AMD Vega Architeture Speculation

I don't think AMD really needs HBM2, or even HBM1. I played around 290X memory over/underclocking, I found that lowering the clock to 1GHz (4GHz effective) which means 256GB/s (from stock 320GB/s) I don't lose much performance, nor does if I overclock to it's highest 1625MHz strap (giving 416GB/s) making much performance difference.

My estimation is AMD wanted to have experience with interposer manufacturing and apply it to it's APU in the future. From what I know building interposer is hard since it's huge (Fury interposer is 1011mm2, almost twice the size of Fury itself) and different from normal chip manufacturing.
 
I had read that upping the ram to 2250 from 2000 in the 480 gave a measurable increase in perf. It will depend on the game imo...
 
Anyhow, let's see Vega on the 25th and hopefully, it'll be both versions ( 10 and 11), even if they're still a couple of months away from retail release and only engineering samples are shown.


My largest concern is HBM 2.0 and if that really is what's still holding up the release.....Is it ready or isn't it?
 
shadow001 said:
Anyhow, let's see Vega on the 25th and hopefully, it'll be both versions ( 10 and 11), even if they're still a couple of months away from retail release and only engineering samples are shown.


My largest concern is HBM 2.0 and if that really is what's still holding up the release.....Is it ready or isn't it?
Click to expand...

What is happening on the 25th?

If there is a Vega Nano replacement that maybe my first buy since the Nano system has a capable CPU in it. For my main system I will need to update to Zen or Intel 2011 first to bother. I would rather AMD get Vega right, even if there is a slight delay.

I am not too impressed with the Pascal Titan X nor the price. Meaning the 1080Ti is a non-start since it will perform less than the TitanX. Obviously Nvidia was not too impressed either since they used the same name from last generation??? :confused:
 
noko said:
What is happening on the 25th?

If there is a Vega Nano replacement that maybe my first buy since the Nano system has a capable CPU in it. For my main system I will need to update to Zen or Intel 2011 first to bother. I would rather AMD get Vega right, even if there is a slight delay.

I am not too impressed with the Pascal Titan X nor the price. Meaning the 1080Ti is a non-start since it will perform less than the TitanX. Obviously Nvidia was not too impressed either since they used the same name from last generation??? :confused:
Click to expand...


Another capsaicin style presentation is what I heard, this time to present the Vega series.......First one as we all know, was when the R480 was officially presented.


So that is what could be done with a 150 watt max power budget, now let's see what they can cook up with a 300 watt maximum and hopefully using HBM 2.0.
 
Well if it's a Vega variant it's probably going to be the same as Pascal GP100 because of extremely limited HBM2 availability, I hope you have $129,000 laying around to get one.
 
Faceless Rebel said:
Well if it's a Vega variant it's probably going to be the same as Pascal GP100 because of extremely limited HBM2 availability, I hope you have $129,000 laying around to get one.
Click to expand...

Can't really compare, this is like saying Quadro is the same as a gaming consumer card.

DXG-1
2x 16 core Xeon Intel processors
8x P100 GPUs
512GB DDR4 memory
128GB HBM2
4x 2TB Samsung SSD
Multiple high bandwidth network
and a 3200w power supply.
 
Faceless Rebel said:
Well if it's a Vega variant it's probably going to be the same as Pascal GP100 because of extremely limited HBM2 availability, I hope you have $129,000 laying around to get one.
Click to expand...

Can't really compare, this is like saying Quadro is the same as a gaming consumer card.

DXG-1
2x 16 core Xeon Intel processors
8x P100 GPUs
512GB DDR4 memory
128GB HBM2
4x 2TB Samsung SSD
Multiple high bandwidth network
and a 3200w power supply.
 
Faceless Rebel said:
Well if it's a Vega variant it's probably going to be the same as Pascal GP100 because of extremely limited HBM2 availability, I hope you have $129,000 laying around to get one.
Click to expand...


I'd be just happy to know what the largest Vega is packing in terms of hardware, so that we know that the eventual gamer oriented version is worth waiting for ( or not )....


Now that Nvidia tipped it's hand regarding the titan-X and i'm sure reviews will follow soon enough after the 2nd of august, AMD will also know what they need to be competitive with, and by Nvidia's own estimates the titan-X should be roughly 30~40% faster on average than the GTX1080 already on store shelves.


Can Vega be in the same performance ballpark is what remains to be seen.
 
shadow001 said:
I'd be just happy to know what the largest Vega is packing in terms of hardware, so that we know that the eventual gamer oriented version is worth waiting for ( or not )....


Now that Nvidia tipped it's hand regarding the titan-X and i'm sure reviews will follow soon enough after the 2nd of august, AMD will also know what they need to be competitive with, and by Nvidia's own estimates the titan-X should be roughly 30~40% faster on average than the GTX1080 already on store shelves.


