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

ethomaz on 28 June 2013

No difference after all.

The eSRAM can run at 1TB/s that will not change the fact the biggest pool of RAM runs at 68GB/s... you can't use the eSRAM for all just for cache some little data (32MB).

8GB@176GB/s vs 32MB@192GB/s

"Now that close-to-final silicon is available, Microsoft has revised its own figures upwards significantly, telling developers that 192GB/s is now theoretically possible. Theoretical peak performance is one thing, but in real-life scenarios it's believed that 133GB/s throughput has been achieved with alpha transparency blending operations (FP16 x4)."



Around the Network
manny10032 on 28 June 2013
MaulerX said:
manny10032 said:
Did they actually improve it or is it just a theoretical figure?

 

 

Even the PS4 bandwidth is a "theoretical" max. They all deal in theoretical maxes. But regardless, yes, it's an actual improvement over the previous theoretical max.


Ahh thanks for the explanation!



coolguy on 28 June 2013

sounds like MS spin to me



VITA 32 GIG CARD.250 GIG SLIM & 160 GIG PHAT PS3

Nsanity on 28 June 2013
ethomaz said:
No difference after all.

The eSRAM can run at 1TB/s that will not change the fact the biggest pool of RAM runs at 68GB/s... you can't use the eSRAM for all just for cache some little data (32MB).

And... why people sum the bandwidths?

Is this your opinion or some else's from NeoGaf?



Azerth on 28 June 2013

So just in terms of the x1 is this good news? Not comparing it to the ps4



Around the Network
ethomaz on 28 June 2013
Nsanity said:
ethomaz said:
No difference after all.

The eSRAM can run at 1TB/s that will not change the fact the biggest pool of RAM runs at 68GB/s... you can't use the eSRAM for all just for cache some little data (32MB).

And... why people sum the bandwidths?

Is this your opinion or some else's from NeoGaf?

What do you mean?

Edit - I undertood now... no it's always my opinion... I didn't even read the gaf before write that... why?

And I read wrong and fixed... they are not suming the bandwidth.



ethomaz on 28 June 2013
kowenicki said:
BOOM!!

hilarious.

Why?



MaulerX on 28 June 2013
theprof00 said:
disolitude said:
theprof00 said:
wow looks like disolitude is wrong again :D
Guess they did go for higher bandwidth in the end, despite dis saying they didn't need it.


Actually no, it proves me right. It shows that X1 didn't need touse GDDR5 RAM. Perfectly capable with DDR3 + esRAM

I thought your point of that thread was that their setup was perfectly in sync with their output, and they didn't need the high speed of gddr5. If I'm wrong about that, my bad. That's how I interpreted it, although I do see that a large part of it was about gddr5, I thought it was more about the speed

 

If anything, it proves that the term "gddr5" is somewhat overplayed. Yes, gddr5, by nature, has a much higher bandwidth than  DDR3. But that doesn't mean that gddr5 is the only means by wich to achieve or obtain high bandwidth. At the end, Microsoft went with low latency DDR3 (for their own W8 app ambitions) while simultaneously having a boost in bandwidth with the esRAM.

 

There will always be a debate as to wich solution is better/worse but by judging this DF article they seem to believe that any performance advantage that the PS4 has over the X1, will not be evident until perhaps a few years down the line. And even then, it'll probably just be relatively minor (ie slightly higher resolutions). The thing I got the most is that these systems will be much closer in performance than we were initially let to believe.



ethomaz on 28 June 2013

What I think....

MS is trying to compensate the IQ drop compared with PS4 using magical post-processing... the 32MB eSRAM will be used to that... post-processing... the render will be made in the 68GB/s RAM yet.

32MB eSRAM is enough to post-processing 1080p output... but the render part will have better IQ in PS4... you can't render in 32MB.



theprof00 on 28 June 2013
MaulerX said:

If anything, it proves that the term "gddr5" is somewhat overplayed. Yes, gddr5, by nature, has a much higher bandwidth than  DDR3. But that doesn't mean that gddr5 is the only means by wich to achieve or obtain high bandwidth. At the end, Microsoft went with low latency DDR3 (for their own W8 app ambitions) while simultaneously having a boost in bandwidth with the esRAM.There will always be a debate as to wich solution is better/worse but by judging this DF article they seem to believe that any performance advantage that the PS4 has over the X1, will not be evident until perhaps a few years down the line. And even then, it'll probably just be relatively minor (ie slightly higher resolutions). The thing I got the most is that these systems will be much closer in performance than we were initially let to believe.

In regards to speed.

The ps4 still churns out more flops. Showing that the RAM is faster doesn't mean the system is more powerful. In fact, this is probably just spin, as we already know that 3GB is already dedicated to the OS, and kinect will require its own processing as well. Those things will take up bandwidth.

The original bandwidth they were talking about was probably the actual speed available to games. This new "theoretical" number is probably something like with kinect disabled, or some such other band-sinks being disabled.

Not really a point I need to argue though. Give it a week or two and we'll have a dev saying "new speed is fake". MS sure have been getting called out left and right recently.