Not Expecting Xbox One eSRAM Bottleneck To Be Patched

You know thats not true. If any thing X1 has more room for optimization and growth. Unlike the straight forward PS4. It's going to be the lazy devs console.

And when are buying a console Torms? You and Thuway don't even own consoles and talk a lot of smack. lol

Hard to compete with the PS$ being the worlds most powerful video game console to have ever been created.

Lems can't catch a break ;'(

Would be a treat, too bad it isn't happening.

Both of these consoles are weak sauce. They both have huge bottle necks. Xbox one home of the non P's. PS4 home of the 30 fps.

Carry on.

Dx 12 will save the xbone.

@ZombeGoast: Resogun is 1080p 60fps locked. Eat your words

Will Ryse's graphical level be enough in a few years? How will that compare to what's on the competitor's system? I don't know, even if it's not an excuse, it certainly isn't a favorable argument.

READ THE ARTICLE THE GUY DID NOT MAKE RYSE hes never worked on the x1 ESRAM PATCHED? sounds like babble

Who needs games when you have TV and Kinect?

Well duh, you can't "patch" a hardware bottleneck. Those thinking some SDK changes would overcome that were just fooling themselves.

You can compensate for it using software but you'll never completely patch it out and it will always be something devs have to work around.

A hardware bottleneck can only be removed with new hardware or bypassing the bottleneck all together. Considering the ESRAM is used for low-latency RAM to compensate for the lower throughput of DDR3, it's not something you can just avoid using.

@Wasdie: Correct me if I'm wrong as Ive said Im still learning this stuff. It sounds like theoretically you can write to both sets of ram. They are not independent of each other. Does this mean that you could write certain aspects of the frame buffer into esram things that need to stream faster, and things that need to read slower through the ddr3?

bu bu but t3h powerz 0f the cloodze :(

Exactly, no, I don't, but Salt_The_Fries implied that it's "enough" for the system: "ask yourself one question: does Ryse in its current shape and form need optimization? If it does, it's minuscule at best."

Are you really in this conversation?

Well yeah, that's exactly why they have that 32mbs of fast ram. The problem is that 32mbs isn't enough for a 1080p image, especially when you need 60 frames per second. It's a straight up hardware bottleneck. They should have gone with GDDR 5 system ram like Sony did if they wanted 1 pool of unified memory. They could have done a more PC like architecture and gone like 2 gbs of GDDR5 vram (more than enough for 1080p) and then like 6 gbs of DDR3 system ram.

Ok I was under the impression that you don't have to write everything to just the 32mb. The example of a frame buffer I saw it had specific things targeted to specific ram addresses. I figured you should be able to write certain objects such as background textures, sky's, and etc to the ddd 3 ram and save the 32mb for the things in the foreground as well as updating the camera position. Ill also find this post where a developer from trials gave an example of writing a frame buffer that would fit in esram

edit: posted by Sebbi

"MJPs g-buffer layout is actually only two RTs in the g-buffer rendering stage and one RT in the lighting stage. And a depth buffer of course. Quite normal stuff.

On GCN you want to pack your data to 64 bpp (4 x 16 bit integer) render targets because that doubles your fill rate compared to using more traditional 32 bpp RTs (GCN can do 64 bit filling at same ROP rate as 32 bit filling).

I assume that the packing is like this:

Gbuffer1 = normals + tangents (64 bit)

Gbuffer2 = diffuse + brdf + specular + roughness (64 bits)

Depth buffer (32 bits)

Without any modifications this takes 40 megabytes of memory (1080p).

The lighting step doesn't need extra 8 MB for the 4x16f RT, because compute shader can simultaneously read and write to the same resource, allowing you to to lighting "in-place", writing the output over the existing g-buffer. This is also very cache friendly since the read pulls the cache lines to L1 and the write thus never misses L1 (GCN has fully featured read & write caches).

It's also trivial to get this layout down to 32 MB from the 40 MB. Replace gbuffer1 with a 32 bit RT (32 MB target reached at 1080p). Store normal as 11+11 bit using lambert azimuth equal area projection. You can't see any quality difference. 5+5 bits for tangents is enough (4 bits for exponent = mip level + 1 bit mantissa). 11+11+5+5=32. Also if you only use the tangents for shadow mapping / other planar projections, you don't need them at all, since you can analytically calculate the derivatives from the stored normal vector.

This layout is highly efficient for both g-buffer rendering and lighting. And of course also for post processing since all your heavy data fits in the fast memory. Shadow maps obviously need to be sampled from main memory during the lighting, but this is actually a great idea since the lighting pass woudn't otherwise use any main memory BW at all (it would be completely unused = wasted)."

