AMD Fury series coming soon.

LordEC911 said:
That is just wrong.
AIBs and their repair houses have the skills and tools to fix and replace components on PCBs, such as GDDR5 ICs.
Amkor is the only one that can "replace" a HBM stack on the interposer, if that is even possible.
Hynix is the only one that can "fix" a defective HBM stack, if that is even possible.
Click to expand...

Yeah, while I can technically see it being possible to replace the interposer, I can't think of a situation where it would ever be economically feasible, outside of failing test right after assembly of the GPU and HBM stacks onto the interposer. Might be possible then to just desolder, clean and reflow on a new interposer. And yeah, not possible at all with the HBM stack, since it's using TSVs to connect the chips, so the silicon layers are literally bonded together. Well, I suppose it might be technically possible in a fun engineering sort of way to see if you actually could.

All this talk of interposers reminds of working on the POWER4 chips, where they used this funky flexible interposer between the mainboard and the MCM. Instead of using solder balls, it used tiny little balls of gold thread, which dislodged from the interposer at the slightest brush. You want tedious, I'll show you tedious, it's looking at a 3400+ connection interposer under a lab microscope, trying to find the slightest imperfection that you attempt to fix with the most technical of tools, a needle.
This shows it perfectly, right along with missing and shorted pads: http://www.cpushack.com/wp-content/uploads/2013/08/IMG_7193.jpg

Edit:
Related to the HBM stacks, I suppose it might be possible to bin them if they have a "failed" chip? I.e. you could sell 512MB stacks that have 2 out of 4 chips functional. Probably not useful right now, since capacities are so low and only 4 chips on the stack, but maybe in the future with with a lot more stacked it will be done. Who knows, but I suppose something like that might allow HBM to trickle down to the lower end, since these binned chips would have to be fairly cheap.
 
Last edited:
madgun said:
So this means if one of the stacked module is faulty the whole package is a waste. Yikes!

This will make one hell of an expensive exchange.
Click to expand...
I would imagine it must be doable to repair/replace it but it wouldn't be at the AIB's repair center.

JBark said:
Edit:
Related to the HBM stacks, I suppose it might be possible to bin them if they have a "failed" chip? I.e. you could sell 512MB stacks that have 2 out of 4 chips functional. Probably not useful right now, since capacities are so low and only 4 chips on the stack, but maybe in the future with with a lot more stacked it will be done. Who knows, but I suppose something like that might allow HBM to trickle down to the lower end, since these binned chips would have to be fairly cheap.
Click to expand...
Hynix sells only "Known Good Stacked Die" so that really isn't AMD's problem to worry about. There has been some papers and talk about being able to "repair" the stack after it is already assembled. But again, that is something that Hynix would do in-house.
 
Last edited:
More updates from Gibbo:

"Hi there


I did a test last night, I closed up my case but left the radiator just in the bottom of the case exhausting its heat inside the case, this is a really bad idea by the way, but I wanted to see how hot the card would get.

After running Heaven 4.0 all night, the maximum temperature was 58c and the fan speed never exceeded 16%, starting point percentage for the fan is 15%, so it was quieter than the case fans and I was able to sleep with no issues at all.

Here is a picture, its in a Corsair case with the smoked side panels, so its not easiest to see and ignore the SSD's just randomly hanging in the bottom of case:"

IMG_3015.jpg


So loosely hanging drawing in a mix of ambient and case air and low fan speed not going over 60 C

He also mentioned his stock voltage was not hitting 1200mhz. and that if he did hit 1200-1300 things would get interesting. So with voltage unlock maybe at around 1300 mhz we would start to see Fury X matching 980 Ti OC in most benches. If it goes over that to 1400mhz+ being faster than G1? Question is how well does it respond to voltage increase (how much increase in TDP) and when will they unlock voltage.
 
lee0539 said:
More updates from Gibbo:

"Hi there


I did a test last night, I closed up my case but left the radiator just in the bottom of the case exhausting its heat inside the case, this is a really bad idea by the way, but I wanted to see how hot the card would get.

After running Heaven 4.0 all night, the maximum temperature was 58c and the fan speed never exceeded 16%, starting point percentage for the fan is 15%, so it was quieter than the case fans and I was able to sleep with no issues at all.

Here is a picture, its in a Corsair case with the smoked side panels, so its not easiest to see and ignore the SSD's just randomly hanging in the bottom of case:"



So loosely hanging drawing in a mix of ambient and case air and low fan speed not going over 60 C

He also mentioned his stock voltage was not hitting 1200mhz. and that if he did hit 1200-1300 things would get interesting. So with voltage unlock maybe at around 1300 mhz we would start to see Fury X matching 980 Ti OC in most benches. If it goes over that to 1400mhz+ being faster than G1? Question is how well does it respond to voltage increase (how much increase in TDP) and when will they unlock voltage.
Click to expand...

I always buy new tech when faced with an option and Fury it is.
:D
 
Man, is it wrong that even if the G1 980Ti ends up 10% faster or so, I'd still probably pass on it because it's xbox huge compared to the Fury?
 
he said that the G1 (factory OC'd 980ti) beats fury in synthetics but in games its "far far closer", which could be a positive given that the fury can't really be OC'd much (at least his sample can't as he cant adjust voltage)

once the new afterburner update comes out w/ fiji compatibility we'll see how well it OC's vs an OC'd 980ti

doesnt matter much to me as i'm not buying a new $400 computer monitor to use w/ this $650-700 card
 
Who is this guy, why does he have a Fury X!? Let's go find his house!