Can Vega be in the same performance ballpark is what remains to be seen.
Click to expand...
so that means 20 to 25% in real gaming and 10% in dx12 and vulkan


hope vega boosts GCN for dx12 and vulkan a lot over fury
 
shadow001 said:
I'd be just happy to know what the largest Vega is packing in terms of hardware, so that we know that the eventual gamer oriented version is worth waiting for ( or not )....


Now that Nvidia tipped it's hand regarding the titan-X and i'm sure reviews will follow soon enough after the 2nd of august, AMD will also know what they need to be competitive with, and by Nvidia's own estimates the titan-X should be roughly 30~40% faster on average than the GTX1080 already on store shelves.


Can Vega be in the same performance ballpark is what remains to be seen.
Click to expand...

HBM2 is availability in production? Granted, but the HBM2 won't be 100% ready until next year, so they will have more quantities and then Vega will be ready to get in production.
 
Treeckcold57 said:
HBM2 is availability in production? Granted, but the HBM2 won't be 100% ready until next year, so they will have more quantities and then Vega will be ready to get in production.
Click to expand...


Nvidia did surprise by having both a GP100 using HBM 2.0, and the GP102 variant using GDDR5x and found on this Titan - X card, so it can't be that hard to have a Vega variant also using the latter.


The signal protocol is the same between HBM 2.0 and GDDR5x in that both transfer 4 bits per clock, just that GDDR5 runs at much higher clock speeds over a narrower bus, there's no chip stacking involved and no need for an interposer either.



Unlike Nvidia, AMD does have experience with cards having a 512 bit memory bus, so in combination with GDDR5x, that yields upwards of 600 GB/sec of memory bandwidth.....Leave HBM 2.0 for Vega once the kinks are ironed out......



That chart AMD showed in the last capsaign showed they were going to introduce a new memory technology for the Vega architecture afterall.
 
WTF?.....


"NVIDIA is preparing to launch its flagship graphics card based on the "Pascal" architecture, the so-called GeForce GTX TITAN P, at the 2016 Gamescom, held in Cologne, Germany, between 17-21 August. The card is expected to be based on the GP100 silicon, and could likely come in two variants - 16 GB and 12 GB. The two differ by memory bus width besides memory size. The 16 GB variant could feature four HBM2 stacks over a 4096-bit memory bus; while the 12 GB variant could feature three HBM2 stacks, and a 3072-bit bus. This approach by NVIDIA is identical to the way it carved out Tesla P100-based PCIe accelerators, based on this ASIC. The cards' TDP could be rated between 300-375W, drawing power from two 8-pin PCIe power connectors. The GP100 and GTX TITAN P isn't the only high-end graphics card lineup targeted at gamers and PC enthusiasts, NVIDIA is also working the GP102 silicon, positioned between the GP104 and the GP100. This chip could lack FP64 CUDA cores found on the GP100 silicon, and feature up to 3,840 CUDA cores of the same kind found on the GP104. The GP102 is also expected to feature simpler 384-bit GDDR5X memory. NVIDIA could base the GTX 1080 Ti on this chip."
Click to expand...


Up to 3 versions?

One with GDDR5x and using GP102.<----- already shown
One with 3 HBM 2.0 stacks and using GP100.
Ones with 4 HBM 2.0 stacks and using the same GP100 <----This is the highest end one of them all.



https://www.techpowerup.com/223895/nvidia-to-unveil-geforce-gtx-titan-p-at-gamescom


It's nothing official, but still.
 
I'm skeptical the HBM versions are going to be coming any time soon, especially not for gaming. Between the combination of poor yields on the large chips and the seeming rarity of HBM 2, the card would be way outside of the price range that even enthusiasts would be willing to pay. It would probably be another $3000 "professional" card.

I think they'll hold those back until closer to when Vega is coming out.
 
So the AMD Polaris is graphic IP8 which is the same as Hawaii, Tahiti, and Tonga. However, the Vega is the most likely will be next graphic IP9 since the GCN architecture will be completely overhauled. I just need to get a Sapphire Nitro RX 480 then keeping it for a few months until Vega finally comes out. I like to see the how much Vega improved the tessellation performance with its new GCN architecture. I wonder how much ROP will it going to have?
 
Shadow for real? that has nothing to do with VEGA rumors. At least pretend to be on topic ffs. Pascal does not belong in this thread unless you are talking perf or arch comparisons and even then id argue it should go in other gfx section.

Also vega isn't introducing new memory Navi is.
 
Gandalfthewhite said:
Shadow for real? that has nothing to do with VEGA rumors. At least pretend to be on topic ffs. Pascal does not belong in this thread unless you are talking perf or arch comparisons and even then id argue it should go in other gfx section.

Also vega isn't introducing new memory Navi is.
Click to expand...


And what new memory is that since HBM 2.0 is supposed to come with Vega?
 
You must log in or register to reply here.
Back
Top