AMD Blocks/Disables Overclocking the Memory for the Radeon R9 Fury X

[jaggysnake57]

i was going to reply to some of the fucking stupid comments but then ..... this forum is becoming toxic with the same members spewing hate. 

 

as for this....who fucking cares. How much performance do you really get from the memory clock anyway.....get off your fanboxes 

[ThreePinkApples]

i agree with amd here .. its a new type of technology.... who knows what could happen.... u might eventually break the card....

 

This, AMD might have a very good reason for locking it, they might have had really bad results with pushing it further. 

 

Stop complaining about it, we don't know the reason they did it. Wait for someone to (hopefully) be able to go around the lock, then we can see if it is really dangerous or not. Of course that person might just be lucky with yields, but if it spreads, we will know eventually

[harrynowl]

Given all the reports of very low yields they probably dont overclock all that well anyway.

[Techhog]

Uh, I thought that we all knew that HBM wouldn't be overclockable? And why does it even matter? It's already faster than any GDDR5 speed you can reach. It's not going to bottleneck the card at all.

 

You people bitch over the dumbest things. Do you really think that overclocking it would make a difference? In fact, with the chips right next to the die, why would you even want to overclock it? It'll just limit your GPU overclock. You people will get mad over anything so long as it involves something you had no plans to buy in the first place.

[patrickjp93]

Would overclocking the HBM brick the card by any chance? :huh:

There's always a tiny risk that overclocking anything at all will brick your chip. As for that chance for HBM, I couldn't tell you with precision.

[patrickjp93]

Uh, I thought that we all knew that HBM wouldn't be overclockable? And why does it even matter? It's already faster than any GDDR5 speed you can reach. It's not going to bottleneck the card at all.

You people bitch over the dumbest things. Do you really think that overclocking it would make a difference? In fact, with the chips right next to the die, why would you even want to overclock it? It'll just limit your GPU overclock. You people will get mad over anything so long as it involves something you had no plans to buy in the first place.

512-bit bus using 8GHz GDDR5 would have exactly the same bandwidth as this HBM at 500MHz.

[marldorthegreat]

 

This thread is quickly becoming hilarious, most people are agreeing that partly locking down overclocking on a watercooled flagship desktop card which was presented to us and advertised as some overclockers dream is perfectly okay.

 

You people are such hypocrites.

 

WE DONT KNOW HOW HBM WORKS. IT COULD BREAK IF OVERLCLOCKED. As the dies are stacked so they cannot dissipate heat as well. 

[Kamina]

Lame, lame, lame. My money, my freedom. 

[Techhog]

512-bit bus using 8GHz GDDR5 would have exactly the same bandwidth as this HBM at 500MHz.

I didn't realize it was 500MHz. Still though,hardly any GDDR5 can reach that speed, and it's not going to be a bottleneck. I'm also pretty sure that HBM's issue is it having speed bins.

[ItsAFeature]

I can't see this as any legitimate reason not to buy a Fury. I've never witnessed any tangible benefit to overclocking my GPU memory and with HBM being new and relatively low yield at the moment, I can see why AMD wants users to leave it alone.

[spartaman64]

Uhh, that was legions of people bashing Nvidia?

more like 3 people while everyone else was calling them idiots for caring 

[spartaman64]

this might be a good move or it might be a bad move. if HBM is fragile when it come to overclocking then this is good. but if HBM can handle overclocking while getting a big performance boost then this is bad

[Techhog]

There's always a tiny risk that overclocking anything at all will brick your chip. As for that chance for HBM, I couldn't tell you with precision.

Maybe I just have bad luck, but in my experience GDDR5 almost never overclocks well. Low-yield HBM is going to do even worse, and again it would do more harm than good. In the best case scenario, someone will make a custom vBIOS with HBM overclocking, only to discover that the extra heat reduces the core clock headroom significantly while the memory overcolck brings not even 0.000001FPS improvement. Taking away choice sucks, but if you're not going to buy it because of this, you're an idiot.

