Tadgerot's forum posts
[QUOTE="Tadgerot"] This is thing, any number of rumours could be right, I'm just saying my theory holds as much weight as any other, proof wise. On another note, do we think this speed increase is independent on JIT compression or is it including this benifit. If its included then I can see where the theoretical speed comes from. If not and they are claiming 192 without the compression, well then I dunno what XBO will output but it will be huge. This kind of makes me Beleive my source even more. "Seems accurate but they do not have detail of the 'move engine' and the esram. The 'move engine' as they put it is out sdma engine which has a desperate encode and decide engine which can operate a synch... They can decode video memory compressed jpegs directly to esram for render by the texture pipe line. Anyhow will see I guess. I've played with the decompress engine myselfm its pretty powerful."tormentos
They made a bandwidth claim it has nothing to do with JIT compresion or DME.
Yep that's what I am thinking, in which case the XBO is gonna be more powerful than any of us thought with the inclusion of them move engines.[QUOTE="Tadgerot"]Hang on let me get this straight...I am getting confused :( If 102 still holds true it means no downclock, is that right? The DF journalists also found out from MS that the memory through put was an 88% increase. Now if I was the journalist and I was asked to write up the DF article, I would look at 102 and think, ok that's the bandwidth. Then I would increase that number by 88% rounded up. Which makes 192. I don't understand how people started taking numbers of the gpu clock to suit there needs, especially when it says in the article that there has been no downclock. So to summarise, is it possible the editor just added 88% to 102 to get that figure? Seem more likely that the editor downclocking the gpu and adding them together?tormentos
Because the whole theory was based on reading and writing at the same time,MS tryed to hype reading and writing by ESRAM as something new,when it is in fact one of the leaks,first you have to believe that MS just found out this,and second this is well known,how do you think peopel came with the whole 170GB/s bandwidth for the xbox one.?
102 + 68 = 170 GB/s.so basically this just rise the 170GB/s number by 22GB and not the 88% MS claims now.
Since the whole theory was based on reading and writing at the same time,192GB/s translate into 96GB/s.
Which doesn't match up with 102GB/s which would be 800mhz 128 bits.
But 750mhz 128 bits fits perfectly with the 192 number,and this is funny because DF claim that it hasn't been a downgrade base on its source,but MS has refuse to reveal the xbox one GPU clock speed,because they know people will get the TF count as soon as its know.
So MS could be even saying to DF there is no downclock on the GPU,while the speed it 750mhz,because MS never revealed the GPU speed.
This is thing, any number of rumours could be right, I'm just saying my theory holds as much weight as any other, proof wise. On another note, do we think this speed increase is independent on JIT compression or is it including this benifit. If its included then I can see where the theoretical speed comes from. If not and they are claiming 192 without the compression, well then I dunno what XBO will output but it will be huge. This kind of makes me Beleive my source even more. "Seems accurate but they do not have detail of the 'move engine' and the esram. The 'move engine' as they put it is out sdma engine which has a desperate encode and decide engine which can operate a synch... They can decode video memory compressed jpegs directly to esram for render by the texture pipe line. Anyhow will see I guess. I've played with the decompress engine myselfm its pretty powerful."[QUOTE="LanceSSJ"]the point of his tweets is to say that THE GPU IS NOT DOWNCLOCKED. Thread doneFar_RockNYC
it's going to take about 4-6 hours for cows to read that and another hour for them to understand it.
That is great, I have been pissing my self so much since I read it, I just had to post.I respect neogaf, but there are some serious cows on there and I think the techie peeps don't look at the wider picture enough. Ok this is my take. The journalist writing up the article on digital foundry states that there is an 88% increase on 102.4 GB/s. I think he or she simple multiplied 102.4 by 1.88 to obtain 102 + 88 % performance. I could be wrong but it seems more likely than MS secretly downclocking the gpu by 50mhz and not telling partners or developers.Posted in neogaf
Not possible. Because that means the GPU clock rate is doubled (Calm your shit Kotaku, I'm using this as an example). They are saying they are doing a read AND write at the same time. So it "doubles" the bandwidth... (102x2 = 204GB/s)
Except they said 88% increase... = 192GB/s... but... where the **** did that number come from? Why wouldn't it be able to be a 100% increase? Well... because you'd have things trying to right and read to conflicting spots at the same time.
So that's why they say "realistically" it's 133GB/s....
But wait... if the 88% increase isn't account for those conflicts and data misses... then how did they make up that number?
So...currently.. they say:
128 (bit) x 800 (clock rate) = 102.4GB/s
Well.. downclocking the GPU from 800 to 750 (going down 50) will get you...
128 (bit) x 750 (clock rate) = 96GB/s
Double that since they are saying you can read/write at the same time...
96 x 2 = 192GB/s
So... several of us are thinking this 88% thing is just a bs number they came up with after changing around their GPU clock and "double accessing" the RAM.Raziel831991
I don't know why everyone seems to Beleive there is a downclock? Can someone enlighten me? It says in the DF article that this is not the case.Posted in DF comments about this topic
Reads more like creative accounting to disguise a GPU downclock from 800MHz by 50MHz.
Raziel831991
750 (freq) * 128 (bits) = 96 GB/s
96 GB/s * 2 (simultaneous read/write) = 192 GB/s
How much does MS pay you to write this garbage, Leadbetter?
Log in to comment