• 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

As said above, we already know the clock is lower. Performance isn't necessarily IPC or clock. It could be the memory controller is much better, or the cache is faster, or the dispatcher is better. Regardless of where the real world gains come from, all that matters is it's there.

It's a shame these APU's don't allow for L3 cache, as I bet that's a good 5% performance loss before they've even started.

For some reason I really don't like the idea of buying a CPU without L3 cache.:confused:
 
http://www.techpowerup.com/195320/amd-a10-7850k-and-a10-7700k-apu-specifications-detailed.html

Nothing we don't already know but lists and specs a new part, the A10-7700k.

8 ROPS and the 128bit DDR3 memory controller is seriously going to hurt what those 512 GCN cores can do, you'll be looking for a fast set of DDR3 2400 and above to feed this chip and get anything good out of the GPU.

Don't see the point in the expense of that ram, unless you already own the ram, (Or you are using a real m-itx htpc case which doesn't feature any expansion slots for pcie cards)
for considerably cheaper you could buy a 7750 gddr5 or equivalent and have considerable better performance than the 512igpu
 
Yeah agreed, their are decent budget options from both Intel and AMD, for me right now the Haswell Pentiums paired with a dedicated GPU like a HD 7750+ is a similar priced more balanced overall system, performs a lil better and uses less power. Current AMD APU's do have their place in small for factor PC's as well though, where space is very limited. I'm hoping Kaveri is a decent step up from a 6800K, the GCN SP's and SR cores might make it a better option than a cheap CPU / dedicated GPU setup. It will prob all come down to pricing.

This was the same argument that was put forward when Llano was around and trinity, a cheaper intel dual core and a dedicated gpu provides a better system on the basis of gaming. An Amd quad provides a cheaper faster system that handles htpc better that can rip using programs like handbrake better than an intel system dualie for the budget.
 
That's a stonking board for the money, teppic. Too big for my liking though, and no wifi :(

Think I'll plump for the Gigabyte F2A88XN-WIFI

I was thinking of that myself, but I'm happier with micro ATX because of the better expansion options (I might want to add a sound card and a discrete graphics card).
 
Don't see the point in the expense of that ram, unless you already own the ram, (Or you are using a real m-itx htpc case which doesn't feature any expansion slots for pcie cards)
for considerably cheaper you could buy a 7750 gddr5 or equivalent and have considerable better performance than the 512igpu

Indeed, but not everyone here cares about cost.
For me, I'll have a great deal of fun building my Kaveri boxes and seeing what I can eke out them.

Tweaking tiny boxes is just as much fun as clocking the nuts off gaming behemoths imo.
 
It's a shame these APU's don't allow for L3 cache, as I bet that's a good 5% performance loss before they've even started.

For some reason I really don't like the idea of buying a CPU without L3 cache.:confused:

Depends on the application and how it uses the cache and memory subsytem,
For example clock for clock Llano was about 3-5% faster than phenom II without an L3 cache.
 
Depends on the application and how it uses the cache and memory subsytem,
For example clock for clock Llano was about 3-5% faster than phenom II without an L3 cache.

Just had a look up on Llano cores.
Lol, wouldn't 8 of them with L3 cache been much better than Bulldozer was :p?

As far as I remember, Athlon II and Llano cores weren't massively difference from each other in end performance, so the L3 cache as far as I'm aware would have made the Phenom II's and Llano's around the same?
 
Last edited:
Just had a look up on Llano cores.
Lol, wouldn't 8 of them with L3 cache been much better than Bulldozer was :p?

As far as I remember, Athlon II and Llano cores weren't massively difference from each other in end performance, so the L3 cache as far as I'm aware would have made the Phenom II's and Llano's around the same?

Llano used 1meg per core l2 cache instead of the 512 in phII/Athlon II and had some tweaks to the architecure/ also had a reworked memory controller with slightly higher bandwidth than PhII/Thuban. With exception to the die harvested fm1 Athlon which is Llano rebadged ie (athlon x4 631 etc) so basically Llano.
It depended on the application but providing the app was accessing l2 cache data then Llano showed gains. Of course in l3 dependent apps and games which were l3 sensitive the phenom II would still win or be equal to.

