The Kaveri Pre-Launch Thread (A10-7800 and A10-6800k @3,5 Ghz)

[inf64]

Djohny updated his post with iGPU+cpu compute scores:

 

[ShintaiDK]

Nothing new in terms of VLIW4 to GCN.

 

[inf64]

Well who cares why it scores so much, the point is it does outperform previous chip by big amount within same power envelope. Similar goes for x86 benchmarks on previous page and CPU limited games(720p).

 

[DaZeeMan]

R.E. this link:
http://foro.noticias3d.com/vbulletin/showthread.php?t=421821

The one thing we all need to keep in mind with his 4.55 GHz overclock is that he's not using the integrated GPU (reviewer was having driver issues, so he used a dGPU). So we don't really know how much the GPU is able to overclock yet. It could hold the maximum stable overclock back a bit when active, so of course keep that in mind when looking at the numbers.

So while I am hoping to see similar overclocks with the GPU enabled, I'm not getting my hopes up yet.

 

[monstercameron]

R.E. this link:
http://foro.noticias3d.com/vbulletin/showthread.php?t=421821

The one thing we all need to keep in mind with his 4.55 GHz overclock is that he's not using the integrated GPU (reviewer was having driver issues, so he used a dGPU). So we don't really know how much the GPU is able to overclock yet. It could hold the maximum stable overclock back a bit when active, so of course keep that in mind when looking at the numbers.

So while I am hoping to see similar overclocks with the GPU enabled, I'm not getting my hopes up yet.

flank3r and co. have mentioned on the igp 900+mhz OCs and estimated 1GHz stable is possible.

 

[inf64]

R.E. this link:
http://foro.noticias3d.com/vbulletin/showthread.php?t=421821

The one thing we all need to keep in mind with his 4.55 GHz overclock is that he's not using the integrated GPU (reviewer was having driver issues, so he used a dGPU). So we don't really know how much the GPU is able to overclock yet. It could hold the maximum stable overclock back a bit when active, so of course keep that in mind when looking at the numbers.

So while I am hoping to see similar overclocks with the GPU enabled, I'm not getting my hopes up yet.

He tested the performance of Kaveri Vs Trinity with discrete GPU (7950 OCed) in 720p and 1080p...

 

[monstercameron]

I wonder why are they breaking NDA or are they not bound by it?

 

[inf64]

I wonder why are they breaking NDA or are they not bound by it?

That forum users has no NDA signed, he just bought the chip from some retail store.

New AIDA gpgpu benchmarks:

 

[monstercameron]

darn I asked a manager at my local tigerdirect all he said is that stuff like that just poof,show up in their inventory...

 

[Erenhardt]

This is interesting:
Stock:

+31% GPU memory read
OC:

+37% GPU memory read
Quite a gain that will definitely help in games.

 

[SPBHM]

my GPU memory write and read is a little over half of that with discrete graphics, but copy is like 4x, r/w is basically limited by PCIE or there is something else going on?

still, pretty nice gains on opencl, and if it can go as high as 1GHz, it's higher than what many 7750s can do but, I think it's going to be really hard to translate all this additional GPU power to real world gaming with 128bit DDR3.

 

[monstercameron]

this is like a never ending seesaw with AMD, sometimes it looks excellent and others times it can be disappointing...
oh AMD stahp playing with my emotions!

in any case a 30% bandwidth uplift, with a more powerful gpu should yield fantastic results...shouldn't it be better than a 7750 ddr3?
they usually have ddr3-1600 http://www.powercolor.com/global/products_features.asp?id=426
or am i missing something?

 

[SPBHM]

this is like a never ending see-saw with AMD, sometimes it looks excellent and others times it can be disappointing...
oh AMD stahp playing with my emotions!

in any case a 30% bandwidth uplift, with a more powerful gpu should yield fantastic results...shouldn't it be better than a 7750 ddr3?

most 7750 DDR3 uses 1600 memory, so yes, it should be better (as long as you use faster ram, theoretically you can use this with 64bit 1066 ram or something lol)

 

[DaZeeMan]

This thread might be easier to read...
http://www.overclock.net/t/1404574/steamroller/2770

From Post 2767 in that thread (page 277)

Finally i have drivers. I used a 5800K because is the same frequency than kaveri to compare ipc performance.