Posted in this thread: http://forum.beyond3d.com/showthread.php?p=1830129#post1830129

I feel developers would be better off not even looking to do anything with the ESRAM.

It was only added in the last minute to bring up the bandwidth of the console stats wise to try and shift more consoles and not to give Sony a reason to advertise how there console has multiple times the bandwidth of its physical memory.

that makes no sense.

@FoxbatAlpha said:

@hoosier7 said:

@FoxbatAlpha:

Dude you are not going to win with me no matter what you show. The PS4 is based off of horsepower. What the ONE is based off of hasn't been even used yet and requires less powerful hardware. In a year or two if the Xbox is still doing the same thing with sub 1080p and hasn't made a benefit from tiled resources, cloud compute and DX12, then I will hand you this victory.

Until then 1080p hasn't been a game changer for the PS4 like cows claim. The differences are very minor and in some instances I still think the ONE has better detail. Don't forget about the other mulitplats that are better on the ONE also. The tide is already turning and the Full POWA of the ONE hasn't even started.

Fairly round bout way of dodging the weaker Cpu statement.

Deviating much? "I'd be happy if Ryse-quality graphics became standard this gen."

So... "excuses?"

Yes.

Simple objective text interpretation man...

Umm...yeah it does!

8GB of DDR3 is a lot slower then 8GB of GDDR5, Microsoft knew that, Sony knew that...everybody knows that.

Microsoft then adds 32Mb of ESRAM to then bring up the bandwidth, as ESRAM is expensive but insanely quick. This then brings the Bandwidth of the consoles physical memory lower but not by much leaving Sony unable to capitalise on it and use it to their advantage as a selling point for their console about how much quicker its bandwidth is over the X1.

The truth of it is still that 32Mb isn't really that much and there isn't a lot of stuff you can logically fill it up with. The new batmobile model takes up 160mb in itself, this then clearly means that 99.9% of the game goes through the much slower 8GB of DDR3 memory where 100% of PS4 games go through the much quicker 8GB of GDDR5 memory.

Hope that clears it up for you.

That won't be a problem in the near future. Especially with DX12, tile resources and new graphic engines.

This problem will only get worse vs the ps4 there is no secret sauce. MS is trying everything they can to minimize the gap in power vs the ps4.

1. They cloud will increase the xflops powah

2. There is really no noticeable difference between 720p vs 1080p cause we upconvert to 1080p

3. The dx12 will save us

4. We will use eSRAM so there is no bottle neck vs the ps4

5. We got new rendering and stacked secret chips

Did I miss any other damage control ??

It's ok. Let them believe what they want. They think they know everything. Another year or so from now they will be like. "WTF. How is the XB1 doing that!?!?" lmao!!

It doesn't explain why your initial post calls for devs to ignore 32mb of the fastest ram on the system just to keep Sony from advertising the hardware differences. Hardware is nice but games sale consoles not hardware and ignoring that 32mb of fast ram for marketing purposes is ridiculous.

32mb is not a lot of ram but if I am correct (currently looking for the developers guide or power point for x1) the 32mb of esram and the 8gb of ddr3 are viewed as the same pool of memory. You do not have to pass in a linear fashion between both. You can write to esram and ddr3 for the same frame buffer. Furthermore I posted an example of a frame buffer layout that could make it feasible to fit a 1080p frame buffer into 32mb of memory. I personally cannot test it but the members of that forum are active developers. So while I thank you for regurgitating the same talking points it is not needed as it has been beat to death on this forum. Does that clear it up for you?

xboys think a software upgrade will physically upgrade the hardware

The eSRAM is only 32mb vs the 8gb of ddr3. On pc when your gpu runs out of memory and has to go to system memory your games take a massive performance hit. Sure exclusives can try to work around this problem but that will take time and money and still wont address the gap in gpu's (x1 vs ps4). MS trired to be slick and overclock to try to distract xbots yet again.

Yet multiplats show that the ps4 has a major advantage and cost 100 bucks less.

thia is not true

7790 has 14CUS

Xbox One has only 12 working CUs and 10% os this is reserved for kinect so its 10,8

about 30% lower end than 7790

noy to mention the lower clocks and DDR3 memory

X1 is a joke

Yes, lets not take one of the worlds best engine creators seriously.

XBottleneckOne.

Why are you so willfully ignorant?

@Indicud: I can't know.

and teh cloud, keep on hoping!

i can already see it, in 3 years: "guys microsoft said teh directx 13!!!!"

