 14/16 nm, AMD might well not trail so perilously behind Intel in process nodes.  And once Moore's Law dies off, AMD might well have just as good of process nodes available as Intel.  But even the first of those are a ways off, and any server-oriented products that AMD launches in the meantime would probably be massive money-losers.  So AMD might well have canceled them.  And you can't release a desktop version of a server chip that you didn't make.

AMD isn't abandoning the server market entirely.  They are, however, changing their focus to APUs, for server applications that can leverage the GPU.  They've already made Kabini, and will finish Kaveri soon, so releasing server versions of both chips doesn't cost them much.  Kabini could also be a nifty server product for programs that are fine with a zillion low power, low performance cores, but need stronger CPU cores than ARM can provide.

so are the APus being released (Kaveri) be significantly more powerful than the current generation APU? Cause using my Phenom 965 atm i see no reason to go to an APU yet.

I'm going to go out on a limb and say that the top bin desktop Kaveri (A10-7800K?) will be roughly competitive on the CPU side with a Core i5-4440, which is a lower clocked Haswell quad core (turbo up to 3.3 GHz).  AMD will narrow the gap with Intel substantially, but not catch up.  Kaveri will launch early next year.

-----

I should probably back up and observe that AMD has announced a new 12/16 core server chip (two dies, so the desktop version will be 6 or 8 cores) for next year with Piledriver cores.  It will probably mean that they replace the current Vishera chip with the tweaked Piledriver cores in Richland.  So that will presumably come to desktops as well, likely early next year.  But that will only mean an extra ~5% performance or so as compared to current generation FX processors.

Thanks Quiz

TheLizardbones

Holy jeebers. I just looked at the prices on those AMD APUs. If their performance is "good enough", you could save hundreds of dollars on a gaming rig.

**

Er, nevermind. My years old 5770 outperforms the 8670D, the integrated graphics in the A10-6800k. They would need to improve significantly to even be "good enough".

Yes, I know the 5770 lacks some DX11 stuff, but raw horsepower is pretty important. I wish it was possible to have the Nvidia 7x0M chips on a microATX board for a micro gaming system.
:-(

Cleffy

Since I will buy a Discrete GPU on any Desktop I buy, it will be a difficult buy until at least 2015 to get a CPU that does not waste half its die on a GPU. However, by that time Hybrid GPU drivers should be significantly better.

Judging by the PS4 and XBox360 chips, the next Desktop oriented AMD CPU will use a single massive L3 Cache with many CPU cores.  No more split upon split design.