Fiji and HBM.

[xfia]

amd has said around 70% of vram is allocated and wasted along with bandwidth needs being reduced.
there is something some testing is not showing in regards to real world performance but mantle is not the same as dx12.
but anyway the real world numbers for dx12 that have been demonstrated is 30 percent performance improvement for nv and 100 percent for amd.

 

[bug]

i would not have posted that if i seen the article.. the forbes article was just rich in content and could spark several conversations but it appears some people just need a hug or something..
good thing i got several threads to help fight off the troll haha

And yet, even in the article from Forbes you missed the picture showing that the chip is actually larger. 596 sq mm vs 390X's 438 sq mm - that's 36% larger.
Sure, the total chip+memory area is significantly reduced, but lager chips are still more expensive to make. You'll be able to save $1-2 on the smaller PCB, but the total cost still goes up. And having a larger chip with the memory right next to it means more power to dissipate per unit of area.
There's no miracle at work here.

 

[xfia]

i ignored both the comments about the size because the measurement is obvious and meant to be..
it seems smaller faster cooler running products with the best design efficiency ever are not enough for some people

 

[LightningJR]

amd has said around 70% of vram is allocated and wasted along with bandwidth needs being reduced.
there is something some testing is not showing in regards to real world performance but mantle is not the same as dx12.
but anyway the real world numbers for dx12 that have been demonstrated is 30 percent performance improvement for nv and 100 percent for amd.

70% of VRAM used is wasted huh, so if a game is using 4GB of VRAM it should really only be using about 1.2GB. Or using 8GB of VRAM but should only be using 2.4GB?! I have to say that would be wonderful if true but it's such a huge number I don't know how anyone can believe it.

You're right Mantle is not the same as DX12, Mantle is specifically tailored to GCN, developed by AMD themselves if anything Mantle should give better performance numbers. Mantle is very similar to DX12, do the research, Mantle has the API overhead extremely reduced, it has direct access to all GPU features, even Microsoft has said DX12 will imitate Mantle.

I would love the link to these "real world" numbers for DX12 showing 30% improvement for NV and 100% for AMD. That would be very interesting to read.

 

[Ebo]

I really dunno xfia, but I will put it as kindly as I possibly can.

Flooding forum with different treads regarding the exactly same thing, just with a different angle......well thats not good behavior in my book, end of story.

We are a lot of people looking forward to see what the thingy actually can do, but until NDA is lifted nobody can publish and people can only speculate which is the pit you have been in for quite some time, and it makes peoples eyes roll and think oohhh no, not again.

That you are exited about the new tech, thats allright, but pure speculation and bringing old news in a different bottle dosent make it new all over again. Do like the rest of us, wait until the tests are here then we have something to talk about, namely face value and NOT just the words from the marketing department.

 

[qubit]

@xfia This really is getting old as the others have said. Why not just reserve your comments for when the card is released and W1z does his review? Then we'll all know exactly how it performs.

 

[xfia]

70% of VRAM used is wasted huh, so if a game is using 4GB of VRAM it should really only be using about 1.2GB. Or using 8GB of VRAM but should only be using 2.4GB?! I have to say that would be wonderful if true but it's such a huge number I don't know how anyone can believe it.

You're right Mantle is not the same as DX12, Mantle is specifically tailored to GCN, developed by AMD themselves if anything Mantle should give better performance numbers. Mantle is very similar to DX12, do the research, Mantle has the API overhead extremely reduced, it has direct access to all GPU features, even Microsoft has said DX12 will imitate Mantle.

I would love the link to these "real world" numbers for DX12 showing 30% improvement for NV and 100% for AMD. That would be very interesting to read.

the video showing amd's accomplishment in api overhead for dx12 was locked by a mod. im telling you those are real numbers.
yes yes i know how they are alike but dx12 is better and they have been able to up real performance number and not just take away bottlenecks for budget cpu's.

think we could end the troll and argument about facts? you people are just making anyone that actually wants to talk about it intellectually go away.

 

[the54thvoid]

think we could end the troll and argument about facts? you people are just making anyone that actually wants to talk about it intellectually go away.

You are impossible. You remind me of someone else who went on a long 'vacation'.