Im benching gpgpu now, stay tuned

Awesome opencl performance.

From Post 2770 in that thread (page 277)

Oc in igp is 900mhz in bios + 6bclk. Gpuz cant apply that changes so the real frequency is near 1ghz. Fully stable.

He shared some Open CL benches on page 278. Looks like the same as above.

 

[inf64]

With 2133+ memory it should likely be faster than 7750 DDR3 version. Looks like drivers are still not officially finalized (reviewers must have some sort of working driver though). Probably on the 14th AMD will officially release the Kaveri supporting catalyst.

@ DaZeeMan

Previous page .

 

[monstercameron]

maybe 14.1 will drop with mantle...I wager bf4 mantle and/or the strardock[?] demo drops the same day.

 

[frozentundra123456]

most 7750 DDR3 uses 1600 memory, so yes, it should be better (as long as you use faster ram, theoretically you can use this with 64bit 1066 ram or something lol)

Clockspeed is lower on kaveri as well, isn't it? In any case, the real comparison is to 7750 GDDR 5. I don't know why anyone who is interested in gaming would use the DDR 3 version.

 

[SPBHM]

Clockspeed is lower on kaveri as well, isn't it? In any case, the real comparison is to 7750 GDDR 5. I don't know why anyone who is interested in gaming would use the DDR 3 version.

yes 800 vs 720, but with 1600 ram I think the impact from the GPU clock wouldn't be that high on average, and faster memory should win...

but I completely agree, 7750 reference model used GDDR5 as the vast majority of 7750s sold over the years, and the DDR3 model is simply a very poor choice, probably only made by people with no clue about what they are buying.

 

[monstercameron]

yes 800 vs 720, but with 1600 ram I think the impact from the GPU clock wouldn't be that high on average, and faster memory should win...

but I completely agree, 7750 reference model used GDDR5 as the vast majority of 7750s sold over the years, and the DDR3 model is simply a very poor choice, probably only made by people with no clue about what they are buying.

think about the OCed clkspd, 1GHz with ddr3-2133-2666, that could be a huge uplift over a 7750-ddr3 sku.

 

[DaZeeMan]

I'm still hoping to see some DDR3 2400+ CL9/10 Benches at some point, once the NDA's are lifted. Looks like djohny24's DDR3 timings are 2544 @ 11-13-13-32.

I'm quite happy to see him sharing his benches though!

 

[parvadomus]

That's some nice performance gain on gaming!

 

[PPB]

Looking good so far. If HSA adoption rate on production 3D renderers is good, the era of having a big fat CPU for rendering can finally be over

EDIT: For non spanish speakers, the bencher said he saw really good gains on minimum framerates. Considering this is key for a good gaming experience, seems like AMD nailed at last a good gaming CPU for most people.

 

[Shivansps]

With 2133+ memory it should likely be faster than 7750 DDR3 version. Looks like drivers are still not officially finalized (reviewers must have some sort of working driver though). Probably on the 14th AMD will officially release the Kaveri supporting catalyst.

@ DaZeeMan

Previous page .

Dont forget that the cpu also uses it.

 

[mikk]

purely workload dependant clock for click piledrive wins some K10 wins some.


edit:lol wrong cpu.....
http://www.anandtech.com/bench/product/203?vs=699

...and wrong link. They have a turbo, hence why meaningless for IPC comparisons.

 

[SPBHM]

...and wrong link. They have a turbo, hence why meaningless for IPC comparisons.

it's a shame they didn't test using more software, but here a comparison at the same clock
http://www.hardwarecanucks.com/foru...15-amd-vishera-fx-6300-fx-4300-review-15.html

I would like to see llano vs trinity vs kaveri at same clock (CPU, RAM AND IGP) just for fun