Given that GDDR5 is much faster than DDR4 and DDR3 RAM, why aren't we using it as a general system memory?

Given that GDDR5 is much faster than DDR4 and DDR3 RAM, why aren't we using it as a general system memory?

Attached: 1687218364283.jpg (770x660, 33K)

Other urls found in this thread:

gamersnexus.net/guides/2826-differences-between-ddr4-and-gddr5
twitter.com/NSFWRedditGif

lattency

>GDDR5 is much faster than DDR4
it's not

Because GDDR6 already exists

gamersnexus.net/guides/2826-differences-between-ddr4-and-gddr5

Read up

>Why can't a GPU be used as a CPU?

>he isn't using GDDR9 yet

Attached: ishygddt.jpg (815x666, 329K)

GDDR5 is not part of the GPU
Plus a graphics card could run Linux if you got the documents from the manufacturer and implemented a new arch

So basically the latency doesn't matter for GPUs because they run at a lower clock?

latency matters for anything, but it's compensated for in GPUs by having thousands of processing cores. CPUs, on the other hand, have far fewer cores.

basically, yeah. GPUs have slow clocks but a great number of cores, which means bandwidth is more valuable than latency. CPUs are the other way around.

because RAM designed to be soldered to the same PCB like 2 cm away from the processor will never reach the same clocks going through two socket interfaces over a longer distance and with shared address/control lines.

I use core2duo and 512 ddr2

Nothing preventing soldering gddr5 onto the motherboard. I'm not suggesting to use the graphics card memory.

Based. DDR2 systems are /comfy/

Attached: 1538706649824.png (600x500, 42K)

The PS4, PS4 Pro, and Xbox One X uses GDDR5 for system memory.

even socketed CPU would probably prevent GDDR5 from reaching anything close to nominal frequencies.
also, it's fucking retarded to have one-size-fits all for host memory -and- no ability to transfer mem to new mobos, etc.
if anything you would want HBM on-package as L4 and downclocked DDR4/5 on the motherboard for high capacity at lower power.

considering GDDR5's higher bandwidth, would't the slower access time be compensated by the ability to fill the CPU's cache relatively faster?
Or are cache misses too common to make it worth it?

And why is the GDDR on my gpu clocked to like 3500 mgz while my DDR4 is 2400? Wouldn't that give me less latency rather than more? I thought the benefit for GDDR was the bus width rather than the frequency, now I'm confused. Or is that just the chips frequency needed to fill the bus byte by byte, and the bus runs at another, lower frequency?

Can they run Arch though?

And coincidentally, they run games with far more FPS on-paper-specs-wise than PC.

evidently, yes.

Latency, but also price and availability, video cards were on GDDR5 well before ddr4 ram became a thing.

Theoretically, video card memory could be accessed like regular ram, like through the PCI bus. Technically it's possible. Stable? fuck no, better to just buy more ram and cede the theoretical capacity to purpose-built components.

a pointless 1-2% overall gain i'm sure