Xbox One in trouble.

[Garson007]

Xbox One: Bluray player, media player and games console + can control my decoder (if DSTV allows it)

Is the DSTV decoder HDMI-CEC compatible? If it is they should allow channel switching.

 

[Roman4604]

Is the DSTV decoder HDMI-CEC compatible?

Not from what I've seen.

If it is they should allow channel switching.

Think the XBO comes with an IR blaster. Hopefully MC will sign on as a partner.

 

[Garson007]

Think the XBO comes with an IR blaster. Hopefully MC will sign on as a partner.

I'm not so sure about that. I think it mostly relies on CEC set top boxes, from what I've read anyways.

 

[Ancalagon]

Why does a CPU and a GPU have a cache? Latency. The CPU and GPU will be sharing these caches on HSA hardware. Increasing the size of these caches will improve the performance of low-bandwidth latency intensive tasks, which do exist - especially under any sort of GPU-CPU data sharing arrangement where memory will constantly have to be offloaded off of the registers and into the cache.

Both APUs have will on die L1 and L2 cache in any case. Shared memory in this case is only necessary once a particular job is finished and information needs to be passed back.

But still, you cant define A) how "bad" the latency really is on GDDR5, B) what the maximum allowable latency for any particular task is. You know the latency is better, but you dont know how that translates into real world performance. You have no idea.

The RAM latency would have to be reported by the manufacturer. So I'm not sure what either is. What I do know is that eSRAM is built for low latency and GDDR5 is built for high bandwidth.

So you dont really know what the difference is - one could have 5% better real world latency than the other. You have no idea.

I am not.

Then why are you the only person on the planet who feels the consoles are evenly matched?

You're assuming that the memory calls will happen in bursts. I'm specifically talking about situations where the bandwidth does not matter and latency is more important. So the 3,4 or whatever magnitude of latency is huge (especially if it's even higher).


Latency will be too much for certain tasks. Specifically when those tasks are repeatedly going back and forth xxxx times during a frame.

But you still dont know how much latency is too much or how bad the latency is on GDDR5. Doesnt matter whether bandwidth matters, it matters whether the latency is good enough! As I said, doesnt matter if your latency is 0.0000000000001 if 0.0000001 will do.

It works for graphics and other bandwidth intensive computations. Graphics is not latency sensitive. That's a bull**** call, that is. Every frame generated does not need continuous calls back to memory to generate, so almost any latency faster than a frame is more than good enough.

See now I know you are trolling because clearly you have no idea what you are talking about. Do you even code?

Latency matters for graphics, because during frame rendering, the GPU has to make several calls to its own memory for texture reads. If your latency on texture reads is huge, guess what? Frame time goes up!

To use the simplest possible example. To say that graphics is not latency sensitive shows you have absolutely no idea how rendering works at all.

 

[prod]

can you two girls stop flirting already?

Neither console has even been released, wait for the games, gfx, and user experience...then decide.

 

[Garson007]

Both APUs have will on die L1 and L2 cache in any case. Shared memory in this case is only necessary once a particular job is finished and information needs to be passed back.

Incorrect, many jobs will not fit the cache sizes. The sharing of the memory will likely have to happen multiple times before a final answer is given.

But still, you cant define A) how "bad" the latency really is on GDDR5, B) what the maximum allowable latency for any particular task is. You know the latency is better, but you dont know how that translates into real world performance. You have no idea.

A) It will be worse, which for all intents and purposes are more than enough. B) It depends on the task, that's the entire point of giving developers more options. They might feel like implementing a new algorithm and it will either work within the tolerances allowed or it won't. Again, if you were to calculate on the GPU portion and use the logic of the CPU portion then you will require low latency to pass between the two constantly.

So you dont really know what the difference is - one could have 5% better real world latency than the other. You have no idea.

5%? Now you're joking. It's going to be a couple of 100% in the least. It's true that I don't hold this information, but there is a latency reason for the existence of embedded RAM and a latency reason why CPUs do not use GDDR RAM.

Then why are you the only person on the planet who feels the consoles are evenly matched?

