[WCCF] AMD Radeon R9 390X Pictured

[Paratus]

Pay the Russians to take one up to the ISS next time they send a resupply mission. Problem solved.

This is why our 386's that run the ISS are faster than any other computer in the world. :sneaky:

 

[Vesku]

Assuming that's Fiji...

It could be pure promo imagery, or it could be a custom 295x2 as well

Whether rendered or real the top plate would be consistent with dual sided HBM. Or just an odd graphics effect they slapped on there. ;p

 

[Cloudfire777]

Its obviously not a teaser for the old R9 295X2 thats been on the market for over a year.
"Most advanced" is because of HBM.

Think people

 

[3DVagabond]

It ALL matters actually. Doing well currently is no guarantee to always grow the business, taking the above approach is short sighted and has killed many companies.

When your cash cow has a finite time in the market place it'll make you do desperate things.

 

[3DVagabond]

Whether rendered or real the top plate would be consistent with dual sided HBM. Or just an odd graphics effect they slapped on there. ;p

Not this ^^^

An where is this "dual sided HBM" coming from? HBM operates via an interposer not attached to/through the PCB.

Its obviously not a teaser for the old R9 295X2 thats been on the market for over a year.
"Most advanced" is because of HBM.

Think people

This ^^^

 

[Glo.]

http://www.fudzilla.com/news/graphics/37790-amd-fiji-aims-at-849-retail-price

I think this could be interesting

 

[3DVagabond]

http://www.fudzilla.com/news/graphics/37790-amd-fiji-aims-at-849-retail-price

I think this could be interesting

They're still trying to support their dual GPU predictions now that it appears obvious that the 390X is a single GPU.

 

[Glo.]

They could mean R9 395X2 or whatever that is called.

As been proved many times, HBM1 does not allow more than 4 GB of RAM for single core.

8 GB would be possible for dual GPU setup.

 

[goa604]

They could mean R9 395X2 or whatever that is called.

As been proved many times, HBM1 does not allow more than 4 GB of RAM for single core.

8 GB would be possible for dual GPU setup.

As been mentioned many times, dual link interposer.

 

[desprado]

http://www.fudzilla.com/news/graphics/37790-amd-fiji-aims-at-849-retail-price

I think this could be interesting

If this is true than GG AMD.

 

[Silverforce11]

If this is true than GG AMD.

What's the % chance where Fud's predictions are true?

 

[desprado]

What's the % chance where Fud's predictions are true?

That is why i said if so it is maybe or maybe not.

 

[Creig]

If this is true than GG AMD.

Just like Nvidia was GG with a $1,000 Titan, right?

 

[desprado]

Just like Nvidia was GG with a $1,000 Titan, right?

No because people know what they are paying for and i bet that if AMD really is selling at that price than it almost impossible for AMD get back in the game.

 

[PPB]

No because people know what they are paying for and i bet that if AMD really is selling at that price than it almost impossible for AMD get back in the game.

So it is ok when Nvidia sells a 1K GPU, but AMD cant sell a same performing one for 150 less and with WC included.


Is there any other deduction made by your 3rd grade logic that you should tell us? We are all ears (or eyes).

 

[desprado]

So it is ok when Nvidia sells a 1K GPU, but AMD cant sell a same performing one for 150 less and with WC included.


Is there any other deduction made by your 3rd grade logic that you should tell us? We are all ears (or eyes).

with 4GB it is nearly impossible enough said.

 

[PPB]

with 4GB it is nearly impossible enough said.

Oh, your 3rd grade logic also deducts it is a 4GB card. Interesting. Anything else? It's time to spill all that knowledge and enlighten us, pal :awe:

Infraction issued for personal attack.
-- stahlhart

 

[tviceman]

So it is ok when Nvidia sells a 1K GPU, but AMD cant sell a same performing one for 150 less and with WC included.


Is there any other deduction made by your 3rd grade logic that you should tell us? We are all ears (or eyes).

Look at this in a vacuum instead of insulting someone. AMD is having a hard enough time selling cards for $300. For 40-50% more performance, who will want to spend nearly 3x more money for said performance improvement?

 

[nvgpu]

http://techreport.com/review/28294/amd-high-bandwidth-memory-explained

This first-gen HBM stack will impose at least one limitation of note: its total capacity will only be 4GB.

http://www.pcper.com/reviews/Genera...Memory-HBM-Architecture-AMD-Plans-Future-GPUs

The first iteration of HBM on the flagship AMD Radeon GPU will include four stacks of HBM, a total of 4GB of GPU memory.

http://hexus.net/tech/news/graphics/83221-amd-spills-details-hbm-memory/

Memory partner Hynix has openly stated that first-generation HBM will use 2Gb slices resulting in 256MB on each of the four-hi-stack modules. That's a total of 1GB per stack or 4GB for the four supported stacks. This means, quite clearly, the Radeon R9 390X, the first recipient of HBM memory, will very likely have 4GB of HBM memory, not 8GB or a Titan X-matching 12GB.

http://www.bit-tech.net/hardware/memory/2015/05/19/an-overview-of-high-bandwidth-memory-hbm/1

Back to the here and now, there have been rumours flying around recently that AMD's next generation of GPUs will be limited to 4GB frame buffers when using HBM, and we can confirm this is the case.

 

[PPB]

Look at this in a vacuum instead of insulting someone. AMD is having a hard enough time selling cards for $300. For 40-50% more performance, who will want to spend nearly 3x more money for said performance improvement?

