• 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

I want to see how the Kaveri core performs against Lynfield / Bloomfield so i turned of HT and clocked it at 3.7Ghz, i can put them up here if anyone wants to see the results now, if not i will wait for the review.
 
humbug said:
I want to see how the Kaveri core performs against Lynfield / Bloomfield so i turned of HT and clocked it at 3.7Ghz, i can put them up here if anyone wants to see the results now, if not i will wait for the review.
Click to expand...

Depending on the situation it's either near the same or lower than anyone should want it to be at a rough guess?

As the clock for clock improvements are inconsistent.
 
Martini1991 said:
Depending on the situation it's either near the same or lower than anyone should want it to be at a rough guess?

As the clock for clock improvements are inconsistent.
Click to expand...


We don't know what the performance is yet. the stock performance results for CineBench R15 we have in this thread so far range from 302 to 333 and anything in-between.

And half of them claiming the stock clock as 3.5Ghz with the other half 3.7Ghz.

Its the usually nonsense, all of them claim to have one while none of them do.
 
stulid said:
Wonder what Vcore they used for 4.5Ghz as mines not 100% stable in certain tests and im not wanting to go too stupid.
Click to expand...

What board are you using? 4.4-4.45 seems the norm with sensible volts (sensible being up to 1.45v) from the leaks I've seen so far.

Crank the iGPU up to 1100+, that's where the best gains are with smoking fast DDR3 2400+ with tight timings.
Are you using the 13.25 drivers?
I wouldn't be too fussed where the CPU tops out tbh, I doubt that chip at 4.559 on the H80 is representative of the average clocks of these chips and is on the upper side.
I think he got lucky there and the Asus A88XM-PLUS isn't anything special as boards go, been running one for a good while now in my HTPC, 8 sata 3 ports is awesome.

Best board for clocking these is the Gigabyte GA-F2A88X-UP4, why the hell they can't make boards that beefy in m-ATX and m-ITX is beyond me.
I hope to god we see some beefy m-ATX and m-ITX (RoG ITX please Asus) boards soon and not the pitiful ones available atm.

I'd also like to see more boards with display port, one Asus board has it and and the Gigabyte UP4, that's it lol.
 
Last edited:
Legion said:
What board are you using? 4.4-4.45 seems the norm with sensible volts from the leaks I've seen so far.

Crank the iGPU up to 1150+, that's where the best gains are with smoking fast DDR3 2400+ with tight timings.
I wouldn't be too fussed where the CPU tops out.
Click to expand...

A Sniper, but I increased the igp too 1000Mhz and the scores were not effected or went down lol.
 
If those benchmarks are remotely accurate, the openCL speed up is down right ridiculous, the cache improvements in copy speeds are amazing also(obviously not particularly interesting to end user).

I don't know enough about the openCL tests, are they using the gpu anyway, only, or using cpu/gpu and a huge portion of this increase is from HSA/memory copy latency being eliminated.

This is the kind of thing I hopefully want to see answered, I don't know if you'll be able to turn HSA off, to compare with/without. It's going to be difficult I would guess to pick out specifically where improvements are coming from with a better compute gpu, better cpu cores, better cache, unified memory. It will be hard to point to a particular benchmark and go "unified memory is clearly making that faster" because it will make say part of the difference to say a compute benchmark, but with a much faster gpu in compute... how can you quantify how much of the increases is the gpu, how much HQ/hUMA.

It's a heck of a lot of interesting tech. HSA should be compatible straight off with Windows(as 8 was always implemented to work with HSA), and the hardware as always leads (most) software. So it's an important step to getting dev's to push for HSA coding, and many already are but much of the software is due in the next 12-24 months.

In a couple years APU's won't be seen as budget or gaming parts, there will simply be plenty of software where a quad core + compute gpu will smash a hex/octo core to pieces. ARM, AMD, Apple, Qualcomm, Samsung all want every day software to be accelerated as much as possible, as soon as possible.
 
I think It's Not exactly But you see where i'm coming from. A game that stutters about at 3fps is almost unplayable. 30fps is very much so playable. It's just 60+ is better. It's my opinion.
 
stulid said:
A Sniper, but I increased the igp too 1000Mhz and the scores were not effected or went down lol.
Click to expand...

Default for these is 720mhz? could be there's not that much headroom on the GCN cores.
After looking around some more, I think it's likely that ~900 is going to be more like the norm.

There'll be some protection / throttling kicking in at 1000 if the iGPU cant handle it.

Try it with Easytune6 and take it up slow, save a load of time on reboots.
 
Last edited:
Legion said:
Default for these is 720mhz? could be there's not that much headroom on the GCN cores.
After looking around some more, I think it's likely that ~900 is going to be more like the norm.

There'll be some protection / throttling kicking in at 1000 if the iGPU cant handle it.

Try it with Easytune6 and take it up slow, save a load of time on reboots.
Click to expand...

GCN Cores = HD 7### / R7/9 2##, even a bad one will do 1150Mhz a golden one 1350 though thats very rare.

The GPU architecture can do a lot more than 900. if anything its Motherboard and or BIOS limitations.
 
You must log in or register to reply here.
Back
Top Bottom