They are not evenly matched. How often do I have to say that? They are different. Given generic use of resources, as we do today in games, it is quite an obvious conclusion to make that PS4 will handle it with more ease. That is not what I'm arguing.

But you still dont know how much latency is too much or how bad the latency is on GDDR5. Doesnt matter whether bandwidth matters, it matters whether the latency is good enough! As I said, doesnt matter if your latency is 0.0000000000001 if 0.0000001 will do.

Again, it restricts the amount of calculations you can pass back and forth before the game is no longer rendered in pseudo real time tolerance levels.

See now I know you are trolling because clearly you have no idea what you are talking about. Do you even code?

I work with numerical simulations on a daily basis. Working mostly with Fortran and C.

Latency matters for graphics, because during frame rendering, the GPU has to make several calls to its own memory for texture reads. If your latency on texture reads is huge, guess what? Frame time goes up!

To use the simplest possible example. To say that graphics is not latency sensitive shows you have absolutely no idea how rendering works at all.

Graphics is not latency sensitive. Latency affects graphics, sure, but the bandwidth is of far more importance than latency - hence the use of high bandwidth high latency memory. It does not have to do the amount of memory calls that other calculations have to do.

 

[Mike Hoxbig]

Getting bored of this geek talk, please stop guys

 

[Ancalagon]

Graphics is not latency sensitive. Latency affects graphics, sure, but the bandwidth is of far more importance than latency - hence the use of high bandwidth high latency memory. It does not have to do the amount of memory calls that other calculations have to do.

I'm a game programmer - as in 3d videogames, much like those which run on both consoles - and I used to work in the real time broadcast graphics industry. Trust me when I say graphics is a latency sensitive operation.

It doesnt matter if your graphics memory has 1000000GB/s of bandwidth if the latency will push frame rendering times to 100ms on decent hardware.

 

[Garson007]

I'm a game programmer - as in 3d videogames, much like those which run on both consoles - and I used to work in the real time broadcast graphics industry. Trust me when I say graphics is a latency sensitive operation.

It doesnt matter if your graphics memory has 1000000GB/s of bandwidth if the latency will push frame rendering times to 100ms on decent hardware.

I think we're miscommunicating. When I say high latency, I do not mean 100ms. Neither do I mean USB latency (whatever that is). I'm talking about the latency at the hardware level. It's high-latency in comparison to other random access memory technologies. That doesn't mean that it's suddenly not super fast, it's just slower in comparison. I'm sorry if I came across as insulting. I don't really know much at all about graphics processing, but I do notice the opportunity for faster real time calculations when I see it.

 

[prod]

i like video games.

 

[Swa]

True. But at $100 less, not forcing Kinect on you and having not shown intent to use DRM they are the lesser risk.

Probably better to wait till mid next year to see how this plays out.

It's all about publishers intent and if they have DRM for the XB1 they will likely have it if the PS4 supports it.

$100 difference will probably be about R1000 in SA - R1000 is nearly two games. And you get a weaker console if you pay the extra cash.

It would depend on the free* games. If XB1 comes bundled with 2 more it may likely outsell the PS4. It's about two different markets here, the XB1 for entertainment and the PS4 for gaming only. Most people will only hear the entertainment part. Sony made the wrong ad with something that MS could always have changed.

Yes, DDR3 has marginally lower latency than GDDR5 - however, the bandwidth of GDDR5 is triple that of DDR3. That is why it is used in graphics cards! If graphics cards are not sensitive to the latencies of DDR3, why would a gaming console that is doing mostly graphics work?

DDR3 has significantly lower latency. Don't downplay the latency of GDDR5 to make it keep up with DDR3. The bandwidth of GDDR5 is only double that of dual-channel DDR3.

Why would GDDR5 cripple in it when that is exactly what ALL current existing GPGPUs cards, such as the Tesla, use?

Are you affiliated with MS in any way?

Because they only do graphics work or parallel computing. Loading a single 32-bit integer from 128-bit memory is already a waste. That's why triple-channel memory never gave Intel an advantage. Loading anything below 128-bit data structures will not give any performance gain and will take a hit on the latency. That's practically every read for CPU work. Do you honestly think a gaming console or pc does minimal CPU work?

