Especially when there's video that could have been used to argue the same point.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
DigitalFoundry: X1 memory performance improved for production console/ESRAM 192 GB/s)
- Thread starter ekim
- Start date
- Status
NoirVisage
Banned
I demand to know your sources!
IHaveCandy
Banned
Memory bandwidth is less vital when processing power shrinks. It is more necessary for PS4 to have a higher bandwidth because it can crunch data faster.
qa_engineer
Member
Where did you get the TMU numbers from?
Vestal
Gold Member
remember playing virtua tennis over that 56k connection!!!!
and that sexy ass controller with pressure triggers, was so revolutionary back then, awesome for baseball games.
Suikoguy
I whinny my fervor lowly, for his length is not as great as those of the Hylian war stallions
Two reasons;
1.) It looks like BS, simply because you don't, under any normal circumstances, discover a sudden 88% increase in bandwidth months before a console launches.
2.) It might have inadvertently confirmed a down clock, which is quite amusing since it appears to be spin in the first place.
The downclock would be only very marginal, and probably not have much of an effect.
Same. I fondly remember playing NFL2K, Sonic Adventure, Power Stone, and Ready 2 Rumble all day and all night on launch day. Some of those haven't aged well, but the console itself was amazing for its time.
And the ads! "It's thinking."
http://youtu.be/EH9Xx-nXTo8
Yeah, whoever decided it would be a good idea to flood the market with hardware in a short span of time should have their mouth peed in. Sega CD, 32X, Saturn, Dreamcast...it was too much, too fast.
McHuj
Member
Each CU in GCN has 4 TMU's.
shinra-bansho
Member
This thread is fun.
If you have 10 apples and previously had 5 apples you have 100% more apples than previously. 50% more apples than an original amount of 5 apples is 7.5 apples. Perhaps the comment was ignored because it was based on a false premise.
IHaveCandy
Banned
Alright if last gen was the move to HD, I want this gen to be the move to high IQ. I don't want to see aliasing anywhere. It ruins high quality visuals so much. Those 6 extra CUs on the PS4 better at least give me extra high IQ on games designed to not look too gimped on Xbox One.
Man, a eSRAM with only 32 MEGA BYTES. I highly doubt it can do something significant in terms of graphics.
Moreover, industry will get rid of DDR3 as VRAM in a year or two. This will lead GDDR5 price down and industry uses it as standard VRAM.
So Xbox One specs will seem dated after 2 years.
Moreover, industry will get rid of DDR3 as VRAM in a year or two. This will lead GDDR5 price down and industry uses it as standard VRAM.
So Xbox One specs will seem dated after 2 years.
Suikoguy
I whinny my fervor lowly, for his length is not as great as those of the Hylian war stallions
This is up to developers to an extent, you should really get into PC gaming if IQ is a such a significant priority.
The scenario posted omits a couple things: caches, and memory controllers.
Looking at the GCN whitepaper, you'll see a block with Z$, C$ and Color ROP/Z Stencil ROP units. The Z$ is the depth buffer cache and the C$ is the color buffer cache.
It's only in the case of a miss (and when flushing) that you'll actually need to go out to main memory (or eSRAM) to fetch whatever information you need, and I hope that scheduling is smart enough to pipe fragment data from wavefronts to ROP groups that have already been working on that portion of the screen, as to minimize reading/flushing of the caches.
Also, in the case of depth (depending on operating mode) the rasterizer can cull entire triangles based on early-Z, and this process is sped up by having a hierarchically compressed Z buffer for fast min/max depth on a region.
Lastly, as the diagrams in the whitepaper show, there are several individual memory controllers. I don't know much about how memory controllers work, but I'd assume that they can each individually read or write on a given cycle (please correct me if I'm wrong here).
IHaveCandy
Banned
Suikoguy
I whinny my fervor lowly, for his length is not as great as those of the Hylian war stallions
I do play PC games. It lets me know how much poor IQ ruins console games. TTT2
Yeah, then you know that it's a trade off situation, more graphics on the screen vs higher IQ.
I'd love for games to come with a priority option you can set in the settings menu.
I remember playing one of the NFL2K games online. Me and my brother on the same console, taking on all comers.
The first game we played online--ever--was Chu Chu Rocket.
He later replied in the same thread that he wasn't really saying the PS4 GPU was weaker. He was just talking about how to get most of the X-One.
