• 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.

Tomshardware - Xbox Series X SSDs Will be Powered by Phison Controller, Report Claims

Pagusas

Member
Jun 9, 2006
11,406
1,847
1,400
Prosper, Tx
Everyone saying this is expenses af needs to remember companies like Sony and MS are litterally buying these things in massive bulk quantities. More so then any other company. They will get sweetheart deals.

these are the type of contracts that allow these controller groups to even exist.
 
Last edited:

SenjutsuSage

Member
Jan 23, 2010
9,456
372
915
The more I think about it the more the Xbox Series X reveal trailer becomes that much more impressive to me. They quite literally teased everything in there, fast switching between games, the fast loading of the game world, the power and visual capabilities. I can't wait to play Halo Infinite on this thing.
 

Panajev2001a

GAF's Pleasant Genius
Jun 7, 2004
14,177
2,491
1,975
Yeah sure buddy, Mark Cerny also said PS4 Pro is 8.4 TF using fp16 and secret sauce.
Ok, the claim he made that “running FP16 code the PS4 Pro GPU processes data twice as fast and if all the code were FP16 it would process it as if it were an 8.4 TFLOPS machine” is not true because?

Considering the audience was a technical site for industry devs, Gamasutra, and the way he conducts himself in presentations and interviews I am quite sure he was not going there in PR bullshitting mode. Not all execs are like that ;).
 
  • LOL
  • Like
Reactions: Gone and Kokoloko85

X-Fighter

Neo Member
Jan 10, 2020
107
115
150
Not to rain on anyone's parade, but if it's expensive or very new it won't be in these consoles. They never use the newest of newest tech because most of the time that's too expensive and doesn't work in console economics. Not even when you buy them in bulk...
 
  • Like
Reactions: existensmaximum

Rightisright

Member
Dec 19, 2019
540
597
290
Wether true or not, the only consistent rumours about SSD’s always put Sony’s solution as faster, not XSX solution as slow. Big difference.
What rumors? There was an article saying that the PS5 SSD was exceptionally fast and might beat Xboxs, however the developer they interviewed was Counterplay who are doing Godfall and don't even have a XSX Dev kit, and nowhere in the interview did they even say the PS5 had a faster SSD.
One thing is for sure, the rumors this gen have contradicted each other. I'm thinking this has been done on purpose to try and hide any real leaks in the mountains of fake ones.
 

LordPezix

Member
Feb 7, 2017
944
615
365
Maybe it’s the Epic moment with Gears of war, they pushed for 512 MB ram, which ended up in the system.
Very well could be. In either case this will be a boon for their system that devs and customers alike will benefit from.
 

Lady Bernkastel

Formerly 'Scathach Eilean'
Mar 8, 2018
1,062
1,395
570
Ok, the claim he made that “running FP16 code the PS4 Pro GPU processes data twice as fast and if all the code were FP16 it would process it as if it were an 8.4 TFLOPS machine” is not true because?

Considering the audience was a technical site for industry devs, Gamasutra, and the way he conducts himself in presentations and interviews I am quite sure he was not going there in PR bullshitting mode. Not all execs are like that ;).
And how effective was it for game devs except for being a fodder for console wars in the internet ?
Same as PS2 being twice as powerful as Dreamcast and "Power of Cell".
 

it_wasn't_me

Member
May 22, 2017
1,083
1,092
375
So, 7 GB is over 100x speed. That contradicts MS statement - 40x improvement.

Lots of assumptions and jumping to conclusions based on a rumor from digitimes on the supplier.

Phison controller can mean a few models, not just the E18. E19, for instance, is only 1.6W and will allow them 3.75GB/s which will fit the "X40 times faster" claim if we look at the Xbox One HDD as a 50-100MB/s drive.
 
Last edited:
  • Like
Reactions: longdi

X-Fighter

Neo Member
Jan 10, 2020
107
115
150
So, 7 GB is over 100x speed. That contradicts MS statement - 40x improvement.

Lots of assumptions and jumping to conclusions based on a rumor from digitimes on the supplier.