Since the XB1 has two memory pools, one CPU exclusive and one shared (GDDR5?), it can run both types of tasks at the same time without one impacting the other. This is where the XB1 can be an all round entertainment system without taking a performance hit. If the PS4 did the same it would tank the gaming.

More complexity = more things that can go wrong.

Complexity equates to interworking of parts. With ram dedicated to different tasks and keeping the two os's isolated as much as possible there is actually less complexity.

Which announcement was better received? Sony's or MS's? Do I really need to dig up news articles for this one?

http://www.forbes.com/sites/davidthier/2013/05/21/first-round-of-xbox-one-vs-ps4-goes-to-microsoft/

See, this is the problem. It is a gaming console. It should do that well first, and then accommodate everything else. A higher price tag might then be justified.

You can't just tell the core market "well if you want better games then this isn't for you".

That's why MS is trying to initiate a shift where the core market is no longer just about gaming. It's a gamble and like Dreamcast it may not pay off. But PS3 and Xbox now has everything Dreamcast introduced where their previous consoles didn't. We'll see if the market is ready for it but I think they should be commended for their initiative. You also seem to think that the dedicated gaming market is the only one to cater for.

First, how is latency calculated? Latency is calculated in the number of memory cycles it takes for a load or store to complete.

There you're wrong. Latency is in nanoseconds. The CL values are a factor of the clockspeed and latency. If it takes 8 cycles and you double the clockspeed it will take 16 cycles.

And latency only matters if its too much. It doesnt help to have latency that is 0.0000000000s if you can tolerate up to 0.00001s. Its only if latency goes past your threshold that players will notice, because the time taken to generate a frame goes up, which the user will see as input lag.

No, latency is also a matter of which tasks you perform. You keep treating it like the GPU is the only factor. If that was the case they could couple a high end graphics card with a 486. The higher latency will be seen on CPU intensive tasks. Some games will be faster on the PS4 and some will be faster on the XB1. Different strokes and all that...

If the GDDR5 latency was flipping terrible, we would have an issue. But we know its not terrible, and you know how we know that? Because GDDR5 is used in graphics cards, which is the most latency sensitive part of rendering. They even use GDDR5 for professional broadcasting (industry I used to work in). If the latencies are good enough for broadcast, its good enough for games.

You can't make such comparisons. Graphics + high bandwidth = able to accommodate much higher latency. CPU + low bandwidth = needs low latency. Guess which one broadcasting falls into.

 

[Garson007]

Since the XB1 has two memory pools, one CPU exclusive and one shared (GDDR5?), it can run both types of tasks at the same time without one impacting the other. This is where the XB1 can be an all round entertainment system without taking a performance hit. If the PS4 did the same it would tank the gaming.

Uh, mmm... XBO has the embedded eSRAM and DDR3, if reports are correct. I'm not sure where your information comes from that the eSRAM will be restricted to the CPU, especially given that MS chose an HSA capable platform (and therefore the whole memory sharing thing doesn't work like it did before; the sharing is dynamic and both can address all memory space).

 

[Swa]

Uh, mmm... XBO has the embedded eSRAM and DDR3, if reports are correct. I'm not sure where your information comes from that the eSRAM will be restricted to the CPU, especially given that MS chose an HSA capable platform (and therefore the whole memory sharing thing doesn't work like it did before; the sharing is dynamic and both can address all memory space).

I'm not talking about eSRAM. MS stated that 3GB of the 8GB main memory will be dedicated to the system and CPU. What is not known is if the other 5GB will be GDDR5 and if the 3GB will be available to developers not running applications in the OS space.

If it is then the options are better. If it's GDDR5 then the conversation is moot anyway and if not then XB1 will excel at CPU intensive games while the PS4 will excel at Gfx intensive games.

eSRAM will have minimal impact on none repetitive use of data and doesn't get my interest. Where the biggest difference will be seen is in the latency and bandwidth of DDR3 vs GDDR5.

 

[B3ns]

Xboxone has 8 gigs of ddr3. 3 gigs to run the os. 5 gigs available to developers. The os will always have the 3 gigs, it's not available for anything else.