[Techhog]

this might be a good move or it might be a bad move. if HBM is fragile when it come to overclocking then this is good. but if HBM can handle overclocking while getting a big performance boost then this is bad

Yeah... No. Even if you doubled the speed, it would not bring a noticeable performance boost. If it could, then overclocking the 980 Ti's or Titan X's GDDR5 should bring a big performance boost. It doesn't.

[patrickjp93]

Maybe I just have bad luck, but in my experience GDDR5 almost never overclocks well. Low-yield HBM is going to do even worse, and again it would do more harm than good. In the best case scenario, someone will make a custom vBIOS with HBM overclocking, only to discover that the extra heat reduces the core clock headroom significantly while the memory overcolck brings not even 0.000001FPS improvement. Taking away choice sucks, but if you're not going to buy it because of this, you're an idiot.

Well, remember GDDR5 is quad pumped. A 100MHz overclock is really 400MHz-worth. I don't think AMD is justified to do this. Had review samples been sent out on time and reviewers were given express warnings on VRAM overclock thresholds to tell their flocks, everyone here would've been much happier.

[Techhog]

Well, remember GDDR5 is quad pumped. A 100MHz overclock is really 400MHz-worth. I don't think AMD is justified to do this. Had review samples been sent out on time and reviewers were given express warnings on VRAM overclock thresholds to tell their flocks, everyone here would've been much happier.

I know how GDDR5 works. It just that I've heard that the bins are actually very high. I think that what's needed is a better explanation of why. It's still not a big deal though, because there isn't a GPU on Earth which will be bottlenecked by this bandwidth while Fiji's overclocks likely would be hurt by extra heat from the HBM chips. You people just love to complain about everything.

[Lukiose]

Don't see the problem really.. as far as i am concerned overclocking the core clock is definitely far, far, far more important than increasing the memory clock.

I doubt it matters at all below 4K.

[patrickjp93]

I know how GDDR5 works. It just that I've heard that the bins are actually very high. I think that what's needed is a better explanation of why. It's still not a big deal though, because there isn't a GPU on Earth which will be bottlenecked by this bandwidth while Fiji's overclocks likely would be hurt by extra heat from the HBM chips. You people just love to complain about everything.

I'm just annoyed at the uncompromising defense of AMD when Nvidia got huge flak for disabling laptop overclocking when it was causing huge OEM problems and headaches from stupid users who keep trying to claim warranties after overclocking when it's difficult to detect if the GPUs were overclocked. This is the same damn thing down just 1 step, on  DESKTOP ENTHUSIAST CARD!

[Sammael]

WaitWait wait wait...

You are imagining that people will come up with excuses for amd? On this forum?

Have you been living under a rock?

3,5gb fiasco was legions of people defending Nvidia, but if amd so much as farts they get Shit for it.

Either you haven't been paying attention or you live in opposite world.

Also funny how you want make it more unbalanced by promoting amd bashing even more. Pathetic attempt.

 

 

For the record I did not think it was a big deal that nvidia restricted overclocking of mobile gpus, different vendors have different notebook sizes with different thermal constraints based on the configurations used, such that not all 970ms perform the same and are as capable of being pushed to the same degree.   So why not tamp down on that.

[Sammael]

A more important question is whether games are constrained in performance by the current memory clocks.  Are they?  If not why would this be an issue?

[Guest Generallee]

It's not like they gave you 3.5Gb instead of 4

 

Also memory clocks are basically irrelevant since they don't impact your holy fps by much

 

[GER_T4IGA]

Some people are really into hardmodding cards.

 

Is that even possible these days? lol

[Sammael]

I'm just annoyed at the uncompromising defense of AMD when Nvidia got huge flak for disabling laptop overclocking when it was causing huge OEM problems and headaches from stupid users who keep trying to claim warranties after overclocking when it's difficult to detect if the GPUs were overclocked. This is the same damn thing down just 1 step, on  DESKTOP ENTHUSIAST CARD!

 

 

You are blinded by your own perceptions I think.  You come across in comments as generally quite negative towards amd on the whole.  I'm virtually certain you prefer nvidia products.  You have said in the past you are not fond of AMDs project to push forward HSA computing in part because you despise coding for it and find it a chore or torturous or worse.  Again, there is a general negative vibe and energy wafting from you about amd in general.  Which is fine, but I think that is what is coloring what you see. 