Phison controller can mean a few models, not just the E18. E19, for instance, is only 1.6W and will allow them 3.75GB/s which will fit the "X40 times faster" claim if we look at the Xbox One HDD as a 50-100MB/s drive.
Indeed, that would actually make much more sense. It could be the E18, but I think that's way too expensive, it's more likely to be the E19, and maybe if we're lucky the E16.
 

Panajev2001a

GAF's Pleasant Genius
Jun 7, 2004
14,177
2,491
1,975
And how effective was it for game devs except for being a fodder for console wars in the internet ?
Same as PS2 being twice as powerful as Dreamcast and "Power of Cell".
Now you are dodging the question and moving the goalpost by also trying to bring Cell and PS2 into the argument.

FP16 is so ineffective that some devs were using it to save on bandwidth back when it had no advantage processing speed wise, then nVIDIA added it to their shader ALU’s, then AMD added it in Vega, integrated it in PS4 Pro custom GPU, kept it for Navi and successors, and offers it in both XSX and PS5 GPU’s.

Now, back to you explaining what Cerny actually said to a game industry tech website is misleading because...?
 
  • Like
Reactions: thelastword

nikolino840

Member
Dec 30, 2018
1,681
1,130
485
35
So, 7 GB is over 100x speed. That contradicts MS statement - 40x improvement.

Lots of assumptions and jumping to conclusions based on a rumor from digitimes on the supplier.

Phison controller can mean a few models, not just the E18. E19, for instance, is only 1.6W and will allow them 3.75GB/s which will fit the "X40 times faster" claim if we look at the Xbox One HDD as a 50-100MB/s drive.
 
  • Like
Reactions: CatLady

Ar¢tos

Member
Oct 24, 2017
3,007
3,625
645
Are there any other controllers with better Random Read speeds, even if with significant less Sequential read speed ? (or any other pcie4 controllers at all?).
Aren't Random Read speeds much more important for gaming than Sequential? (idk if these things are directly proportional)
 
Last edited:

DoctorEkkoUK

Member
Dec 14, 2014
577
515
550
Faster drive speeds make the visuals “richer?”

I’m not sure I follow.

Is PCIe 3.0 truly not “fast enough” for ultra high resolution textures?
I think it's about texture streaming like that spiderman demo sony released where you can whizz through a huge map at high speed and the new storage solution is able to stream in the large amounts of textures fast enough to keep up.
 

ethomaz

Member
Mar 19, 2013
27,264
7,090
835
37
Brazil
These are two good options:

1) The PCIe 4.0 controller with high speeds but expensive
2) The PCIe 3.0 controller slower but cheaper

Let’s which one MS choose.
 
Aug 28, 2019
907
975
305
No way, at least nvme pcie4 in ps5.

According to leaks, ps5 is faster than fast, so it is even more exciting. Based Mark Sony may grant us his secret reRAM sauce.
I've personally already explored why ReRAM is not going to happen with these systems. No chip has yet to be produced at big capacities (GB range) for commercial purposes, and we're already into 2020. Even once the chip got produced, it'd have to go through a lot of testing and pre-production, which takes months. Fabs would have to be set up, evaluated, certified, and people hired to man the machines and assembly.

In other words, it's not happening. If Sony had a ReRAM announcement, they'd of said it at CES given it's a technology applicable to the broader electronics market. The only ReRAM-like PCM memory on the market right now is Intel's Optane, and while it provides DRAM-style benefits (in the persistent memory variant) for bit/byte-level data alterability, speed-wise it isn't much faster in use than these upcoming behemoth PCIe 4.0 NVMe SSD drives. Chances are very likely ReRAM would follow suit in terms of real-world performance, but they haven't even managed to produce a chip in decent sized-capacity for commercial use yet.

At this point given how similar ReRAM and 3D-Xpoint are in terms of technology it's possible teams are testing taking 3D-Xpoint and altering it into something like ReRAM, but these experiments are still ongoing (if they're happening). So in the end, buying into the ReRAM (or even 3D-Xpoint/Optane) fantasy train with these systems is just that: fantasy. But since these systems will be providing low-level access to their storage memory to act as a cache and memory-map it as well (think AMD's SSG cards), that will be more than good enough for next-gen gaming.

So, 7 GB is over 100x speed. That contradicts MS statement - 40x improvement.

Lots of assumptions and jumping to conclusions based on a rumor from digitimes on the supplier.

