By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - Microsoft - DigitalFoundry: X1 memory performance improved for production console/ESRAM@192 GB/s

disolitude said:

 PS4 is most likely utilizing 1 Ghz memory clock GDDR5 RAM, while Xbox One is using 1066 Mhz for its DDR3 ram. You could argue that ESRAM may add more power usage for Xbox One though...

There is absolutely no need to invent numbers by you, again and again and again and again and again.....
We KNOW the command clock of the gddr5 chips is 1.375GHz. We know it is LOW POWER gddr5 chips, hence only 176G/s instead of 192G/s. READ THE F"CKING Samsung or Hynix MANUALS for power requirements. PS4 uses less power for memory than the XBox One.



Around the Network
drkohler said:

WTF???? Power is used to switch transistors and load/unload capacitors. All the power is transferred to heat in the end. This is really a no brainer: The more powr a circuit requires, the more heat it generates. Basic electronics.

Yeah... for the same component.

GDDR5 and DDR3 are different components... it is like a CPU from Intel consume the same power but is more cool than AMD.

I mean... I said GDDR5 require less power than DDR3 but it generate more heat DDR3.



Player1x3 said:
sth88 said:
Great news! Sounds like Xbox One games should be able to keep up with PS4 games for the most part; as long as Microsoft can keep it fairly close, they should be good to go as far as graphics are concerned.


Uh...all this article does is debunking the downclock rumor and and counts for a theoretical peak performance of both pools added together...PS4 still has much better RAM and GPU


I hope people do know that the PS4 will not reach peak 176 as well.  The CPU takes it chunk, audio, Sub systems ect.  when all is said and done, bandwidth to the GPU part of each console will not be something that hold either console back.



ethomaz said:
A question....

The eSRAM clock runs at the GPU clock or it can run async?


That is a good question. I think you have to consider both cpu and gpu. But generally it's ok if memory is faster than the fsb of cpu/gpu - it just must not be lower...



walsufnir said:
ethomaz said:
A question....

The eSRAM clock runs at the GPU clock or it can run async?


That is a good question. I think you have to consider both cpu and gpu. But generally it's ok if memory is faster than the fsb of cpu/gpu - it just must not be lower...

For 192GB/s you have two options... 750Mhz or 1500Mhz... but MS are saying they can read and write at the same time so that give us 750Mhz for eSRAM.

Or the eSRAM runs at lower clock of GPU.... or the GPU is running at 750Mhz too.

I think (no sure too) that the CPU runs at 2x the GPU clock in AMD APU (not a rule but in most case run in that config).



Around the Network
JoeFlex said:
128 bytes per block multiplied by the GPU speed of 800MHz offers up the previous max throughput of 102.4GB/s. It's believed that this calculation remains true for separate read/write operations from and to the ESRAM.However, with near-final production silicon, Microsoft techs have found that the hardware is capable of reading and writing simultaneously.

 

Hahahahahahahahahahahahahah. Some MS marketing guy has found a way to add numbers, again (probably hovering stoned in the Cloud, somewhere)

Look, IF the clock is 800MHz AND the bus size is 128bit, then the maximum throughput of the esram is 102GB/S, regardless whether you read or write).  END OF STORY.

IF they have achieved 192GB/s (in theory) this means that: clock 750MHz bus width 256bits. This is a major redesign of the APU so my money is on some Microsoft PR guy adding numbers as he pleases.



drkohler said:

Hahahahahahahahahahahahahah. Some MS marketing guy has found a way to add numbers, again (probably hovering stoned in the Cloud, somewhere)

Look, IF the clock is 800MHz AND the bus size is 128bit, then the maximum throughput of the esram is 102GB/S, regardless whether you read or write).  END OF STORY.

IF they have achieved 192GB/s (in theory) this means that: clock 750MHz bus width 256bits. This is a major redesign of the APU so my money is on some Microsoft PR guy adding numbers as he pleases.

They are adding numbers to that read and write but the main questions is... 192GB/s (96GB/s) means 750Mhz and not 800Mhz anymore.

The downgrade rumor gain more credibility... at least in eSRAM clock.



Good for xbox one owners. But it wot matter. I n the end it will matter in service and games.



ethomaz said:
drkohler said:

WTF???? Power is used to switch transistors and load/unload capacitors. All the power is transferred to heat in the end. This is really a no brainer: The more powr a circuit requires, the more heat it generates. Basic electronics.

GDDR5 and DDR3 are different components..

I mean... I said GDDR5 require less power than DDR3 but it generate more heat DDR3.

No,no, no. ddr3 and gddr5 do EXACTLY the same thing. They charge/discharge a capacitor in a bitcell to figure out if there is a 1 or a 0. The path to reading the bitcell is a lot more complex in gddr5 than in ddr3 (because gddr5 has a lot more options and has to clock a lot faster), but the whole circuitry in both cases is made of transistors, tons and tons of transistors. There is no magic that says transistors in gddr5 create more heat than the same transistors in ddr3. The power you sink into a chip is converted into heat in the end.



drkohler said:

No,no, no. ddr3 and gddr5 do EXACTLY the same thing. They charge/discharge a capacitor in a bitcell to figure out if there is a 1 or a 0. The path to reading the bitcell is a lot more complex in gddr5 than in ddr3 (because gddr5 has a lot more options and has to clock a lot faster), but the whole circuitry in both cases is made of transistors, tons and tons of transistors. There is no magic that says transistors in gddr5 create more heat than the same transistors in ddr3. The power you sink into a chip is converted into heat in the end.

I don't know they are the same... thanks.