 

Go read comments and responses that are either pro or against amd and nvidia.  The pro nvidia comments have more likes, the pro amd comments tend to have fewer.  That cool crab guy with that amd sniping comment on the very first page in this thread is a clear example.  the amd defense is not so rebust as you suspect it to be, it just seems that way to you because of your own antagonism towards the company and their efforts.

[osbios]

How about we just take care that the information is out there for other people to know so they can decide for them self?

The critic against nvidia in case of the slower 0.5 GiB of the 970 was all about them withholding that information.
And in case of limiting there mobile chips max. voltage it was the fact that a new driver was forcing this limit onto the users if they didn't want to stick with older drivers.

 

As a counter example take the double precision performance of the 980, 980TI or Titan X. Its actually quite bad, but thats ok! Because the infomation is avaible to all buyers and they can decide if the need it.

Fiji is not even out in the wild and we already got the information that the memory won't run faster then the stock 512 GiB/s. So its all up for the potential buyers to decide if they are OK with that.

[patrickjp93]

You are blinded by your own perceptions I think.  You come across in comments as generally quite negative towards amd on the whole.  I'm virtually certain you prefer nvidia products.  You have said in the past you are not fond of AMDs project to push forward HSA computing in part because you despise coding for it and find it a chore or torturous or worse.  Again, there is a general negative vibe and energy wafting from you about amd in general.  Which is fine, but I think that is what is coloring what you see. 

 

Go read comments and responses that are either pro or against amd and nvidia.  The pro nvidia comments have more likes, the pro amd comments tend to have fewer.  That cool crab guy with that amd sniping comment on the very first page in this thread is a clear example.  the amd defense is not so rebust as you suspect it to be, it just seems that way to you because of your own antagonism towards the company and their efforts.

I buy whatever's best for me at the time. The last time my family owned an AMD CPU was before Core 2 Quad. My little brother has 2 7970GHz in XFire. I personally have a pair of Titan Blacks on my home workstation and a GTX 570 in my college rig that is very due for replacement. My negativity towards AMD is pessimism in regards to their long-term future based on the various movements in the industry and the markets at large, not towards what they can produce now. My analysis of Zen is based on all available information regarding Excavator, the Zen core/cluster diagrams, and AMD's claim of 40% IPC gains over Excavator. That still even theoretically dumps it right around Haswell, maybe 1-2% above, depending on the clock speeds we see.

 

The HSA thing is a terrible design. There are actually books on how to design a programming language for ease of use. HSA and HSAIL make x86 assembly seem like a cakewalk. Is that my fault? I've been in the trenches of C++ 98 and C++11 (thank God for 11) combined with OpenCL, OpenGL, and all the markup languages such as OpenMP and OpenACC which allow for very rapid code transformation to be highly parallel and be offloaded to an accelerator. If HSA is not that easy, it won't be adopted. That's just the facts. Intel spends so much money and man power on refactoring the CUDA libraries of their clients to use OpenMP, OpenCL, and OpenACC for their Xeon Phi accelerators, and those are code bases multiple hundreds of millions of lines long. If AMD develops a new standard that can't be used to refactor code as easily, there's no way it'll take off in HPC. I don't blindly dislike AMD. They make so many boneheaded decisions both in marketing and software it ruins the good they do in hardware. That is as objective as you can get.

 

I have no antagonism with AMD's efforts, only its methods. And the defense of AMD is not robust, but where is the due aggression? Nvidia is held to standards of perfection beyond reason but 90% of this thread has said "Meh, don't care." or "I buy the new tech reasoning." It's not defense so much as it is people rolling right over and buying it. There were better ways for AMD to handle it if this was the case. Give early reviewers a very detailed description of the memory overclocking limitations and instruct them to point out that warning in the review. Do that much and AMD's golden. Because then you have the warning to not do it, and the reviewer can provide proof of the limitation in testing.