Phison controller can mean a few models, not just the E18. E19, for instance, is only 1.6W and will allow them 3.75GB/s which will fit the "X40 times faster" claim if we look at the Xbox One HDD as a 50-100MB/s drive.
Again I ask, why would they gimp a PCIe 4.0 compatible APU with likely between 16-24 PCIe 4.0 lane interfaces, to simply x1 (2GB/s effectively)? You can argue that perhaps in having a lot more CUs they've cut out some PCIe lanes but I can't picture either system's APU supporting less than x12 PCIe 4.0 lane interfaces.

We don't have a lot of information on what state XSX development was in when the comment was made, or if they were even referring to the internal custom SSD storage (that I think both systems will be using hence likely using these type of memory controllers to achieve 7-8GB/s speeds) or (more likely) the connection speed for the secondary user-optional storage drive? Which, since it'd be acting more as cold storage, 2GB/s is sufficient enough.

These are logical conclusions to draw considering the other areas these system manufacturers are pushing their specifications. Even if it's to conserve power draw there is no reason to limit the internal cache drive to 2GB/s, these aren't mobile devices.

Are there any other controllers with better Random Read speeds, even if with significant less Sequential read speed ? (or any other pcie4 controllers at all?).
Aren't Random Read speeds much more important for gaming than Sequential? (idk if these things are directly proportional)
If you want sufficient random access speeds you need a different type of memory technology built for sufficient random access. NAND is not that type of memory. NOR is, but it's also in much lower size capacities, and can get pricey, plus the write speeds are extremely slow compared to NAND (the read speeds are faster however, even faster than SLC NAND).

3D-Xpoint (persistent memory variant) and ReRAM were two options some of us were entertaining...that some of us still are entertaining actually xD....but the latter is nowhere near ready for commercial deployment and the former would still cost about $1-$1.50 per GB even at bulk pricing the likes of Sony or MS could secure, which would limit the total amount in either system to about 64-128GB, versus the 1TB we'll be seeing from both system's (potentially) customized low-level access memory-mapped internal SSDs.
 
Last edited:

Rightisright

Member
Dec 19, 2019
540
597
290
So, 7 GB is over 100x speed. That contradicts MS statement - 40x improvement.

Lots of assumptions and jumping to conclusions based on a rumor from digitimes on the supplier.

Phison controller can mean a few models, not just the E18. E19, for instance, is only 1.6W and will allow them 3.75GB/s which will fit the "X40 times faster" claim if we look at the Xbox One HDD as a 50-100MB/s drive.
Actually, the Xbox One Xs drive is rated at 140MB/s. So 40 x 140MB/s = 5.6gbs.
Not sure where you guys get you info from.
 

Trimesh

Member
Jun 8, 2019
245
278
290
These are two good options:

1) The PCIe 4.0 controller with high speeds but expensive
2) The PCIe 3.0 controller slower but cheaper

Let’s which one MS choose.
The ordinary price considerations go out the window once you start dealing with sufficiently high volumes, though. Unless they are using some licensed IP that they have to pay a per-unit royalty on, there is no real reason that the 4.0 controller should be significantly more expensive than the 3.0 one - and if you give them an EYP of (say) 10-15m units they will want the business even if the margins are small (not least because having recurring orders like that gives you a lot more flexibility in booking fab time).

Obviously, this only works for custom parts that carry no risk of cannibalizing their higher margin business - but that's exactly what you would want for a console anyway and if you have already projected that your higher margin business will be enough to cover the NREs for the development then any extra you can make selling stuff that's not going to impact your primary market is just gravy.
 

it_wasn't_me

Member
May 22, 2017
1,083
1,092
375
Actually, the Xbox One Xs drive is rated at 140MB/s. So 40 x 140MB/s = 5.6gbs.
Not sure where you guys get you info from.
The same from where you got X1X drive speed ( if it is like that ). I've mentioned Xbone, didn't i? Not the mid-gen upgrade.
 