Ok jkz. But cant he do some benches now and release it?
 
shadow85 said:
Who is this guy, why does he have a Fury X!? Let's go find his house!

Ok jkz. But cant he do some benches now and release it?
Click to expand...

AMD knows who he is, so he's intentionally being vague w/ the information as he doesn't wanna be blackballed probably, he works for OCUK which is an online tech store
 
Reality said:
AMD knows who he is, so he's intentionally being vague w/ the information as he doesn't wanna be blackballed probably, he works for OCUK which is an online tech store
Click to expand...

He is the head of graphics at OCUK i.e. he manages orders, supply, pricing etc. They got an example Fury X and he's been playing with it since.

As a re-seller, I don't believe he's technically breaking any NDA as he likely never signed one. That said, he's not revealing any real numbers yet as he likely doesn't want to draw the ire of AMD.
 
lee0539 said:
More updates from Gibbo:

"Hi there


I did a test last night, I closed up my case but left the radiator just in the bottom of the case exhausting its heat inside the case, this is a really bad idea by the way, but I wanted to see how hot the card would get.

After running Heaven 4.0 all night, the maximum temperature was 58c and the fan speed never exceeded 16%, starting point percentage for the fan is 15%, so it was quieter than the case fans and I was able to sleep with no issues at all.

So loosely hanging drawing in a mix of ambient and case air and low fan speed not going over 60 C
Click to expand...

If this is true. I'm SOLD! There is nothing I like the most than a nice cold beer&GPU :D
 
So loosely hanging drawing in a mix of ambient and case air and low fan speed not going over 60 C

He also mentioned his stock voltage was not hitting 1200mhz. and that if he did hit 1200-1300 things would get interesting. So with voltage unlock maybe at around 1300 mhz we would start to see Fury X matching 980 Ti OC in most benches. If it goes over that to 1400mhz+ being faster than G1? Question is how well does it respond to voltage increase (how much increase in TDP) and when will they unlock voltage.
Click to expand...

All the benchmarks that have been released show the Fury X beating the 980ti. Not really sure what you're talking about. Overclocking wise, its hard to know what the limits of the chip will be. Previous generation GCN products usually topped out at around 1300mhz for standard cooling. Fiji is a new iteration of GCN so its improvements could help or hurt its overclocking ability.
 
ebduncan said:
All the benchmarks that have been released show the Fury X beating the 980ti. Not really sure what you're talking about. Overclocking wise, its hard to know what the limits of the chip will be. Previous generation GCN products usually topped out at around 1300mhz for standard cooling. Fiji is a new iteration of GCN so its improvements could help or hurt its overclocking ability.
Click to expand...

So you're referring to benchmarks that have been released by AMD? Not a very reliable source to go on, of course they are going to paint the Fury X in a good light. Unless of course you can point me to an independent review of the Fury X vs. 980ti, and if you can, by all means do so.

Too many people are jumping the gun on these comparisons.
 
Last edited:
he said that the G1 (factory OC'd 980ti) beats fury in synthetics but in games its "far far closer", which could be a positive given that the fury can't really be OC'd much
Click to expand...

why is this positive? i dont get it, at best fury x is about as fast as an oced 980 ti at the same price;

At what point did you people gave up Fury X will be a Titan X killer ? might be me, but with hbm and 4000 shaders i was expecting it will run over 980 ti /Titan X
 
VonGriffin said:
why is this positive? i dont get it, at best fury x is about as fast as an oced 980 ti at the same price;

At what point did you people gave up Fury X will be a Titan X killer ? might be me, but with hbm and 4000 shaders i was expecting it will run over 980 ti /Titan X
Click to expand...

Fury X may well be the 980 Ti/Titan X killer at 4k and over resolutions, and possibly with Crossfire vs SLI.
 
He also said he was running 1080 p and games were running at 1080p that he was testing. I don't think that is a good indication of performance during games.
 
980ti and Fury X start at the same price point and it seams Fury X will be (at least at first) more expensive
And some of us game at 1440p and some use 1080 at 120hz

Before you continue using half arsed arguments remember Nvidia has most of the gpu market share if AMD wants to pull out of the value market it needs to deliver a clear flagship card and not just a 980 ti alternative without gameworks, hdmi 2.0 and less vram
 
Fury X will outperform the Titan X by around 5-10% at launch and that's without optimized drivers.
 
I think the 980 TI and Fury X, after both are overclocked, will have almost the same performance. With slight advantage going to the 980 Ti as it has a larger buffer.
 
yourgrandma said:
Us 120hz/144hz users as well.
Click to expand...

Yes, even at 1440p I'll probably end up with two fury's trying to keep fps high, especially with secondary monitor. I gave up on surround monitors, it's not that great. 4k is a gimmick at best right now.
 
frunction said:
Yes, even at 1440p I'll probably end up with two fury's trying to keep fps high, especially with secondary monitor. I gave up on surround monitors, it's not that great. 4k is a gimmick at best right now.
Click to expand...

Yup, starting with one at 1440, will see if 2 are needed. 4K at 60hz doesn't work for me at all, just too used to 120/144hz.
 
You must log in or register to reply here.
Back
Top