 

[LightningJR]

the video showing amd's accomplishment in api overhead for dx12 was locked by a mod. im telling you those are real numbers.
yes yes i know how they are alike but dx12 is better and they have been able to up real performance number and not just take away bottlenecks for budget cpu's.

uhh, That shouldn't keep you from giving me the link to the video of the performance numbers.. A locked thread.. how does that relate to a link to proof of the numbers, just give me the link to the video in this thread? I also don't know how it is you KNOW those are real numbers?? Also how you KNOW they upped real world performance... Am I just suppose to just trust you? I am sorry but you have no argument now if you want me to just take your word for it...

You've lost me... Now you just sound like a fan boy.

 

[xfia]

ahh it took me awhile with the trolling but it could be possible we are already seeing nv at more optimal levels of optimization and the performance advantage of gcn is already being seen
^ now that is actual speculation

i would rather have you in the other thread since you are actually interested and this was meant to really only be for fiji efficiency but i keep fighting trolling instead of getting conversation.

 

[Ebo]

think we could end the troll and argument about facts? you people are just making anyone that actually wants to talk about it intellectually go away.

Ohhh i beg you dont dig that hole any deeper than you already are in.

You know for a fact, that a new OS thats going to be released on 29 next month comes with a perfectly tuned dx12 at launch ?............comon its Microsoft, yes it will work from the start but is it finetuned....not in a million years. Otherwise you will not see any optimizesation to the core of the OS, the next many months or ever ?.
Have you ever seen that from anything Microsoft have sold until now, because if you have, ive missed out on something ever since win 3.1. Ive never seen that before, but again it could be that you mean Windows Me, but im not quite sure.

 

[lilhasselhoffer]

Someone's sounding like a big shiny dickhead. I wonder who that could be?^^^

Perspective.

Forbes Article source, June 22nd
http://www.techpowerup.com/forums/t...-efficiency-to-the-28nm-fiji-silicone.213719/
Youtube, June 20th
will Microsofts DX12 performance unlocking potential prove to be friend of AMD and foe to NV?
Forbes article, June 19th
http://www.techpowerup.com/forums/threads/hbm-update.213634/
AMD promo video, June 22nd
http://www.techpowerup.com/forums/t...-to-amds-incredible-dx12-api-overhead.213722/
Forbes article, June 22nd
http://www.techpowerup.com/forums/t...matically-reduce-vram-bandwidth-needs.213721/
Forbes article, June 20th
http://www.techpowerup.com/forums/t...performance-with-r9-390x.213647/#post-3301534


For those counting, in the span on 3 days (19th-22nd) 6 new threads were generated. Of those threads, 4 linked to Forbes and 2 linked to youtube for the purposes of promotional material. All of the posts relate either directly or tangentially back to the new Fury and its HBM memory. Objectively, I think we can agree that this is either somebody reaching out to make Forbes relevant, or reaching out to evangelize AMD. Neither of these things is acceptable in a place like TPU, where this sort of activity would be called being a shill if it hit the front page.


So, I may have spoken with my wallet and supported AMD. At the same time, this flood of reinterpreted news from a second and third hand source isn't useful as a bunch of new threads. If all of it is relevant, how about you include it in the pre-existing ones (the threads that took me less than 60 seconds to find in my original post here). Supporting AMD, Intel, or Nvidia is one thing, but spamming a forum because you seem to have little or no awareness that you are doing so is unacceptable etiquette.

 

[Aquinus]

Am I the only person who noticed that the title says "Silicone" and not "Silicon". I guess HBM makes them fake titties bounce.

amd has said around 70% of vram is allocated and wasted along with bandwidth needs being reduced.

AMD also said that Bulldozer was going to trample Sandy Bridge and we saw how that turned out. Speculation is one thing, treating it as if it were fact like you are currently doing is hypocritical.

 

[xfia]

Perspective.

