Blazkowicz
Legend
So that means locking away 17.5% users (75% multiplied by 30%), just because. That's silly, those Vista and 7 32bit work fine.
Haswell vs Kaveri
- Thread starter AnarchX
- Start date
Somehow I'd missed this:
http://videocardz.com/48716/amd-mantle-true-audio-support-coming-later-month
It doesn't seem very likely to ever show up in anything other than a demo, but it could be pretty cool.
http://videocardz.com/48716/amd-mantle-true-audio-support-coming-later-month
It doesn't seem very likely to ever show up in anything other than a demo, but it could be pretty cool.
Note the date.
That's pretty much the last slide I could find that mentions extending it to discrete.
The ISCC and the Fusion Development Summit roadmaps that followed shortly after omit that reference, and I haven't seen it mentioned going on two years.
Somehow I'd missed this:
http://videocardz.com/48716/amd-mantle-true-audio-support-coming-later-month
It doesn't seem very likely to ever show up in anything other than a demo, but it could be pretty cool.
It's nice to know AMD are still thinking of it as a viable option but yeah, I can't see any developer ever going to the trouble to implement this. Perhaps 3 years from now when HSA capable GPU's make up the majority of the gaming market (i.e. Intel are also in on the game) and there's some kind of standard than can be coded to we'll see this. But even then that's pretty optimistic.
They bothered to put GPU PhysX in titles too, why not this?
I need to check with AMD at some point and see what the plan is there. Certainly the last time I talked to them, the plan was to have coherency for dGPUs over PCIe, in order to give those products the ability to run monolithic-style HSA applications.
Rurouni
Veteran
If Kaveri 3D performance level is acceptable, what is the equivalent if I want to go with Intel for the same price with similar 3D performance? A combo of CPU + GPU is okay as long as it's around $200 (preferably less).
I think I want to update my PC and I just need enough grunt to run games @720p30 + decoding anime mkvs with those damn animated subtitle (that my Athlon II X3 struggles!). Is a single core Kaveri performance better than a single core Athlon II at the same clock? I assume it is, but I just want to make sure.
I think I want to update my PC and I just need enough grunt to run games @720p30 + decoding anime mkvs with those damn animated subtitle (that my Athlon II X3 struggles!). Is a single core Kaveri performance better than a single core Athlon II at the same clock? I assume it is, but I just want to make sure.
Only Iris Pro 5200 is competitive with the higher-end (65W+) Kaveri iGPUs right now. Given that it's not that widely available (it's not socketed so you need to get OEM motherboards/systems) and currently costs more, you'll likely have to look at discrete GPUs. I think there were some suggestions earlier in the thread, but ultimately a pretty inexpensive dual core + a decent ($80-100) GPU will generally outperform Kaveri, albeit with higher power draw.
Rurouni
Veteran
I can get Intel G870 with AMD R7 250 for similar price as the most expensive Kaveri (7850K). Is that a good choice or I should just go with Kaveri? I don't think i3 is an option unless you want to go sub $100 with the GPU (I could get either 6670, 7730, GT630, or push it to GT640).
Just a reminder that this is with the local pricing in my country. What is the minimum GPU for something that's comparable to Kaveri?
Just a reminder that this is with the local pricing in my country. What is the minimum GPU for something that's comparable to Kaveri?
Hmm, the R7 250 is going to fall somewhere in the same range as Kaveri... probably slightly faster but it's hard to say for sure. I haven't had time to go over the benchmarks in detail so perhaps someone else could chime in on that particular comparison.
FWIW there's a pretty huge jump in performance (~double) if you can afford a R7 260 (~$20 USD more MSRP). That one will wipe the floor with Kaveri for a pretty small $ delta.
FWIW there's a pretty huge jump in performance (~double) if you can afford a R7 260 (~$20 USD more MSRP). That one will wipe the floor with Kaveri for a pretty small $ delta.
UniversalTruth
Veteran
So, apparently the i-cache size scales with the associativity, although in an "odd" manner, which means the bank structure is unchanged
Why is the voltage on your image so low at so high frequency, just 0.528 V at 3.9 GHz, while the image below reveals something different, 0.792 V at 1.7 GHz?
And another question- why are the L1 Data and L1 Inst. caches only 4-way and 3-way while on Intel processors they are 8-way?
Does it introuduce some type of advantage?
Where are the skylake disclosures?
I'd love to hear more about that!
Physx runs on any Nvidia GPU which is a big percentage of the target market. This would specifically require a Kaveri APU which is only a tiny percentage of the target market and if it also needs to be in combination with an AMD dGPU that percentage gets even smaller again. It would of course be a very different matter if this worked with all iGPUs from say IVB/Llano upwards and in combination with both AMD and Nvidia GPU'S. That would finally give me a reason to upgrade the old Sandybridge.
As an user of a KVR based rig (bought the CPU on day one) I would like to point out that the current state of affairs is rather disappointing. The monitor going into torpor requiring a power cycle to awaken is not as fun as it seems initially (note that things work fine sans the 7850k, so the probability of it being the culprit is high).
I was speaking to the multitude of crashes in seemingly every game they tested. Especially hearing about OS corruption as one of the symptoms...
My apologies, my brain said Skylake when it should have been Broadwell.
While it is possible to stay coherent across the PCIe, to some extent, it is simply not possible to "run monolithic-style HSA applications" on a PC dGPU, due to PCIe latency.
PCIe bandwidth is fine, and GPU video can hide modest latency, but compute requires low latency, and dGPU compute on common memory will not have that. Tight HSA-style computation means using whichever is most appropriate: CPU or GPU. That means moving computations repeatedly back and forth between CPU and GPU, and doing that with overall practicality depends on low movement overhead. No conventional dGPU video card can possibly support that, because of PCIe.
http://www.xbitlabs.com/news/cpu/di...th_Chip_with_Sixteen_Cores_on_Single_Die.html
xbitlabs discovered an 8 module / 16 cores cpu only steamroller
xbitlabs discovered an 8 module / 16 cores cpu only steamroller