 

[Ancalagon]

It would depend on the free* games. If XB1 comes bundled with 2 more it may likely outsell the PS4. It's about two different markets here, the XB1 for entertainment and the PS4 for gaming only. Most people will only hear the entertainment part. Sony made the wrong ad with something that MS could always have changed.

If MS was going to bundle free games with the XBO, they would be doing it in the States already. Remember how low down we are on MS's priority list.

DDR3 has significantly lower latency. Don't downplay the latency of GDDR5 to make it keep up with DDR3. The bandwidth of GDDR5 is only double that of dual-channel DDR3.

68 Gb/s vs 176 GB/s. In this particular case, just under triple.

Because they only do graphics work or parallel computing. Loading a single 32-bit integer from 128-bit memory is already a waste. That's why triple-channel memory never gave Intel an advantage. Loading anything below 128-bit data structures will not give any performance gain and will take a hit on the latency. That's practically every read for CPU work. Do you honestly think a gaming console or pc does minimal CPU work?

Where did I ever say that? In any case, if a GPU is used for GPGPU tasks - and make no mistake, that will happen with both consoles - it will likely be reading in vectors in any case, which will be either 128 or 256 bit depending upon precision. So, the GPU wont be reading a single 32 bit signed int unless its a constant.

Since the XB1 has two memory pools, one CPU exclusive and one shared (GDDR5?), it can run both types of tasks at the same time without one impacting the other. This is where the XB1 can be an all round entertainment system without taking a performance hit. If the PS4 did the same it would tank the gaming.

Wrong, system memory is shared - the DDR3 pool is also the graphics card memory. In case you havent noticed, DDR3 is only used on low end graphics cards, which should tell you something.

Do some more research on the architecture of the XBO and then maybe you can debate with us here.

That's why MS is trying to initiate a shift where the core market is no longer just about gaming. It's a gamble and like Dreamcast it may not pay off. But PS3 and Xbox now has everything Dreamcast introduced where their previous consoles didn't. We'll see if the market is ready for it but I think they should be commended for their initiative. You also seem to think that the dedicated gaming market is the only one to cater for.

Which sold the most - the Wii, the Xbox360, or the PS3? The Wii. Which console was the most capable? The PS3 (counting the fact it has a Blu ray player built in).

That should tell you something. Yes, the market has changed, but sometimes, gamers just want something that will let them play their favourite games.

There you're wrong. Latency is in nanoseconds. The CL values are a factor of the clockspeed and latency. If it takes 8 cycles and you double the clockspeed it will take 16 cycles.

Nope, wrong. Latency is measured in nanoseconds (or milliseconds in the case of networks), but it depends upon the number of cycles and how long those cycles each take. If you double the clockspeed, it is still 8 cycles, its just that those 8 cycles are half the length. Garson, you know your stuff, explain this to them.

No, latency is also a matter of which tasks you perform. You keep treating it like the GPU is the only factor. If that was the case they could couple a high end graphics card with a 486. The higher latency will be seen on CPU intensive tasks. Some games will be faster on the PS4 and some will be faster on the XB1. Different strokes and all that...

The CPU is the same on both but the PS4 has the more powerful GPU. Its really a no brainer.

You can't make such comparisons. Graphics + high bandwidth = able to accommodate much higher latency. CPU + low bandwidth = needs low latency. Guess which one broadcasting falls into.

Its actually sensitive to both, and everything.

 

[saffakanera]

I'm just terribly relieved MS came to their senses and can now provide a healthy competition, PS4 only would have been boring! Competition breeds strong 1st party titles!

 

[Swa]

If MS was going to bundle free games with the XBO, they would be doing it in the States already. Remember how low down we are on MS's priority list.

I don't see it being sold in the States. Pre-order yes but details (including hardware) is still sketchy.

68 Gb/s vs 176 GB/s. In this particular case, just under triple.

So you're referring to the transfer rate. I was referring to the data bus.

Let's look at the figures. DDR3-2133 has a transfer rate of 17GB/s (136.5Gb/s) per module. For dual-channel it's 34.1GB/s (273Gb/s). Sony is claiming 22GB/s (176Gb/s) on a 1375MHz clock. MS claims 68.3Gb/s which does not square with the rate for DDR3. It's the rate for a 32-bit bus which is only half a module actually and 1/4th the AMD data bus.

The actual thing that makes GDDR5 special is its ability to read and write at the same time on a split bus. Something that is great for the type of data gfx use. Not so much for random integer reads/writes. Assuming the CPU is only working with single integers at a time the 68.3Gb/s figure is pretty accurate. Since Sony is using a different metric it's not nearly triple. They are using a higher clockspeed so it's about 1.29 times faster or 88.1Gb/s.

BUT raw memory bandwidth is a useless figure. The latency of DDR3-2133 is 11-14. GDDR5 sacrifices latency and apparently lots of it. I don't know the actual figures but have seen some in the region of 8-10 cycles. At 1375MHz it can only afford 4 extra cycles to match the highest latency DDR3-2133.

Where did I ever say that? In any case, if a GPU is used for GPGPU tasks - and make no mistake, that will happen with both consoles - it will likely be reading in vectors in any case, which will be either 128 or 256 bit depending upon precision. So, the GPU wont be reading a single 32 bit signed int unless its a constant.

You seem to be implying it by continually referring only to the benefit GDDR5 will have for gfx but ignoring the impact it will have on the CPU.

Wrong, system memory is shared - the DDR3 pool is also the graphics card memory. In case you havent noticed, DDR3 is only used on low end graphics cards, which should tell you something.

Do some more research on the architecture of the XBO and then maybe you can debate with us here.

5GB non-system memory is shared. 3GB is reserved for system (usually CPU work) exclusively. In case YOU haven't noticed AMD has stated the PS4 architecture would not come to pc's or laptops because GDDR5 is terrible for CPU intensive work.

Which sold the most - the Wii, the Xbox360, or the PS3? The Wii. Which console was the most capable? The PS3 (counting the fact it has a Blu ray player built in).

That should tell you something. Yes, the market has changed, but sometimes, gamers just want something that will let them play their favourite games.

It tells me sales are more related to marketing. The Wii had something the other two didn't so the people that bought it saw it as the most capable. Yes hardcore gamers just want something that will be good at playing games. There's however another market that's left out with this mindset. It's too early to speculate.

Nope, wrong. Latency is measured in nanoseconds (or milliseconds in the case of networks), but it depends upon the number of cycles and how long those cycles each take. If you double the clockspeed, it is still 8 cycles, its just that those 8 cycles are half the length. Garson, you know your stuff, explain this to them.

Let me explain this to you. Latency depends on time. Take 10ms (0.01s) latency with a clock of 1000Hz. It will take 10 cycles to make up that latency. Latency measured in time units stay the same so if you double the clockspeed to 2000Hz it will take 20 cycles. If you don't believe that take a look at the latency figures for DDR3 and think why that is the case. Sometimes if you overclock you can trim it closer because latency doesn't make up full clock cycles. So back to the example if latency is actually 9.4ms it will be 10 cycles for 1000Hz because each cycle is 1ms. For 2000Hz it will be 19 cycles because each cycle is 0.5ms.

The CPU is the same on both but the PS4 has the more powerful GPU. Its really a no brainer.

It's not so simple. The CPU is the same but on the PS4 it will be constrained by the higher latency. You keep ignoring the impact of the CPU and only take the GPU into account. The XB1 also has the eSRAM that will improve latency for repetitive memory access so for those cases it may in fact be more powerful. Which one will be better at games? It depends on whether it's CPU (XBox) or GPU (PS) intensive. I don't know why this logic needs to be rehashed.

 

[Chevron]

I'm sorry swa, but you're the last person I'd go to looking for actual facts. You do have "faith" though. So maybe stick to PD.

 

[Swa]

I'm sorry swa, but you're the last person I'd go to looking for actual facts. You do have "faith" though. So maybe stick to PD.

So you don't have any facts to refute it? Well done in derailment.

 

[B3ns]

All this tech talk, just read this link

http://www.anandtech.com/show/6972/xbox-one-hardware-compared-to-playstation-4

The hardware professionals explain it exactly how it is. All the facts are there.