• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

AMD "Kaveri" Desktop APU Launch Date Revealed

btarunr

Editor & Senior Moderator
Staff member
Joined
Oct 9, 2007
Messages
47,668 (7.43/day)
Location
Dublin, Ireland
System Name RBMK-1000
Processor AMD Ryzen 7 5700G
Motherboard Gigabyte B550 AORUS Elite V2
Cooling DeepCool Gammax L240 V2
Memory 2x 16GB DDR4-3200
Video Card(s) Galax RTX 4070 Ti EX
Storage Samsung 990 1TB
Display(s) BenQ 1440p 60 Hz 27-inch
Case Corsair Carbide 100R
Audio Device(s) ASUS SupremeFX S1220A
Power Supply Cooler Master MWE Gold 650W
Mouse ASUS ROG Strix Impact
Keyboard Gamdias Hermes E2
Software Windows 11 Pro
A leaked AMD slide detailed key dates in the product launch cycle for the company's next-generation desktop APU, codenamed "Kaveri." The chip will be officially announced on December 5, 2013. This is when AMD will post press-releases about the chip, but lack of market availability will make it more of a soft-launch. The chips should begin to ship to OEM vendors by late-December, 2013. It won't be until mid-February, 2014, that you'll be able to buy one of these chips in the retail channel. Built in the new socket FM2+ package, supporting the company's new A88X chipset, "Kaveri" will feature CPU cores based on the next-generation "Steamroller" micro-architecture, which offers an incremental (conservatively, around 10 percent) performance upgrade over "Piledriver;" a new GPU based on the Graphics CoreNext micro-architecture, TrueAudio technology, support for faster DDR3 memory standards, and PCI-Express gen 3.0 bus interface.



View at TechPowerUp Main Site
 
Last edited:
Awaiting reviews in February!

Dedicated audio core processor. More amd true audio?
 
Last edited:
"New SteamrollerB core" ? Is this the name for the core without the L3 cache or a new, improved version of SR?
 
Rumors stated here say the highest end kaveri brings 832 GCN stream processors.
I don't really believe that they could feed all of those with just the dual channel DDR3 interface that is supported on FM2+ motherboards, since that's even more stream cores than the HD 7770 has.

I wonder what trick they'll have up their sleeves. Stacked GDDR5? On chip cache? They surely have the tech for that last option, judging by the Xbox1 chip design(GCN core count is also very close to XB1). We'll see soon enough I suppose.
 
"which offers an incremental (conservatively, around 10 percent) performance upgrade over "Piledriver;""

Nothing to get excited about then.
 
"which offers an incremental (conservatively, around 10 percent) performance upgrade over "Piledriver;""

Nothing to get excited about then.

I suspect thats IPC, which would scale nicely to clock speeds. I think this is more about how powerful the iGPU will be and how much less die space the CPU cores use as opposed to raw CPU performance. This is an APU afterall.

From a HPC standpoint, shared memory pointers between CPU and GPU is pretty awesome. It can enable tighter coupling of GPU and CPU components at both the hardware and software levels.
 
Because CPU manufactures are focusing on low power designs for mobile, we won't be having many things to get exited about in the near future. :(
 
I wonder what trick they'll have up their sleeves. Stacked GDDR5? On chip cache? They surely have the tech for that last option, judging by the Xbox1 chip design(GCN core count is also very close to XB1). We'll see soon enough I suppose.

Probably just like Xbone,they might using higly superscalar eDRAM.There might be side effect though,branch prediction will be mostly excessing resources feeding data to Decoder.Unless AMD developed "new and faster" L1 cache and separate data path to Floating Point and Integer,we will not see AMD adopting faster DDR soon.
 
Probably just like Xbone,they might using higly superscalar eDRAM.There might be side effect though,branch prediction will be mostly excessing resources feeding data to Decoder.Unless AMD developed "new and faster" L1 cache and separate data path to Floating Point and Integer,we will not see AMD adopting faster DDR soon.

with that many shaders, and enough eDRAM, if true, this is probably gonna be a massive chip:eek:
 
Last edited:
Brings them back to the IPC they had with Phenom II before all this Bulldozer nonsense started.

No..Piledriver pulled them back to Phenom II level...10% improvement for Steamroller is very disappointing if true, there was talk of a near 30% IPC improvement.

Looking more and more like they weren't able to manage it...:cry:
 
10/12 core AM3+ FX please :respect:
 
10/12 core AM3+ FX please :respect:

I was really hoping for that aswell when I decided to pass on piledriver so I could buy that successor chip.

No leaks whatsoever regarding a potential AM3+ refresh though, and I've personally got the feeling that its just not gonna happen. For one, they seem to have switched to another way of power management in both new GPU's and APU, but the AM3+ socket doesn't support that.
 
with that many shaders, and enough eDRAM, if true, this is probably gonna be a massive chip:eek:

GPU wise yes,but not CPU side.Technically AMD can put eDRAM to role as Global L2 shared and memory controller HUB,thus enabling dual DDR5 for desktop just like PS4.But in doing so you need very-very fast L1 cache and another data path for Floating Point and Integer modules while maintain full X86 instruction set at Thread Fetch.All AMD need now is affordable DDR5 :D
 
If they could or did improve their latency issues and caching logic to take advantage of a single DDR5 chip on the board it would do wonders for the overall performance.
 
LOL, time and time again people are totally besides the point of what AMD is trying to do with APUs... Kinda sad really.

Don't get me wrong, I'm skeptical as f**k, but that doesn't mean that I get things mixed up. There's only so much that multi-core and x86 in general can do to performance... that's why we need a damn computation paradigm shift so bad...
 
"which offers an incremental (conservatively, around 10 percent) performance upgrade over "Piledriver;""

Nothing to get excited about then.

That 10% figure is NOT what AMD has said. I can gurantee you single thread performance will increase by at least 15% and per core multithread performance will exceed 20%. But with only 4 cores overaall multi-thread performance will actually be lower than pildedriver 8 core cpus. The saving grace for this chip is the implementation of HSA. When software developers widely adopt HSA (about 2 years) this chip will acclerate performance from 50 to 200% for software programs. That is because HSA will allow your cpu to utilize fast gddr5 memory that is unused in your graphics cards. Sio if you have anything fom a 7850 to a 7970 you will get a big performance boost.
 
That 10% figure is NOT what AMD has said. I can gurantee you single thread performance will increase by at least 15% and per core multithread performance will exceed 20%.

sauceplz.
 
OMFG launch it already!


Also, 3423423-cores CPUs are dead, get over it. As much as I like performance that I can "see" (moar coarz/moar gigglehurtz!!!), HSA has the chance to be something truly revolutionary. Sadly tho, 8 out of 10 people I come across, real life or internet, have no clue what exactly that entails to begin with.

Truth is, x86 is getting pathetic (unless they switch manufacturing materials to something *cough* new... *cough* paradigm shifting stuff *cough*), it has no real long-term future...
 
Back
Top