PSU Post 164
IHaveCandy
Banned
Well that's why I was talking about multiplatform games designed to not look significantly worse on Xbox One. For example, something like BF4. If they are making it to have the same assets and interactability on both PS4 and Xbox One, I want those extra 6 CUs on PS4 at least pumping out extra high IQ.
Even though I'm not a techie, I love reading this stuff. Thank you!
What confuses me is how he says it's more efficient at what it's got to work with. Granted, my knowledge of such things is limited, but wouldn't better efficiency mean better performance?
Vestal
Gold Member
I have a feeling that this Gen will be the online everything Gen, with services up the wazooo.
not really happy about that, since I know we are going to get games with online components duck taped on.
ethomaz
Banned
I have no idea where he got the idea that the ESRAM has "zero" latency or that it allows for "free" z-buffer tests.
The other thing he is talking about is that the XBox has tow busses from the GPU to memory, one to the ESRAM and the other one to main memory. He concludes that you can read on one bus and write on the other concurrently. While this may be true, GDDR5 can also issue read and write requests on the same cycle, so I am not sure where the benefit is supposed to be.
NoirVisage
Banned
Ouch..that hurt MY feelings. Sadaharu owns.
This is backed up by what Cerny said at Gamelab 2013:
http://www.youtube.com/watch?v=xHXrBnipHyA&feature=player_detailpage#t=2159s
And here Cerny goes into more detail on why he didn't go with eDRAM + 128-bit GDDR5 vs GDDR5.
http://www.youtube.com/watch?v=xHXrBnipHyA&feature=player_detailpage#t=2159s
His basic idea was the memory system should be "easy to learn but hard to master."
Oh man. I stopped reading GAF mid afternoon on Friday, and come back to this! Insane that there is more drama in this thread about something this trivial (in the scheme of things) comprated to all the Durango/Orbis/Xbox One/PS4 hardware threads so far.
This is nonsense.
There is absolutely no doubt the ESRAM will mitigate the low bandwidth of the main memory. It is only up for discussion whether that ESRAM can offers results comparable to the PS4 and its unified 8GB GDDR5. The use of embedded RAM in itself is not bad hardware design, and 32MB is ample for its designed purpose.
This is nonsense.
There is absolutely no doubt the ESRAM will mitigate the low bandwidth of the main memory. It is only up for discussion whether that ESRAM can offers results comparable to the PS4 and its unified 8GB GDDR5. The use of embedded RAM in itself is not bad hardware design, and 32MB is ample for its designed purpose.
And it will do so for months.
Yes, every memory controller occupies 2x32bit channels of the memory interface independently.
I have enjoyed this. Memories of ps1, dreamcast and blue shadows.
Yeah, Sony knows what they're doing. They're also a hardware company after all, unlike Microsoft.
Exactly, & I just laugh at people that thinks that the PS4 isn't powerful enough when the games that Sony showed off at E3 were ran off of early dev kits & at 4 GB's of GDDR5 RAM.
Dlacy13g
Member
Agreed its been a fun read but at some point (if this happened sometime yesterday then excuse my ignorance) I wish some would have gotten DF to thoroughly spell out what the meant or how they came to their numbers as in were there provided to them without verification, etc...
Suikoguy
I whinny my fervor lowly, for his length is not as great as those of the Hylian war stallions
Some users have tried by tweeting them and the like, and many comments on the article page itself. No answers were given.
I'm surprised Digital Foundry is going to shit up it's reputation over it.
To be fair, they probably tried to get clarification from Microsoft, and their tech guys are furiously working with their PR guys on how best to explain it without explaining a god damn thing.
I know how MS could have got their original bandwidth figures out by 88%.
They probably worked them out in Excel.
They probably worked them out in Excel.
Laughing at my jokes. I'm a big cheese around these parts. You scratch my back, I'll scratch yours.
A recap of this thread:
- DF link for "X1 eSRAM performance increase by 88%"
- 1st couple pages of people saying good news, and others asking if it makes it better than PS4, and even some saying "CBOAT fail" - CBOAT Fail
- Some members finding that the math is wrong and in conclusion = Downclock!
- Same members stating that the DF article is pure PR from MS
- Finding out that DF's source is MS directly (pure speculation imo) due to a Tweet. Tweet Image
- At the same time we had people not knowing the movie Spaceballs............