It's always been the debate as to whether Amd should have used Llano's architectural changes and included an l3 and then gone for one last x6-x8. They had a more mature 32nm process and so aiming for 4ghz on a k10.5 derived x6-x8 could have been achieved. Not to mention that Amd always drastically overvolted their Cpu's my aunts e-3000 ran a 1.41v at 2.4, seriously with k10stat it runs 1.16v fine.
I don't know if a phII x8 it would have scaled so well in multi-threaded applications in comparsion to Zambezi, it would probably have used more power though.
Llano was built to compete on the mobile side with intel so the l3 had to go.
The reason, for power saving. As their l3 cache is exclusive and can't be powergated without serious rework or latency issues.

Intel from Sandybridge to now have been focusing on mobile performance first and then scaling that chip to desktop. Its the same as expecting Kaveri to scale to more modules and be put on am3 or a new am socket, I just don't it's going to happen, not unless they have a miracle with a smaller node process and go for big cores again. Amd's focus really is cpu-gpu compute and the average user and making money. They are squashed in a shrinking market between Arm and Intel.

From my testing back in the day when I had an x4 620 and a x4 b55 I think the athlon had to be clocked 200mhz above a phII to equal it.

K8 - K10 it had to die eventually, it had a good life though,
Anyway back on topic I will be buying a Kaveri to do some testing,
I have a big report lined up for it
 
HSa isn't a buzzword, HSA is very big though and it depends what you mean. Kaveri will have hUMA and HQ working from day one of Kaveri's launch, both can bring around significant performance benefits instantly for a variety of things. It will increase performance of some stuff instantly, many more things very quickly and a hell of a lot of things over time.


Some things will be faster automatically, any program doing loads of copies in memory to the gpu allocated system memory will save time, copies and get on with the work. There are plenty of optimisations where existing software will simply see a speed up from things like unified memory, then there are longer term things like Java becoming HSA compatible and enabling gpu acceleration in just about any Java program. Unified memory and HQ which will work for things now, will be further leveraged by more HSA software in the future.

It's not a buzzword, it will make an immediate impact, the impact will simply grow over time.
 
HSa isn't a buzzword, HSA is very big though and it depends what you mean. Kaveri will have hUMA and HQ working from day one of Kaveri's launch, both can bring around significant performance benefits instantly for a variety of things. It will increase performance of some stuff instantly, many more things very quickly and a hell of a lot of things over time.


Some things will be faster automatically, any program doing loads of copies in memory to the gpu allocated system memory will save time, copies and get on with the work. There are plenty of optimisations where existing software will simply see a speed up from things like unified memory, then there are longer term things like Java becoming HSA compatible and enabling gpu acceleration in just about any Java program. Unified memory and HQ which will work for things now, will be further leveraged by more HSA software in the future.

It's not a buzzword, it will make an immediate impact, the impact will simply grow over time.

This why i asked, whats the point of L3 when the system memory is unified. :)
 
So with rumors saying there will be no desktop FX chips. Does that mean that AMD are just killing that brand name or does that mean no desktop chips at all?

Also if Kaveri APU's are the only product line, will they be mainstream, like 4670K / 4770K is for Intel, and then they have the 'E' chips for high end, whereas AMD will just have mainstream no high end?

It's confusing where the products will stack against each other..

I.e is the A10 APU 7850K being aimed at 4770K or a lower tiered Intel range? Be nice if AMD would comment on the roadmap leaks.
 
The rumour is based on that dodgy slide I linked to earlier and before any of those sites had picked it up. I suspect they must have either been reading HardOCP,Hexus or OcUK forums.

Also,AMD and Intel never really comment on leaks last time I checked.

Anyway,supposedly tommorrow is when we will get some more official info about Kaveri.
 
Kaveri is the mainstream AMD chip - FX will continue as Piledriver through 2014 as their "high end" option (mid-range really), before being killed off and replaced by excavator IF AMD still see a market for their mainstream server parts (FX being server derived). Otherwise, killed off and no more FX.

It seems pretty straightforward, to be honest. It's time to let FX on AM3 go. It's over.
 
Back
Top Bottom