• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Launches Three Kaveri APU SKUs in February 2014 – Feature Set For A10 and A8 APUs Detailed

So AMD would really need to consider a Triple or Quad Channel memory controller come the DD4 Kaveri refresh if they want to unleash it's full potential

These are not performance parts!
They are for entry level systems and not designed to compete with discrete gfx cards.

AMD are not going to throw money into elaborate memory controller designs for the segment of the market these SKU's are designed for.
 
They will however have the highest performing cores for AMD, so could actually perform better with a discrete card in some games than their FX line up.

Which would be funnies :p

See, price wise, depending on how good these cores are, they could be completely and perfectly viable options for mid end gaming with a discrete GPU.
 
See, price wise, depending on how good these cores are, they could be completely and perfectly viable options for mid end gaming with a discrete GPU.

Or using the integrated graphics a great option for a low to mid end gaming rig with the option to upgrade to a discrete GPU at a later date.
 
Or using the integrated graphics a great option for a low to mid end gaming rig with the option to upgrade to a discrete GPU at a later date.
^ This. Though if it's cpu performance is noticeably faster than Pile Driver based FX cpu's and overclocks like Richland apu's then I could see it being picked up by more than the low end gamers.
 
These are not performance parts!
They are for entry level systems and not designed to compete with discrete gfx cards.

AMD are not going to throw money into elaborate memory controller designs for the segment of the market these SKU's are designed for.

That time will come, late 2014 into 2015 :p These APUs pretty much nullify any dGPU under £80 (which is a huge market), next stop <£120 and soon third tier performance dGPUs
 
That time will come, late 2014 into 2015 :p These APUs pretty much nullify any dGPU under £80 (which is a huge market), next stop <£120 and soon third tier performance dGPUs

Ofc it's a huge market 90% of the people who own a PC do nothing more taxing than flash games on them.

This is where APU's are targeted at the entry level.
I've used them since Llano in my HTPC's and cheap builds for friends and family.

I'll replace my current Richland ITX Steambox with Kaveri once it's a available.
Switching on a 100w box to play less demanding titles is more preferable to me than firing up the gaming rig.

Kaveri is a step forward with HSA and HUMA, but it's going to take a few years for mainstream apps (Mantle aside) to catch up and make use of the HSA tech.
It's also going to be bandwidth starved on DDR3.

Carrizo will be more interesting, DDR4, Excavator cores (which are a much bigger jump in IPC than Steamroller, hence no Steamroller FX CPU's) and a beefier iGPU.
 
Supposedly this popped up last month:

http://i.imgur.com/z8cvdm3.png

z8cvdm3.png

Posted by NostaSeronx on Anandtech forums. It seems to be about the mobile models.

AMD is probably using PCMark and 3DMark so the performance boost over Richland is looking like the Llano to Trinity one. Mobile Richland and Trinity A10 CPUs have base clockspeeds of between 2.1GHZ to 2.5GHZ it seems. Some tidbits from shipping information for Kaveri:

http://www.zauba.com/import-kaveri-hs-code.html

It seems the mobile chips are running at 1.8GHZ,so if this is true,IPC must be much better overall than Richland.

It will be interesting to see the performance boost over Trinity.

Battery life does look improved.

Kaveri benchmark(single channel) against Core i5 2500K(dual channel):

http://browser.primatelabs.com/geekbench3/compare/226070?baseline=223722
http://forum.hardware.fr/hfr/Hardwa...ture-steamroller-sujet_924911_80.htm#t8948140

So the 3.5GHZ to 3.7GHZ range seems to be the base clockspeeds for desktop Kaveri.

The slide comes from the Russian arm of AMD:

http://www.ospcon.ru/files/media/Perminov.pdf
https://docs.google.com/file/d/0B0b9jlx_0kLjclFqLTRTWG9ZX00/edit

All the links came from SA forum members.
 
Last edited:
Last edited:
i think the 4/8 is more about the threads than the cores, from earlier in the article..

"Coming to the architecture details, the AMD Kaveri APU features a full blown Steamroller core architecture with upto four multi-threaded x86 cores which are specifically designed to expand computation efficiency, improve the single core execution and feed the cores faster. We are looking at Quad Core APUs in the Kaveri generation which would be multi-threaded and upto 4 MB of L2 shared cache. Steamroller is AMD’s first big core in some time that would feature the 28nm core process and would essentially mean lower power consumption while pumping out a better IPC rate over Piledriver and Bulldozer."

either way im looking forward to seeing them & 14th of january really not that long! :)
 
It says 12 compute units,so is it 4 cores or 4 modules on the CPU side I wonder??

Compute units are for the GPU side of things 64 stream cores in 1 GCN CU (Compute Unit)
8x64 gives us the 512 cores as listed on the spec sheet.

I'd suspect the 12 compute units is a typo, pretty sure these will launch with a 2 module setup for the A10's.
2 cores in each AMD module.
If they've found a way to introduce hyper-threading into the modules fair enough, not heard or read this before.
 
Last edited:
Compute units are for the GPU side of things 64 stream cores in 1 GCN CU (Compute Unit)
8x64 gives us the 512 cores as listed on the spec sheet.

I'd suspect the 12 compute units is a typo, pretty sure these will launch with a 2 module setup for the A10's.
2 cores in each AMD module.
If they've found a way to introduce hyper-threading into the modules fair enough, not heard or read this before.

Its says heteregenous compute units,so they take the cores/module core it seems and are adding it to the CU count.
 
Back
Top Bottom