Last edited:

Rightisright

Member
Dec 19, 2019
540
597
290
The same from where you got X1X drive speed ( if it is like that ). I've mentioned Xbone, didn't i? Not the mid-gen upgrade.
No you didn't. You just ran with something to try and make it look like MS was going to have a slower drive than is now being touted.
And in fact the 500 GB Xbox One has a maximum theoretical transfer rate of 145 MB/s, per Samsung's own technical data sheet.
Guess this one didn't work out for you.
 

it_wasn't_me

Member
May 22, 2017
1,083
1,092
375
No you didn't. You just ran with something to try and make it look like MS was going to have a slower drive than is now being touted.
And in fact the 500 GB Xbox One has a maximum theoretical transfer rate of 145 MB/s, per Samsung's own technical data sheet.
Guess this one didn't work out for you.
Yes i did. I've said "Xbox One HDD", Not the Xbox One X HDD ( uh, dumb Xbox One names )

Like you said : Maximum theoretical. Cheers!
 
Last edited:

it_wasn't_me

Member
May 22, 2017
1,083
1,092
375
Lol. That's exactly how things are rated. The maximum theoretical speed of the new SSD is 40 times faster than the old hard drive maximum theoretical speed. Derp.
No you didn't. You just ran with something to try and make it look like MS was going to have a slower drive than is now being touted.
Btw. yes i did. I've said "Xbox One HDD", Not the Xbox One X HDD ( uh, dumb Xbox One names )
 

Dontero

Member
Apr 19, 2018
2,207
2,125
565
What does that mean? I don't understand lol
It means that console will have standard SSD.

Goodbye loading screens.
The biggest joke is that people believe that. Most of games will still have loadings but they will be faster assuming data to load will stay the same (and it won't)

Maybe it’s the Epic moment with Gears of war, they pushed for 512 MB ram, which ended up in the system.
The difference is that extra 256vram made huge difference because it was doubling of very limited memory.
Here you have difference in games about 2% maybe 5% at best. Every SSD has similar random read which is actually what is important for games not sequential read. Even latest SSDs are pretty much the same speed when it comes to games as first SSDs from years ago.

This is why buying NVME drive will give you almost no improvement when you have an sata SSD and you want to play games.

So, 7 GB is over 100x speed. That contradicts MS statement - 40x improvement.
Because MS uses real numbers not sequential read which you won't ever see in practice. Every SSD regardless if completely new or years old has pretty much the same random read speed which is what games use. Going from HDD to SSD is huge jump but going from cheap ssd to the most expensive SSD is basically nothing (for games). That sequential read is only great when you have big data aplications where you need to copy huge files from one drive to another.

Say goodbye to sony having the fastest ssd then :messenger_open_mouth:
It's like race who can quicker rise hand. Makes no difference.


These are two good options:

1) The PCIe 4.0 controller with high speeds but expensive
2) The PCIe 3.0 controller slower but cheaper

Let’s which one MS choose.
Cost of controler even one of the most advanced ones is pretty neglegible.
They will go with PCIe4.0 because ZEN cpu they will be using is pcie4.0.

If we get at least 1GB/s sequential reads I'd be happy.
I don't think you understand what you said.
It is like saying: I will be happy if they install huge pipe near my house even if i won't be using it.
Sequential read is meaningless for games which use mostly random small file read.
 
Last edited:
  • Like
Reactions: recursive

TBiddy

Member
Mar 16, 2015
2,690
746
440
Denmark
Btw. yes i did. I've said "Xbox One HDD", Not the Xbox One X HDD ( uh, dumb Xbox One names )
Instead of talking about names, are we in agreement that you were wrong about the speed of the harddrive in the XB1, and thus the potential 40x increase?
 

Kenpachii

Member
Mar 23, 2018
3,139
2,835
600


So, 7 GB is over 100x speed. That contradicts MS statement - 40x improvement.

Lots of assumptions and jumping to conclusions based on a rumor from digitimes on the supplier.

Phison controller can mean a few models, not just the E18. E19, for instance, is only 1.6W and will allow them 3.75GB/s which will fit the "X40 times faster" claim if we look at the Xbox One HDD as a 50-100MB/s drive.
It's the controller on what its maximal capable off. Doesn't mean its going to have 7gb/s ssd performance.

I also wonder how much lanes there ryzen is going to feature, i cant see 24 lanes being enough to power one of those.
 
Last edited:

it_wasn't_me

Member
May 22, 2017
1,083
1,092
375
Denial is a bad look. Even on an anonymous internet forum.