Forbes Article source, June 22nd
http://www.techpowerup.com/forums/t...-efficiency-to-the-28nm-fiji-silicone.213719/
Youtube, June 20th
will Microsofts DX12 performance unlocking potential prove to be friend of AMD and foe to NV?
Forbes article, June 19th
http://www.techpowerup.com/forums/threads/hbm-update.213634/
AMD promo video, June 22nd
http://www.techpowerup.com/forums/t...-to-amds-incredible-dx12-api-overhead.213722/
Forbes article, June 22nd
http://www.techpowerup.com/forums/t...matically-reduce-vram-bandwidth-needs.213721/
Forbes article, June 20th
http://www.techpowerup.com/forums/t...performance-with-r9-390x.213647/#post-3301534


For those counting, in the span on 3 days (19th-22nd) 6 new threads were generated. Of those threads, 4 linked to Forbes and 2 linked to youtube for the purposes of promotional material. All of the posts relate either directly or tangentially back to the new Fury and its HBM memory. Objectively, I think we can agree that this is either somebody reaching out to make Forbes relevant, or reaching out to evangelize AMD. Neither of these things is acceptable in a place like TPU, where this sort of activity would be called being a shill if it hit the front page.


So, I may have spoken with my wallet and supported AMD. At the same time, this flood of reinterpreted news from a second and third hand source isn't useful as a bunch of new threads. If all of it is relevant, how about you include it in the pre-existing ones (the threads that took me less than 60 seconds to find in my original post here). Supporting AMD, Intel, or Nvidia is one thing, but spamming a forum because you seem to have little or no awareness that you are doing so is unacceptable etiquette.

as i said its the dumb trolling and people not understanding what is even meant to be speculated because they dont have more facts that are out there.

 

[Aquinus]

as i said its the dumb trolling and people not understanding what is even meant to be speculated because they dont have more facts that are out there.

Speculation is derived from some level of factual knowledge, otherwise it's not speculation, it's pulling shit out of your ass.

 

[Nosada]

as i said its the dumb trolling and people not understanding what is even meant to be speculated because they dont have more facts that are out there.

 

[xfia]

Am I the only person who noticed that the title says "Silicone" and not "Silicon". I guess HBM makes them fake titties bounce.

AMD also said that Bulldozer was going to trample Sandy Bridge and we saw how that turned out. Speculation is one thing, treating it as if it were fact like you are currently doing is hypocritical.

it has not been showed as fact so i said seems to be in regard to how the latency improvement is affecting vram. what is fact is that vram bandwidth needs will be reduced with dx12 so it may be coming full circle on that one.

 

[Aquinus]

it has not been showed as fact so i said seems to be in regard to how the latency improvement is affecting vram. what is fact is that vram bandwidth needs will be reduced with dx12 so it may be coming full circle on that one.

I cite my experience as a software developer and as a person holding a degree in computer science. Latency does not impact (V)RAM usage (any memory for that matter), period. End of story. You can't store twice as much water in a bucket because the faucet dumps water out twice as fast. Learn how computers work before making erroneous claims like that.

A 10 liter bucket is still a 10 liter bucket regardless of how quickly you can put water into and and take water out of it. All types of memory are more similar to that bucket of water than what you're describing.

 

[dorsetknob]

I wish i could troll like @xfia over multiple threads

 

[xfia]

@Aquinus you may very well be right but take into consideration all the changes taking place with dx12 and how amd said it also that actually keeps up with your point.
they where just stating that what they have found to be true about vram is that 70 percent of what is allocated is wasted and wasnt in regard to a direct link to dx12 so maybe a sign for a initiative to tame vram usage.

 

[mroofie]

do you ?

 

[Aquinus]

you may very well be right but take into consideration how amd said it also that actually keeps up with your point
they where just stating that what they have found to be true about vram is that 70 percent of what is allocated is wasted and wasnt in regard to a direct link to dx12 so maybe a sign for a initiative to tame vram usage.

Which has absolutely nothing to do with HBM and everything to do with DX12. This very well could be had by nVidia as well so what's your point? You seem to have lost it.

 

[CAPSLOCKSTUCK]

pointless, pointless thread.....a waste of keystrokes.

 

[Aquinus]

pointless, pointless thread.....a waste of keystrokes.

Xfia is reminding me of vawrvawerawe...

 

[rtwjunkie]

@xfia This really is getting old as the others have said. Why not just reserve your comments for when the card is released and W1z does his review? Then we'll all know exactly how it performs.

And even THEN, when finally the card and DX12 are BOTH out, we still will only know how AMD and NVIDIA stack up against each other with DX11 improvements, so like you, the intelligent people will wait and see actual comparisons, instead of subscribing to the near-religious fervor heaped on us by xfia.