Your problem is that you look things in a vacuum. AMD cant sell well a $300 product that has been bastardized to death because of a lousy reference design that caused high noise, high temps, made the chip throttle at stock and also hurt even more its power consumption. In reality the aftermarket cards can be had for the same or less money, run cool and consume less power.

Yep, that is the power of having a product stained by a terrible marketing strategy (having such a bad reference design being launched first).

Try to think the opposite way, how is Nvidia being able to sell the Titan for 1K if for 33% of its listing price you can have 60% of its performance? Yeah, you just discovered that the function between price and performance in a GPU stack is not linear, rather than a parabole (and Titan X, despite being a FP64 turd, still sells like hotcakes).

That slide only says 4GB max per stack, how many stacks can have a GPU sitting alongside the die again? We can just ask KNL and its propietary variant, cant we? :awe:

 

[Silverforce11]

That's a very interesting article, it seems AMD is starting the journalist brief train already.

-----------------
"This first-gen HBM stack will impose at least one limitation of note: its total capacity will only be 4GB. At first blush, that sounds like a limited capacity for a high-end video card. After all, the Titan X packs a ridiculous 12GB, and the prior-gen R9 290X has the same 4GB amount. Now that GPU makers are selling high-end cards on the strength of their performance at 4K resolutions, one might expect more capacity from a brand-new flagship graphics card.

When I asked Macri about this issue, he expressed confidence in AMD's ability to work around this capacity constraint. In fact, he said that current GPUs aren't terribly efficient with their memory capacity simply because GDDR5's architecture required ever-larger memory capacities in order to extract more bandwidth. As a result, AMD "never bothered to put a single engineer on using frame buffer memory better," because memory capacities kept growing. Essentially, that capacity was free, while engineers were not. Macri classified the utilization of memory capacity in current Radeon operation as "exceedingly poor" and said the "amount of data that gets touched sitting in there is embarrassing."

Strong words, indeed.

With HBM, he said, "we threw a couple of engineers at that problem," which will be addressed solely via the operating system and Radeon driver software. "We're not asking anybody to change their games.""

 

[Glo.]

As been mentioned many times, dual link interposer.

http://semiaccurate.com/2015/05/19/amd-finally-talks-hbm-memory/

That means if you want to make a high-end modern GPU with at least 4GB on the card, you need four stacks even if you don’t need all the bandwidth. If you want to go to 8GB HBM graphics cards, sorry. That said a four stack GPU should have more than enough bandwidth and capacity to satisfy any modern game at 4K resolutions, by the time the second gen HBM GPUs come out, HBM will be on their second gen with 2-8GB per stack plus double the bandwidth.

FUDzilla is very well informed about coming lineup of AMD GPUs.

 

[tviceman]

Your problem is that you look things in a vacuum. AMD cant sell well a $300 product that has been bastardized to death because of a lousy reference design that caused high noise, high temps, made the chip throttle at stock and also hurt even more its power consumption. In reality the aftermarket cards can be had for the same or less money, run cool and consume less power.

Yep, that is the power of having a product stained by a terrible marketing strategy (having such a bad reference design being launched first).

Try to think the opposite way, how is Nvidia being able to sell the Titan for 1K if for 33% of its listing price you can have 60% of its performance? Yeah, you just discovered that the function between price and performance in a GPU stack is not linear, rather than a parabole (and Titan X, despite being a FP64 turd, still sells like hotcakes).

That slide only says 4GB max per stack, how many stacks can have a GPU sitting alongside the die again? We can just ask KNL and its propietary variant, cant we? :awe:

What makes you think Titan is selling in any significant volume that affects Nvidia's bottom dollar? AMD is financial trouble. They won't claw out of financial trouble selling 5,000 $850 graphics cards. They need a fresh mix of products between $200-500 that are a solid win in every regard.

An $850 graphics card trading blows with a $1000 graphics card will do nothing to fix AMD.

 

[desprado]

That's a very interesting article, it seems AMD is starting the journalist brief train already.

-----------------
"This first-gen HBM stack will impose at least one limitation of note: its total capacity will only be 4GB. At first blush, that sounds like a limited capacity for a high-end video card. After all, the Titan X packs a ridiculous 12GB, and the prior-gen R9 290X has the same 4GB amount. Now that GPU makers are selling high-end cards on the strength of their performance at 4K resolutions, one might expect more capacity from a brand-new flagship graphics card.

When I asked Macri about this issue, he expressed confidence in AMD's ability to work around this capacity constraint. In fact, he said that current GPUs aren't terribly efficient with their memory capacity simply because GDDR5's architecture required ever-larger memory capacities in order to extract more bandwidth. As a result, AMD "never bothered to put a single engineer on using frame buffer memory better," because memory capacities kept growing. Essentially, that capacity was free, while engineers were not. Macri classified the utilization of memory capacity in current Radeon operation as "exceedingly poor" and said the "amount of data that gets touched sitting in there is embarrassing."

Strong words, indeed.

With HBM, he said, "we threw a couple of engineers at that problem," which will be addressed solely via the operating system and Radeon driver software. "We're not asking anybody to change their games.""

That is why they delay it.More problems for AMD.

 

[flash-gordon]

I don't believe when they say "never bothered to put a single engineer on using frame buffer memory better". That's absolutely outrageous for a company that lives in market of minimal advantages.

I do believe they put more effort this time and it can work.

But, what I really believe is that frame times will make a comeback in reviews to test AMD words...