Disk in XB1 is rated at 145 MB/sec. I’ll let you multiply that with 40 and let everyone know the result.
I'm in denial? Nope and someone need to accept facts, though. Xbox fans have a hard time accepting some facts mostly during this gen.

Like i said, wasn't wrong. Xbone S's HDD ( Seagate st2000lm003 ) has Avg. Sequential Mixed IO Speed 54.1MB/s ( btw, this is a PC testing, maybe in Xbone S is somehow higher, but that's about it ). If the MS want to use a some read/write benchmark for SSD improvement in their XSX, they had to account their weakest console.
 
Last edited:

TBiddy

Member
Mar 16, 2015
2,690
746
440
Denmark
I'm in denial? Nope and someone need to accept facts, though. Xbox fans have a hard time accepting some facts mostly during this gen.

Like i said, wasn't wrong. Xbone S's HDD ( Seagate st2000lm003 ) has Avg. Sequential Mixed IO Speed 54.1MB/s ( btw, this is a PC testing, maybe in Xbone S is somehow higher, but that's about it ). If the MS want to use a some read/write benchmark for SSD improvement in their XSX, they had to account their weakest console.
1. You were talking about XB1, not XB1S. Cant even keep your story straight now?

2. The speed of the disk in the XB1 is rated at 145 MB/sec. Go check it out. After that, multiply the number with 40 and let everyone know what that result is.

Here’s a link with the specs: https://www.cnet.com/products/samsung-spinpoint-m8-st500lm012-hard-drive-500-gb-sata-6gb-s/

So yeah, you were wrong, and now you’re desperately reaching for a lifeline.
 
Aug 28, 2019
907
975
305




It's the controller on what its maximal capable off. Doesn't mean its going to have 7gb/s ssd performance.

I also wonder how much lanes there ryzen is going to feature, i cant see 24 lanes being enough to power one of those.
24 Lanes is well more than enough; each PCIe 4.0 lane is roughly 2GB/s of maximum speed. So an APU with 24 of those has a figurative PCIe bandwidth of ~ 48GB/s. A controller capable of 7GB/s would only need 4 lanes; both XSX and PS4 will have way more than that (maybe not 24, but a lot more than 4).

It means that console will have standard SSD.



The biggest joke is that people believe that. Most of games will still have loadings but they will be faster assuming data to load will stay the same (and it won't)



The difference is that extra 256vram made huge difference because it was doubling of very limited memory.
Here you have difference in games about 2% maybe 5% at best. Every SSD has similar random read which is actually what is important for games not sequential read. Even latest SSDs are pretty much the same speed when it comes to games as first SSDs from years ago.

This is why buying NVME drive will give you almost no improvement when you have an sata SSD and you want to play games.



Because MS uses real numbers not sequential read which you won't ever see in practice. Every SSD regardless if completely new or years old has pretty much the same random read speed which is what games use. Going from HDD to SSD is huge jump but going from cheap ssd to the most expensive SSD is basically nothing (for games). That sequential read is only great when you have big data aplications where you need to copy huge files from one drive to another.



It's like race who can quicker rise hand. Makes no difference.




Cost of controler even one of the most advanced ones is pretty neglegible.
They will go with PCIe4.0 because ZEN cpu they will be using is pcie4.0.



I don't think you understand what you said.
It is like saying: I will be happy if they install huge pipe near my house even if i won't be using it.
Sequential read is meaningless for games which use mostly random small file read.
You aren't factoring in the reality that games have yet to be explicitly designed around SSDs as a storage device; vast majority just brute-force use of them and on PC there is OS abstraction (necessary because of all the possible setups, even if drives are conforming to a standard set of specifications).

Consoles will be using semi-custom SSDs more in line with AMD's SSG cards; the drives will be used as a cache and be memory-mapped by their respective console's OS. That's why they will probably not be removable, but support a secondary, optional drive on a slower interconnect.

Now, I admit I'm speculating on the "dual drives support" thing, but it seems kinda obvious XSX and PS5 are going to be utilizing a setup very similar (and improved upon) from AMD's SSG line. Just...nowhere near as expensive (mainly because of much less memory and using non-SLC or MLC NAND).