Patching hardware???

Lol

What a compiling argument by bringing a list of games.

Total buffoonery the xbox one can't optimize pas 1.31 Tf hell not pass 1.28TF which is its limit period,the PS4 has a stronger GPU that doesn't require any mystical coding or some obscure technique.

There is a difference between optimization and braking your back to try to make the code run optimal on the xbox one,there is no much room for improvement on xbox one,and you will see that as time move on,in the xbox 360 case the more it got push the lower the resolution went,i expect the same with the xbox one.

But if you have any other secret sauce please share it with us..lol

You can Turn 10 already did it on Forza 5,the sky since it doesn't change or does anything was allocated on the main pool,anything else was move to ESRAM,but if you watched forza 5 early show,you can see that the spectators were 3d models, but on the retail copy they were replace with photos of people to cut resources consumptions,you can do it in the end basically yield the same results,you can't have dynamic sky or weather if your using the main pool of ram is to slow.

I am sure that is the same reason why MGS5 has dynamic sky while the xbox one version lack it completely.

Hahahaaaaaaaaaaaaaaaaaaaaaaaa......

Man you will be so own next year..hahaha

Didn't you say MS would deliver on this year E3.?

That's not what he says.

He says that developers can't rely on the hardware manufacturer to make the patch so they, already being experienced programmers, make the works-arounds to the hardware limitations for their games when they can.

It is a technical limitation dude.

I can’t speak to whether this will be patched or improved in the future, but I wouldn’t expect it as a developer,” he said to GamingBolt. “Game developers are masters at engineering workarounds to hardware limitations

You don't need a PHD to understand what he is saying.

Ryse look great but is 180 turn from what they do on PC,which is more open games hell Crysis on PC is endlessly open compare to Ryse,which you can't move more than 6 feet from side to side most of the time,and you can't basically explore anything,not to mention that Ryse was say to be 1080p with 150K polygon characters,only to be downgrade to 900p with 85k polygon characters and still Ryse isn't a 30FPS lock game,it has drops into the teens and runs mostly at 26 FPS.

This come from Crytek far from your average weak developer.

Seriously, if instead of using the expensive and problematic Esram, MS had just done something easy like throwing an extra 2 GB of GDDR5 on the side, we woudn't be having this discussion, except for the weak GPU.
Only an idiot would have thought that DDR3 would be fast enough for a GPU...
They are supposed to be the experts, but a decision like that, even if it was taken years ago shows a complete lack of intelligence from the decision makers at MS. I pity those engineers who knew how stupid the top was but could just stay there and watch the fail unfold knowing they would be blamed for it years later. Stupid Mattrick and Balmer and their egos.

Yep and lemmings actually believe that he ESRAM inclusion is some how something to boost performance of graphics,when in reality is something to fix MS choosing DDR3 over GDDR5,if the xbox one had GDDR5 ESRAM would be out of the picture,and the xbox one would perform better,obviously to the point were its GPU allows it.

The xbox one actually cost Dom Mattrick his job in the end just like the PS3 cost Kuturagi his,it was a blonder which cost them the little good faith some had in MS and little good will they won with the xbox 360.

Well, with the 360, the Edram was a nice addition, but the system memory was GDDR3 and not DDR2, so it was fine.

What I can agree with MS is that for multitasking and multiple OS, DDR3 is the best choice. Mabey even 8Gb of GDDR5 and 2GB of slow DDR3 1066Mhz would have been enough. The thing is, with a single SOC, would it even have been possible to have two main pool of system ram accessible? mabey with one SOC that would have been imposible, making the full GDDR5 memory the only true viable option, just like how GDDR3 was for the 360.

As for the PS3, well, Sony got a bit over ambitious with the Cell and it didn't live up to it's expectation. The blu ray was a cost killer, but in the end, it helped destroy HD DVD, so only the Cell was the only true mistake (or more of a failed prototype than a real mistake), with the low amount of Ram (just like the 360). Kuturagi's mistake is nothing compared to Mattrick's complete stupidity.

It would have been a boost if MS had put in 48mb so they could fit 1080p in the framebuffer. But they cheaped out. DDR? 32mb ESRAM? Slower GPU? Proprietary hard drives?

PS3, despite a wonky architecture, was on paper more powerful than the 360. Xbone is not in any way more powerful. Its both weaker and harder to dev for. Its weaker than the PS4, and even getting to that level of weakness is more difficult. No one is gonna bother, they're just gonna stick with 900p for this gen.