New Samsung DRAM Boasts of 12.8GB/s Transfers

Toggle sidebar Toggle sidebar
Status
Not open for further replies.
8 times faster and 87% less power consumption? The mobile industry does look like the driver for development that the PC industry was a few years ago. Not really surprising when we know that the mobile industry ships over a billion units every year while the PC industry ships about a fifth of that.
 
And what cost is associated with connecting 512 lines for communication. Is this achievable in other sitiations like package on package?
 
[citation][nom]rantoc[/nom]Hope its GB and not Gb ? In that case only 1.6GB/sec. Typos have happened before so might as well ask![/citation]

It is right, although my mind does become befuddled when I think what practical uses a 12.8GB/sec transfer on a phone might be. Any suggestions?
 
[citation][nom]ImagineTek[/nom]It is right, although my mind does become befuddled when I think what practical uses a 12.8GB/sec transfer on a phone might be. Any suggestions?[/citation]

so my girlfriend can play the new angry birds... in all seriousness that sounds like an insane amount of bandwidth for a smartphone.Maybe new mobile GPUs will need more bandwidth when they are used in tablets with the larger resolution
 
[citation][nom]ImagineTek[/nom]It is right, although my mind does become befuddled when I think what practical uses a 12.8GB/sec transfer on a phone might be. Any suggestions?[/citation]

Considering most phone 3d "gpu"'s (not that i really consider a phones gpu worthy of the name gpu but that don't belong here) lacks dedicated memory it would soon become a bottleneck when for instance outputting 3d on a higher res external display. Other that that it seems like overkill for a phone when their cpu's have so limited I/O as well as computational power.
 
[citation][nom]mister g[/nom]I'm hoping they're putting as much effort into developing DDR4 desktop RAM, GPUs already have GDDR5 why can't we have DDR4?[/citation]

Most computer systems beside those basic ones that still uses normal memory for 3d graphics would gain little benefit from much faster ram, not enough to make it worthwhile for the big corporations to spend loads of green into the r&d.

The x86 prefetchers for instance do a great job at mitigating that otherwise bottleneck. Try use your mem at 1/2 speed and test some real world scenarios and see the result! Its amazing whet clever engineering can do!
 
They will probably control the release like most tech companies do, in order to get the most out of the speed jump... First release 4 GB, then 8 GB, then finally to 12 GB over 2 years.
 
Perhaps the RAM has a lower power output, but it has to be compensated by the controller, needing to control almost 15x more pins means that the controller also needs to be 15x larger.
 
[citation][nom]ImagineTek[/nom]It is right, although my mind does become befuddled when I think what practical uses a 12.8GB/sec transfer on a phone might be. Any suggestions?[/citation]

I don't about you, but I sure as hell don't mind EXTRA speed. Would you mind gaming with a 6-core i7 980x at 4.5GHz? No game is going to take full advantage of that chip, but you certainly wouldn't complain about it. 😉

That being said, I agree with saturnus that the mobile industry is really stepping its game up. Hopefully we can merge some of its technologies over here on the desktop side. Not that I don't love the idea of having a beastly tower under my desktop. I would just enjoy a little downsizing.

I feel like with the technology we have in phones there is NO reason our powerhouse desktops should be so HUGE and consume so much power. These tiny little handheld smartphones have insanely low power consumption for their performance. Not to mention both power consumption/performance is improving at an astonishing rate. Yet we still have people packing 1000W PSU's in the FULL SIZE towers, granted its for tri-sli for the most powerful GPU's on the planet, but it still blows my mind how these fastly improving smartphones consume a fraction of the power, are completely FANLESS, all in a tiny form factor.

