So someone explain me this DDR5 RAM PS4 getting

[sauravdn]

So, what does that exactly mean? DDR5 which is far superior than DDR3. Is PS4 really getting that as system memory. I do know they a unified which means the CPU and GPU will share the same 8GB DDR5 ram. What happened to DDR4? And PC's only use DDR3 atm. I thought DDR5/GDDR5 was graphics card exclusive only. And if PS4 does get that kind of ram how much powerful does it make than a PC's ram?

Can someone enlighten me on this topic?
:hmm:

Idk the right place to post this so i put it in here.

 

[middlehead]

GDDR isn't exclusive to video cards by any real necessity, just intent due to cost. GDDR5 is based on DDR3, as was the short-lived GDDR4. DDR4 doesn't officially exist yet.

The short version is that the PS4's GDDR5, with the same amount of memory as the XBox One's DDR3, will have about 2.66x the memory bandwidth.

 

[Pantlegz]

It's GDDR5 and it's not really all that fast compared to DDR3, I believe the clocks may be higher but so is the latency. I believe they're doing this, or able to do this, because of the AMD APU that's powering the console. If you want more information than that, Google it.

 

[mmntech]

Wrong forum, but since you're new we'll let it go.

Anyway, DDR and GDDR are two different types of memory. Though the names are similar, they use different technologies. The latter is primarily used for caching graphics and textures, hence the "G". It's significantly faster than system DDR. Though due to general processing GPU technology, you can also use it as system ram. It offers better performance for the types of tasks a console does. The downside is, it's more expensive.

DDR is CPU ram and it's speed is tied to the processor bus speed. It's slower than GDDR but more suited for general tasks and is quite a bit cheaper. DDR3 is the most common and cheapest type today, offering good performance. DDR4 exists but is not yet mainstream.

In terms of performance, DDR3 puts out 6.4Gbits/s at 800mhz while GDDR5 does about 25Gbits/s at the same clock rate.

 

[Staples]

I don't buy the argument that GDDR5 is comparable to DDR3.

The video cards that have DDR3 memory in them are very very slow to their counterparts with GDDR5.

 

[Aikouka]

I'd like to clarify one thing about the X1 and its DDR3 implementation. The one important thing is that you can't look at it like a PC. PCs use 64-bit lanes to implement a single channel of DDR3, and the specific memory chips exist on the DIMMs (the memory silicon that plugs into the slot). It doesn't matter how many chips are on each DIMM, they all operate at the speed that's set by the memory controller. The X1 appears to use direct 16-bit connections to each individual memory chip (16-bit x 16 chips = 256 bit).

I don't buy the argument that GDDR5 is comparable to DDR3.

The video cards that have DDR3 memory in them are very very slow to their counterparts with GDDR5.

Microsoft's contention is that the 32MB of eSRAM is what makes them comparable (has something like 102GB/s bandwidth), but it might require some trickery to make it work well enough. Supposedly, developers prefer the GDDR method because it's straight-forward and easy to use.

DDR is CPU ram and it's speed is tied to the processor bus speed. It's slower than GDDR but more suited for general tasks and is quite a bit cheaper. DDR3 is the most common and cheapest type today, offering good performance. DDR4 exists but is not yet mainstream.

I'm no computer engineer, but aren't all non-serial devices attached to some sort of clock? Also, given you can overclock your memory without overclocking your CPU, they aren't on the same clock.

 

[nenforcer]

In terms of performance, DDR3 puts out 6.4Gbits/s at 800mhz while GDDR5 does about 25Gbits/s at the same clock rate.

At that speed it wouldn't make it any faster than the XBOX 360 or even the XDR in the PS3!

http://www.rambus.com/us/technology/solutions/xdr/index.html

Don't forget to mention that this GDDR5 has a 256-bit memory interface running at 5500MHz for a total bandwith of 176 GB/s as stated in the AT article!

Thats faster than just about all but the very latest high end video graphics cards, I believe.

 

[tweakboy]

haaha thats GDDR3 thats VRAM video memory,,,,, lol

 

[Smoblikat]

I'd like to clarify one thing about the X1 and its DDR3 implementation. The one important thing is that you can't look at it like a PC. PCs use 64-bit lanes to implement a single channel of DDR3, and the specific memory chips exist on the DIMMs (the memory silicon that plugs into the slot). It doesn't matter how many chips are on each DIMM, they all operate at the speed that's set by the memory controller. The X1 appears to use direct 16-bit connections to each individual memory chip (16-bit x 16 chips = 256 bit).



Microsoft's contention is that the 32MB of eSRAM is what makes them comparable (has something like 102GB/s bandwidth), but it might require some trickery to make it work well enough. Supposedly, developers prefer the GDDR method because it's straight-forward and easy to use.



I'm no computer engineer, but aren't all non-serial devices attached to some sort of clock? Also, given you can overclock your memory without overclocking your CPU, they aren't on the same clock.

They are, im old school so I dont have all the new "hip" lingo......but the DRAM speed is tied to the FSB (BCLK now?). The reason you can OC your CPU independently is because CPU's have more than one way to determine frequency, they use the FSB and multiplier, which is independent of the IMC/FSB. So you can raise your multi and not touch the FSB to get higher clocks.

 

[videoclone]

Dont forget the the PS4 has 1152 core and more rop's VS 768 and less rop's on xbox-1

Before you even touch on the 68.3 GB/s VS 176.0 GB/s memory bandwidth

http://www.anandtech.com/show/6972/xbox-one-hardware-compared-to-playstation-4/2

 

[BrightCandle]

Lets take a look at some benchmarks shall we since SiSoft Sandra can happily get latency and bandwidth figures for both types of memory.

