• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Possible Radeon 390X / 390 and 380X Spec / Benchmark (do not hotlink images!!!!!!)

Status
Not open for further replies.
but it's obvious that nVidia are holding back something.

No they dont.
you notice cheap components used in their series to save money. (coilwhine issues) why save money if they hold back? seen the issues with sli drivers lately? it seems more like Nvidia is having a lot of issues that needs fixing.

anyhow, the 390x going to be God like superman.
a brand I can trust as the ceo wont hold up a card with woodscrews for the shareholders.
 
No they dont.
you notice cheap components used in their series to save money. (coilwhine issues) why save money if they hold back? seen the issues with sli drivers lately? it seems more like Nvidia is having a lot of issues that needs fixing.

anyhow, the 390x going to be God like superman.
a brand I can trust as the ceo wont hold up a card with woodscrews for the shareholders.

Since a lot of the non reference ones also suffer from coil whine i doubt its just a simple case of Nvidia using "cheap" components in the 900 series , Only have to lookat how some of the 290 aftermarket coolers were the same as the 700s eries versions to see a lot of brands use the same components for AMD/Nvidia. Read how many gibbo and other retailers are selling so fast thats why we get so many coil whine reports but %wise vs cards sold i dont think its any worse than any other card
 
No they dont.
you notice cheap components used in their series to save money. (coilwhine issues) why save money if they hold back? seen the issues with sli drivers lately? it seems more like Nvidia is having a lot of issues that needs fixing.

What a load of delusional rubbish, Many cards have suffered with coil whine from both manufacturers. What issues with SLI? I see no more than usual and no more than with Crossfire.

But I do hope the 390 series is awesome.
 
Since a lot of the non reference ones also suffer from coil whine i doubt its just a simple case of Nvidia using "cheap" components in the 900 series , Only have to lookat how some of the 290 aftermarket coolers were the same as the 700s eries versions to see a lot of brands use the same components for AMD/Nvidia. Read how many gibbo and other retailers are selling so fast thats why we get so many coil whine reports but %wise vs cards sold i dont think its any worse than any other card

The 970s whine more because there was no reference design. Without one the AIBs just do the bare min.

20nm from 2015, then quick transition into 16/14nm FinFet. 10nm from 2017..

lNTRg3h.png

Seen it before, it's not authoritative. Forget about 20nm for GPUs unless GF can maybe come up with one that works, but it won't be 2015.

28nm for AMD soon, with NV's answer this time next year on 16FF+. Paper launched if they have to. After that AMD's answer-answer on either 20nm FDSOI @ GF or 16FF+.
 
Last edited:
The 970s whine more because there was no reference design. Without one the AIBs just do the bare min.

+1,

If it was within the parameters of every other series of gpu's that preceded and wasn't a widespread problem, there would have been no OcUK commissioning of custom 970's with specific higher quality (glued)vrms to combat the whine-it's pretty simple really.

Hopefully the next round of gpus(regardless the vendor) take steps to help eradicate the issue.
 
It's been rumoured for some time that after such a huge delay for 20nm and with 14/16 on the horizon AMD/Nvidia might just skip it. And 14/16 might be more suited to trialling with a low volume/high value product like a Mid & High GPU. I guess the worrying thing is no mention of timescales. But no news is good news? :)
 
Last edited:
Did you even read it? 20nm is unfit for GPUs.

Give up on 20nm.


kitguru; said:
Based on market rumours, only select graphics processors and, perhaps, console chips, will be made using 20nm process technology.

kitguru; said:
“We will have certain products in 20nm and then we will go to FinFET from there”

So it's not unfit for GPUS at all - as I read it, because 16nm FinFET/FinFET+ uses the same interconnects as 20nm, it's likely they can produce some 20nm products, and move straight to 16nm with very little difference.

Mid range parts are always made first - can't see this being any different - you don't risk high end parts on a new process in case you get yield issues. Low end parts need to be cheap as possible, and don't tend to push the envelope, so they can still be run on the previous process (hence why most lend products are rebrands)
 
28nm with HBM and AIO to cool wouldn't be all that bad. If it's better for AMD until they can transition to a new node.

Feel like holding fire on new GPU's until 2016, because by then we might actually get something really new, good enough to make a significant jump over what we have today. GPU space is in danger of becoming as boring as the CPU space :P

Roll on die shrinks and big performance jump !!
 