I'd say it's about time we start downsize our desktops. SSD's are a great step towards this, extremely high performance(and price 😀), yet low power and barely generate any heat when compared to mechanical hard drives. CPU's are taking a step in the right direction as well, although notebook CPU's are way ahead in terms of power consumption/heat. Now all we need is lower the power consumption/heat on these high performance GPU's, which quite frankly is ridiculous. We have mobile gaming notebooks capable of maxing out almost any game(at 1080p) that barely use more than 150W on full load. There is absolutely NO REASON we should NEED more than 500W for a desktop gaming system. (Maybe for QUAD SLI to play 2560x1600)

Am I crazy to think this? I don't know, it just seems like gaming PC's have been stagnant in the last 5-10 years in terms of size/power consumption/heat. Hardware has definitely been improving, but with such great improvements in efficiency these large towers should be a thing of the past.
 
[citation][nom]mister g[/nom]I'm hoping they're putting as much effort into developing DDR4 desktop RAM, GPUs already have GDDR5 why can't we have DDR4?[/citation]
this is because the cpu r still stuck on the crappy old school x86
 
[citation][nom]ProDigit10[/nom]Perhaps the RAM has a lower power output, but it has to be compensated by the controller, needing to control almost 15x more pins means that the controller also needs to be 15x larger.[/citation]


It certainly will add some space for traces on the board, but the controller shouldn't need to be that much larger to add more traces. Most mobile devices had been moving towards SOC style things, so future mobile cpus might need to be designed for this.

On tablets where this might be geared for, it won't be too big of an issue. Smartphones might have trouble, but I wager this isn't really designed for small smartphones.
 
[citation][nom]rantoc[/nom]Most computer systems beside those basic ones that still uses normal memory for 3d graphics would gain little benefit from much faster ram, not enough to make it worthwhile for the big corporations to spend loads of green into the r&d.The x86 prefetchers for instance do a great job at mitigating that otherwise bottleneck. Try use your mem at 1/2 speed and test some real world scenarios and see the result! Its amazing whet clever engineering can do![/citation]

You're confused, badly. Prefetchers aren't part of programming, it's part of hardware. They have been around since 8086, but what makes memory less important is caches, which make reads to main memory relatively infrequent. Also, a cache always reads in a line, so in a sense that's prefetching, but also, since the Tualatin, the processors prefetch lines as well.

With more cores per processor, bandwidth becomes more important as well. If you're one processor and you've got say seven more, you don't want to be hogging the memory bus. If you are, then they're all waiting. Faster bandwidth means you get release it sooner, and let another core have it, if necessary. It also allows for larger cache lines, if that's desirable in the situation, without unduly increasing the time it takes to read it in.
 
doesn't anyone understand the difference between 8 times and 8 fold?
grab a piece of paper, fold it 8 times, see how many you get. Is it equal to multiplying it by 8 times instead?
 
[citation][nom]mac_angel[/nom]doesn't anyone understand the difference between 8 times and 8 fold?grab a piece of paper, fold it 8 times, see how many you get. Is it equal to multiplying it by 8 times instead?[/citation]

You can fold a standard piece of paper 8 times? Impressive!
 
[citation][nom]sonofliberty08[/nom]this is because the cpu r still stuck on the crappy old school x86[/citation]

You mean by a real processor that actual are worth the name, not a striped down Risc processor not capable of anything more than most mundane things ?
 
[citation][nom]TA152H[/nom]You're confused, badly. Prefetchers aren't part of programming, it's part of hardware. They have been around since 8086, but what makes memory less important is caches, which make reads to main memory relatively infrequent. Also, a cache always reads in a line, so in a sense that's prefetching, but also, since the Tualatin, the processors prefetch lines as well.With more cores per processor, bandwidth becomes more important as well. If you're one processor and you've got say seven more, you don't want to be hogging the memory bus. If you are, then they're all waiting. Faster bandwidth means you get release it sooner, and let another core have it, if necessary. It also allows for larger cache lines, if that's desirable in the situation, without unduly increasing the time it takes to read it in.[/citation]

Confused how? So you mean that the hardware prefetcher in the x86 don't help memory access by filling the cpu cache with data it predicts will be relevant while the memory controllers aren't loaded, instead of jumping down to the slower memory it can operate out of the way faster cpu cache instead when that data gets relevant. The cpu cache don't fill itself magically you know - Its the prefetchers that does that jobb for you!

Wikipedia explains the process pretty well if you search for "Instruction prefetch"
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom