Technical Comparison Sony PS4 and Microsoft Xbox
- Thread starter BRiT
- Start date
- Status
If you are switching back and forth quickly between a game and web browser, I do not see how the browser could need more than 5 or so tabs loaded in memory.
Any more than that could be suspended to the hard drive, if you were to load or resume a tab additional to those five a previous tab would be suspended to the hard drive, or you could tell the system to suspend the game to the hard drive while you browse.
(((interference)))
Veteran
It's because it isn't called 'Jaguar' in MSs documentation.
Michellstar
Regular
Why wouldn´t Ms call them jaguar in their docs?
Devs have to target these processors following AMD recomendations, and not some generic X64 model...
Strange
Devs have to target these processors following AMD recomendations, and not some generic X64 model...
Strange
Betanumerical
Veteran
Same reason they didn't use the AMD nomenclature in there docs even know they could have, nobody knows.
liquidboy
Regular
U make me laugh with your constant assertions of XB1 using stock jaguars.. so wrong...
The comparison would be between the EDRAM and non-EDRAM variants of Haswell to get an idea of the benefits the addition of said EDRAM brings. Regardless the point was less about measuring the performance but more about how Intel's flagship APU design is much more similar to XB1 than PS4.
But if we're trying to suss out low latency benefits, we'll have to separate that from bandwidth freeing benefits. If it's possible.
EG for Haswell if it doubles BW, we'd want to see more than 2X performance increase if I understand correctly.
Then again there are tons of other variables, such as if Haswell even has the execution unit headroom for 2X.
I think the Haswell NDA ends June 3 (dont quote me but I think) so that's good, benchmarks soon.
As we know Intel is super smart, so I'm sure they didn't do this lightly.
And expletive I agree to a point, but the fact is IGP's were running into a severe BW wall.
But for those who continue to act as if Xbone only has the 68GB/s, it's another hole in them. By their thinking Haswell would only have whatever comes over the main bus since ED/SRAM counts nothing in their book.
EG for Haswell if it doubles BW, we'd want to see more than 2X performance increase if I understand correctly.
Then again there are tons of other variables, such as if Haswell even has the execution unit headroom for 2X.
I think the Haswell NDA ends June 3 (dont quote me but I think) so that's good, benchmarks soon.
As we know Intel is super smart, so I'm sure they didn't do this lightly.
And expletive I agree to a point, but the fact is IGP's were running into a severe BW wall.
But for those who continue to act as if Xbone only has the 68GB/s, it's another hole in them. By their thinking Haswell would only have whatever comes over the main bus since ED/SRAM counts nothing in their book.
What's your source for such a confident statement? That is, I assume you have one...
Betanumerical
Veteran
Like to add some more substance maybe? and seriously not using the same name/words isn't really any indication they didn't do it for the GPU and its quite obviously stock GCN.
PCBSD is just a variant of FreeBSD, which has many flavors. Depending on which you use, the RAM requirement goes as low as 24MB.
Sony of course will be using BSD, not FreeBSD and will tune it for their uses. And by not using FreeBSD, they won't have to expose their changes under GPL.
Um.. FreeBSD is not GPL. It's licensed under a standard BSD license (http://www.freebsd.org/copyright/freebsd-license.html)
Makes everything look bad actually
http://www.anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested/6
Looking at 768p (as its actually playable) The i74950 is at 39fps it bats everything but the gt 650m it does much better than even the desktop trinity .
Would love to see a next gen radeon of GeForce come with a 128megs of edram/sram built in. The die size is 264mm2 + 84mm2 . ITs really nice
Or maybe it is a proof Intel is not good making GPUs?
Iris Pro does really well in compute I wonder if anything related to the EDRAM? After all we have heard compute shaders should benefit the most from ESRAM.
http://www.anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested/17
Anand does say this
Crystalwell is the EDRAM...
Are we supposed to take Iris Pro results to correspond to the ESRAM so we can say it will not be a miracle worker (in games), or are they unrelated? I am unsure. Would you need to program more specifically for such a cache?
Drivers and other factors could play in here. Nvidia cards benched here would have better drivers and has more capability in most non-math areas such as texels rate.
http://www.anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested/17
Anand does say this
Crystalwell is the EDRAM...
Are we supposed to take Iris Pro results to correspond to the ESRAM so we can say it will not be a miracle worker (in games), or are they unrelated? I am unsure. Would you need to program more specifically for such a cache?
Drivers and other factors could play in here. Nvidia cards benched here would have better drivers and has more capability in most non-math areas such as texels rate.
Well as Anand points out Iris Pro has just slightely less than double the ALU resources of the HD4600 and the compute performance seems to scale pretty perfectly with that. If the eDRAM were providing an additional benefit (other than providing sufficiently increased bandwidth to allow the ALU's to scale) then we should have seen an even greater performance increase.
That said, I have no idea how the latency of Crystalwell compares to the eSRAM in Xbone.
What we do seem to be able to take from this though is that Intels compute performance seems to be pretty spectacular - maybe even beter than GCN! That bodes well for the possibility of utilising the IGP's for compute work along side discrete GPU's for graphics in the future.
- Status
Similar threads
- Locked