We will have certain products in 20 nanometers and then we’ll go to FinFET from there,

From this they decided to take that everything at 28nm will transition directly to finfet? In what world.. the ONLY way to take that specific sentence is some products will be 20nm and then to Finfet from there(being 20nm). You can't randomly decide they were talking about 28nm.

Such a crap website, it also says most products are 28nm because... they are, everything they have out now is 28nm, no where does it say everything going forward is 28nm, lots of new things on 28nm.

As an aside, Samsung appear to be shipping 14nm wafers in Q1 2015 and it's being suggested that GloFo are actually slightly ahead of them. There is no real hint of what is coming at the moment especially as for one thing there is more than one 20nm process, for a second thing, we might see a 14nm product before we see a 20nm product.

A cpu out say Q2 2015 might have needed to tape out in q3 13-q12 14 and it might only have been possible to work with 20nm then. A GPU out in Q2 2015 might have taped out in Q3 14, and being at least 6 months later may have taped out on 14nm finfet... yet could even launch before the CPU.

GloFo/Samsung look like they will beat TSMC to finfet, but did AMD know that 2 years ago when these products would have needed to pick one or the other?

Yields come in to it, 14nm might be shipping from January 2014, but while you might take a 100mm^2 arm chip, add 10mm^2 in redundancy and hit 40% yields and can make a viable product, a 400mm^2 gpu even with more redundancy built in might be sub 15% yields, maybe lower, and simply not be able to make a volume product nor a financially viable one.

At this stage I don't think 14nm/20nm availability will prevent new GPU's using them, more like yields(not great across the board with finfets at TSMC/Samsung by all accounts) making it not worth while.
 
5% every year like CPU?

Who knows, everything seems to be slowing down though. I guess the market just isn't there to really push development anymore.

Upgrading GPU every two years makes more sense these days..

or redesign of what the gpu should do on a given node.
die shrinks are slowing down and soon stops.
Unless they solve this like nand has done we are in for a tech slow down big time.
 
From this they decided to take that everything at 28nm will transition directly to finfet? In what world.. the ONLY way to take that specific sentence is some products will be 20nm and then to Finfet from there(being 20nm). You can't randomly decide they were talking about 28nm.

Such a crap website, it also says most products are 28nm because... they are, everything they have out now is 28nm, no where does it say everything going forward is 28nm, lots of new things on 28nm.

As an aside, Samsung appear to be shipping 14nm wafers in Q1 2015 and it's being suggested that GloFo are actually slightly ahead of them. There is no real hint of what is coming at the moment especially as for one thing there is more than one 20nm process, for a second thing, we might see a 14nm product before we see a 20nm product.

A cpu out say Q2 2015 might have needed to tape out in q3 13-q12 14 and it might only have been possible to work with 20nm then. A GPU out in Q2 2015 might have taped out in Q3 14, and being at least 6 months later may have taped out on 14nm finfet... yet could even launch before the CPU.

GloFo/Samsung look like they will beat TSMC to finfet, but did AMD know that 2 years ago when these products would have needed to pick one or the other?

Yields come in to it, 14nm might be shipping from January 2014, but while you might take a 100mm^2 arm chip, add 10mm^2 in redundancy and hit 40% yields and can make a viable product, a 400mm^2 gpu even with more redundancy built in might be sub 15% yields, maybe lower, and simply not be able to make a volume product nor a financially viable one.

At this stage I don't think 14nm/20nm availability will prevent new GPU's using them, more like yields(not great across the board with finfets at TSMC/Samsung by all accounts) making it not worth while.

+1

Still no more the wiser if GPU's will come on 28nm / 20nm.

People just doing a lot of guess work, we won't really know until it happens :P

Sure we'll get some decent leak closer to launch.


or redesign of what the gpu should do on a given node.
die shrinks are slowing down and soon stops.
Unless they solve this like nand has done we are in for a tech slow down big time.

Yeah things are def moving slower, we've been on 28nm now for what seems like a lifetime, current cards from past couple years are all within punching distance of each other. Upgrading every 2+ years seems to be the sensible thing to do. If only I could do that myself :P
 
Last edited:
Status
Not open for further replies.
Back
Top Bottom