A 3930k (4Ghz currently) with 4x 2133Mhz DDR 3 chips produces the following:
Bandwidth: 49GB/s
Random Access Latency: 22.3ns or 89 clock cycles.

A pair of 680s (stock):
Bandwidth: 290 GB/s
Random Access Latency: 205.2ns

Its a classic trade off. On the one hand GDDR5 provides much more bandwidth, it does so by doubling the effective data transferred per clock for data , allowing reads and writes at the same time and it can be combined into very wide buses. But one of the consequences is that the latency is dramatically impacted, it can read and write vast amounts of data but it doesn't respond quickly to random access.

Conversely DDR3 has much less bandwidth, but it delivers responses considerably quicker. Its latency is 1/10th that of GDDR5. The nature of GDDR5 is that it matches its usage. A GPU is a massively parallel device, it readis memory in sequential chunks and if one of its processors is stuck awaiting on memory it doesn't matter much. Whereas a CPU is very random in its access of memory and needs to hide as much of the latency as it can or its performance will plummet. That is the memories are designed around the different usage patterns that a GPU and CPU see.

GDDR5 is going to be very bad for a CPU. We already know for example that on a 3930k if we drop the memory from 2133Mhz to 1066Mhz it makes basically no performance difference at all. CPUs are not really impacted that much by bandwidth these days except in particular circumstances. Equally minor changes in latency don't impact them either, but I have no doubt if I could set the memory to take 10 times as long that it would likely halve or worse the performance of my CPU. GDDR5 is not made for CPUs and its certainly not universally better, for a CPU the bandwidth doesn't matter much but the latency really does.

 

[poohbear]

Lets take a look at some benchmarks shall we since SiSoft Sandra can happily get latency and bandwidth figures for both types of memory.

A 3930k (4Ghz currently) with 4x 2133Mhz DDR 3 chips produces the following:
Bandwidth: 49GB/s
Random Access Latency: 22.3ns or 89 clock cycles.

A pair of 680s (stock):
Bandwidth: 290 GB/s
Random Access Latency: 205.2ns

Its a classic trade off. On the one hand GDDR5 provides much more bandwidth, it does so by doubling the effective data transferred per clock for data , allowing reads and writes at the same time and it can be combined into very wide buses. But one of the consequences is that the latency is dramatically impacted, it can read and write vast amounts of data but it doesn't respond quickly to random access.

Conversely DDR3 has much less bandwidth, but it delivers responses considerably quicker. Its latency is 1/10th that of GDDR5. The nature of GDDR5 is that it matches its usage. A GPU is a massively parallel device, it readis memory in sequential chunks and if one of its processors is stuck awaiting on memory it doesn't matter much. Whereas a CPU is very random in its access of memory and needs to hide as much of the latency as it can or its performance will plummet. That is the memories are designed around the different usage patterns that a GPU and CPU see.

GDDR5 is going to be very bad for a CPU. We already know for example that on a 3930k if we drop the memory from 2133Mhz to 1066Mhz it makes basically no performance difference at all. CPUs are not really impacted that much by bandwidth these days except in particular circumstances. Equally minor changes in latency don't impact them either, but I have no doubt if I could set the memory to take 10 times as long that it would likely halve or worse the performance of my CPU. GDDR5 is not made for CPUs and its certainly not universally better, for a CPU the bandwidth doesn't matter much but the latency really does.

U know brightcandle ive askes the same question numerous times & alot of ppl tried answering it without really knowing what the heck they were talkijg about to the point i stopped asking the question, but ur explanation cleared it all up with this post. Thank you, u explained it clearly & simply speaking of its use in a context. Respect!

 

[videoclone]

Until you attach the memory to that CPU we don't know the latency numbers. I thought I could measure it with SiSoft sandra, but it is measuring CPU to GPU memory latency.

So instead I went digging through the datasheets from Hynix and such and found that at 6.0Gbps frequency the following (with data bus inversion off, but it likely needs to be on and then its normally +1 these figures):

GDDR5 timings as provided by Hynix datasheet:
CAS latency= 10.6ns
tRCD = 12ns
tRP = 12ns
tRAS = 28 ns
tRC = 40ns


DDR3 timings for some Corsair 2133 RAM 11-11-11-28
CAS 10.3ns
tRCD 10.3ns
tRP 10.3ns
tRAS 26.2ns


So its up to about 16% higher latency on the raw access timing in measures people are used to with DDR3. The challenge is that the command rate is at half the frequency, so I doubt it translates like this at all.

In converted timings its impressively high because the base clock speed achieved is actually 1500 Mhz so the GDDR5 uses something like 16-18-18-42.

That is enough of a change in timing to have an impact. Some with an AMD CPU llano/jaguar could tell us roughly how much impact it will have. I'll ask in the CPU forum.

The latency thing isn't actually true. Typical GDDR5 memory controllers have high latency, but the modules themselves don't necessarily. Most of what I've seen indicates the opposite; that GDDR5 can have extremely low latency vs DDR3, but building a memory controller to take advantage of that is very difficult. Also, GDDR5 is it's own technological evolution, and not a byproduct of DDR3. GDDR5 is actually QDR ref clock as apposed to DDR.

Overall the GDDR5 CPU memory latency would be very little but GPU Memory performance would be astronomical VS DDR3 as you can tell from any modern video cards when equipped with DDR3 VS GDDR5 the performance is over 100% .. That's what we are looking at with Xbox1 vs PS4