I'd say so, yeah.
10) Frame buffer/ render target can be split up to reside in ESRAM and DRAM.
Xbox One (Durango) Technical hardware investigation
- Thread starter Love_In_Rio
- Start date
- Status
I'd say so, yeah.
10) Frame buffer/ render target can be split up to reside in ESRAM and DRAM.
Billy Idol
Legend
I'd say so, yeah.
10) Frame buffer/ render target can be split up to reside in ESRAM and DRAM.
Well, the devs even gave an example where the sky in a racing game could reside in DDR...that is quite cool imo. The 32MB always seemed a bit small to me...but with this property, it could sure help in a lot of situations.
Also:
11.) A scaler which can be adapted on a per frame basis!!
Killzone: Mercenary demonstrates how good adaptive resolution can work. Now they have hardware support for this in the X1...seems brilliant to me! I hope that devs will use it a lot, exactly the way Graham explained it.
Edit: oh shit, just realized it was already mentioned as 7.)
Was this ever in doubt? I assumed the system was perfectly flexible as the GPU has read/write to both pools. It'd be odd to limit the GPU to only outputting to ESRAM.
There's not a great deal of additional info, I guess. It's important for those reading between PR lines and concluding that there might be a second GPU or the custom processors as something magical to finally put those notions to rest. 8 processors in the audio block explains where the 15 comes from, rather than several secret processors not revealed before. Likewise, it's SI and not VI or GCN 2 or anything uber fancy.
The ESRAM BW also isn't doing anything mysterious. It's just...do we call it dual-ported as there are channels for read and write? 106 GB/s tops in either discrete direction, but bonus BW if you perform simultaneous R/W. The given figure was lowest safe target.
Importantly for me, the design for XB1 has centred around the same value of Wii U - power efficiency. It's an interesting choice rather than going with significant power draw. I'd like to know why they picked that as a primary objective? I wonder if there are long-term objectives for the design more power-sensitive than a mains powered CE device? That's more business discussion, I guess.
It's designed as an (almost) always on device. Same goes for the reason behind the size.
I heard they were thinking of enabling two more CU's. Looks like that was true. Pretty amazing what you learn on forums.
Last I heard they are NOT enabling those 2 extra CU's.
if that was the case, then the total theoretical limit should be 218gb/sec not 204, right?
I read again and again the section about the esram bandwidth, but i stil can't understand how it works in order to get the extra BW.
218 * 15/16 = 204
They fire blank (write) every 8th cycle. so 7 writes + 8 reads 15 mem ops as opposed to 16 total.
No, he actually confirmed it. "Not entirely round".
I dont necessarily see it that way, it gets back to what we thought imo, to target 8GB at a reasonabe cost they more or less figured DDR was the way to go.
I maintain in 90% of alternate universes, Sony would have stuck at 4GB of RAM and Microsoft's decision would have looked very good. It only looks "bad" now because Sony swung for the fences at the last second. And even then, the consolation prize is a lifetime of much lower RAM costs which isn't so bad.
Probably, power draw played in, but I hardly think it was the central tenant on par with Wii U. Wii U drawing 33 watts and XBO should draw well over 100.
What the XBO lacks if anything is some ALU grunt, The DDR/ESRAM is fine.
Overall I wanted to hear more "Low latency ESRAM makes the GPU turbocharged and here's how" type talk, so the article was disappointing to me.
Oh, and I am also disappointed they cheaped out of enabling the two redundant CU's and doing the upclock as well. You are charging $499 for this system, Microsoft...
Not forgetting that the audio processor offload more than 1 core of work, which means that on the CPU side X1 should be around 25% faster than PS4 (this assumes both console require 1 hw thread for OS).
Still, it's a CPU that can deliver 8 hardware threads with solid FP performance (around 100 gigaflops), backed up by quite an high bandwidth.
The state of art ARMv8 A7 is nowhere close in FP performance, and memory subsystem ones.
The reaction to DF interviews are imo not putting the gaming community under a good light to say the least. I got to the point where I actually believe that your average customer, ignorant about anything technical, is less dumb, less biaised, than a lot pretty vocal people on the web acting under the pretense they represent the gaming market.
Anyway interesting, too bad they don't get further into the details, though it is an interview, I would be really interesting to see a lengthy presentation about the arbitration they made, their measurement, etc.
Though most like the main reaction (but at some point why care about dumb asses...) would be the same: pointless flame war.
I think they gave pretty comprehensive overview about the arbitration they made, the esram and how the system is put together as it is, it is just an interview after all.
They spoke about speed bump and what I get from their choices is that bumping the CPU clock speed was more relevant than the GPU.
I won't dispute their conclusion (number of CU vs clock speed), they made their measurement, their analysis, even if they were wrong, I won't be the one able to make the call, devs could if they had access to the measurements made by MSFT => NDA, unavailability of data make it so you believe it or you don't. I will, actually it sounds right.
That is where I agree with Shifty, power constrains had a profound effect on both durango and Orbis, both went with pretty conservative clock speed if you compare those designs to shipping GPU.
Pretty much following their line of thining, one could be better off with say 8 Cus clocked @1GHz, compromising further the raw throughput of the chip but not real world performances (same would apply to the PS4). The issue is power consumption.
So to make it short, if I were to criticize the design in the context of what those 2 high rank engineers did, I would not criticize their choices (I'm not legitimate to do it to begin with and we should be able to asses the end result (shipping games) pretty soon) but possibly the power constrains they had to deal with.
A better Durango could be different than what people on the web would want aka more CUs, more raw theoretical throughput, etc.
My understanding of their pov is that a better durango would be something like this:
6 CPU cores @2GHz
8 CU/16ROPs GPU @1GHz
Keep all the embedded processors
Invest all the "saved" silicon in more eSRAM.