- Oldergamer kept pulling stuff out of his arse and when asked for proof, he kept refusing and his answer was always: "It's just a simple google search away!". Needless to say it didn't go too well for him. And let's not forget this golden post here:
- Ultimatums to Oldergamer by Bish:
Ultimatum 1
Ultimatum 2
Ultimatum 3
- People also found out that specialguy is Rangers from B3D, not that it means much tbh.
- Meanwhile we had some members rambling about the cloud, cloud size, cloud cost, cloud etc....
- Then some members had a persecution breakdown of some sort and believed all of GAF was against them.
- Afterwards people started arguing about full console specs with some stating that the specs were not finalized or it's all about games (in a tech analysis thread...) and also started comparing PS3 and 360 once again.
- Then we started comparing E3 showings of Infamous SS and Forza 4. With Hawk269 saying he played the game at E3 without showing proof. The problem is that according to many, the game was not playable for anyone. Guess now he has more time for his "special showings". Here is Bish's cease and desist: Send me credentials naoooo!
Yep. This has been a pretty damn good thread so far.
- DF link for "X1 eSRAM performance increase by 88%"
- 1st couple pages of people saying good news, and others asking if it makes it better than PS4, and even some saying "CBOAT fail" - CBOAT Fail
- Some members finding that the math is wrong and in conclusion = Downclock!
- Same members stating that the DF article is pure PR from MS
- Finding out that DF's source is MS directly (pure speculation imo) due to a Tweet. Tweet Image
- At the same time we had people not knowing the movie Spaceballs............
- Oldergamer kept pulling stuff out of his arse and when asked for proof, he kept refusing and his answer was always: "It's just a simple google search away!". Needless to say it didn't go too well for him. And let's not forget this golden post here:
- Ultimatums to Oldergamer by Bish:
Ultimatum 1
Ultimatum 2
Ultimatum 3
- People also found out that specialguy is Rangers from B3D, not that it means much tbh.
- Meanwhile we had some members rambling about the cloud, cloud size, cloud cost, cloud etc....
- Then some members had a persecution breakdown of some sort and believed all of GAF was against them.
- Afterwards people started arguing about full console specs with some stating that the specs were not finalized or it's all about games (in a tech analysis thread...) and also started comparing PS3 and 360 once again.
- Then we started comparing E3 showings of Infamous SS and Forza 4. With Hawk269 saying he played the game at E3 without showing proof. The problem is that according to many, the game was not playable for anyone. Guess now he has more time for his "special showings". Here is Bish's cease and desist: Send me credentials naoooo!
Yep. This has been a pretty damn good thread so far.
Something something due to you laughing at SPE cut the cheese. Hurray!
Vestal
Gold Member
You should have included the 2 ultimatums by Bish. This thread has had everything.
Vestal
Gold Member
Read the article. It refers more to user error than software error.
I would normally shy away from such a massive thread derail, but god damn at this thread already.
That article states that 88% of spreadsheets contain errors, not that Excel is causing the errors.
For one, even assuming my specific example were wrong, I assume you can follow the reasoning to a point of being able to address it.
And secondly, a 100% increase to 5 apples gives you 10 apples. 5 apples is 50 percent of 10 apples. Which was why I wrote
Unknown Soldier
Member
I was not aware that DF had a reputation to shit up in the first place. Richard Leadbetter is well known for being a blatant MS and 360 fanboy and because of this almost all of their cross-platform 360 Versus PS3 game comparisons were unreadable biased trash.
Just finished editing Bish's quotes in it.
Suikoguy
I whinny my fervor lowly, for his length is not as great as those of the Hylian war stallions
I thought they were at least had some clout about spec information... guess not.
shinra-bansho
Member
I'm sorry I don't follow what your example was supposed to show and how it relates to the 88%. Feel free to elaborate?
Yes, 5 apples is 50% of 10 apples. Yes, 5 apples is 50% less than 10 apples. While 10 apples is a 100% increase in apples from 5 apples. We've had this song and dance before around the difference in FLOPS. What is that supposed to show with regard to these numbers?
We know 192 is 88% more than 102. What people are seemingly querying is why is the new figure 192. A doubling, i.e. a 100% increase, would be 204.
What exactly are you trying to say with your example?
EDIT: I see you edited. No 10 apples is not 50% more than 5 apples, when using the original amount as a base. It is 100% more apples than 5 apples. 10 is not 50% more than 5.
- Status