Trade off, less PR friendly actually I think it would be a disaster, extremely tough to manage/explain, and a significant jump in power consumption (though I would bet far from unmanageable).
I can only imagine what reactions would be if people were indeed comparing a ps4 as it is now and an Xbox such as described above => web implosion.
Anyway interesting, too bad they don't get further into the details, though it is an interview, I would be really interesting to see a lengthy presentation about the arbitration they made, their measurement, etc.
Though most like the main reaction (but at some point why care about dumb asses...) would be the same: pointless flame war.
I think they gave pretty comprehensive overview about the arbitration they made, the esram and how the system is put together as it is, it is just an interview after all.
They spoke about speed bump and what I get from their choices is that bumping the CPU clock speed was more relevant than the GPU.
I won't dispute their conclusion (number of CU vs clock speed), they made their measurement, their analysis, even if they were wrong, I won't be the one able to make the call, devs could if they had access to the measurements made by MSFT => NDA, unavailability of data make it so you believe it or you don't. I will, actually it sounds right.
That is where I agree with Shifty, power constrains had a profound effect on both durango and Orbis, both went with pretty conservative clock speed if you compare those designs to shipping GPU.
Pretty much following their line of thining, one could be better off with say 8 Cus clocked @1GHz, compromising further the raw throughput of the chip but not real world performances (same would apply to the PS4). The issue is power consumption.
So to make it short, if I were to criticize the design in the context of what those 2 high rank engineers did, I would not criticize their choices (I'm not legitimate to do it to begin with and we should be able to asses the end result (shipping games) pretty soon) but possibly the power constrains they had to deal with.
A better Durango could be different than what people on the web would want aka more CUs, more raw theoretical throughput, etc.
My understanding of their pov is that a better durango would be something like this:
6 CPU cores @2GHz
8 CU/16ROPs GPU @1GHz
Keep all the embedded processors
Invest all the "saved" silicon in more eSRAM.
Trade off, less PR friendly actually I think it would be a disaster, extremely tough to manage/explain, and a significant jump in power consumption (though I would bet far from unmanageable).
I can only imagine what reactions would be if people were indeed comparing a ps4 as it is now and an Xbox such as described above => web implosion.
Last edited by a moderator:
Betanumerical
Veteran
There is no split, we know this as a fact, so can we stop with this rubbish.
Rangers wasn't talking physical split. Cerny confirmed usage split and seeking gpgpu horsepower early on the console life.
You can gate and throttle so hardware doesn't consume its peek at all times, or can chuck in a low power CPU/mobile SOC for low-power stuff. They could have chosen to design a box with 200+ watts draw while gaming and far less while just streaming media if they had wanted.
The article says as much.
OT
"Andrew said it pretty well: we really wanted to build a high performance, power-efficient box,"
"Having ESRAM costs very little power and has the opportunity to give you very high bandwidth. You can reduce the bandwidth on external memory - that saves a lot of power consumption and the commodity memory is cheaper as well so you can afford more. That's really a driving force behind that... if you want a high memory capacity, relatively low power and a lot of bandwidth there are not too many ways of solving that."
That's OT platform comparison. Irrespective of what rivals are doing, MS had a choice for their console whether to go high power draw, high performance, or low-end, and they have chosen the low end. This article admits that, which helps understand some of the technical choices (like no second GPU
). Whether that's the right choice or not is a business discussion rather than technical investigation.Yes, they didn't give any details on how the ESRAM improves performance, such as an example of where in their software testing they are seeing the ESRAM as an enabler. We haven't got any nitty-gritty low-level details, not even timings. How low is their low latency anyway?! So a lot of our questions remain answered, but the rest of the internet has at least some much needed clarification.
Hell, we didn't even get an answer to if XB1 is tier 1 or Tier 2 PRT!
Betanumerical
Veteran
Yes but the point is its dynamic and you don't have to do it therefore there is no real point to be made and that all the CU's are identical there may be scheduling on some that favours GPGPU (more queues?) or something but other then that.
You can gate and throttle so hardware doesn't consume its peek at all times, or can chuck in a low power CPU/mobile SOC for low-power stuff. They could have chosen to design a box with 200+ watts draw while gaming and far less while just streaming media if they had wanted.
The article says as much.
OT
"Andrew said it pretty well: we really wanted to build a high performance, power-efficient box,"
"Having ESRAM costs very little power and has the opportunity to give you very high bandwidth. You can reduce the bandwidth on external memory - that saves a lot of power consumption and the commodity memory is cheaper as well so you can afford more. That's really a driving force behind that... if you want a high memory capacity, relatively low power and a lot of bandwidth there are not too many ways of solving that."
That's OT platform comparison. Irrespective of what rivals are doing, MS had a choice for their console whether to go high power draw, high performance, or low-end, and they have chosen the low end. This article admits that, which helps understand some of the technical choices (like no second GPU
Yes, they didn't give any details on how the ESRAM improves performance, such as an example of where in their software testing they are seeing the ESRAM as an enabler. We haven't got any nitty-gritty low-level details, not even timings. How low is their low latency anyway?! So a lot of our questions remain answered, but the rest of the internet has at least some much needed clarification.
Hell, we didn't even get an answer to if XB1 is tier 1 or Tier 2 PRT!
Well, from his line on GPGPU its seems the ESRAM will come to bear, though he did note that it will discussed better in a later article.
Averagejoe
Regular
I don't think it makes much difference,because that would imply that the PS4 still have 400Gflosp on the compute side that most be compensated by something on the xbox one hardware any way.
14+4 or just plain 18 for all in the end i think is the same.
- Status
Similar threads
