Friday, March 31st 2017

AMD's RX Vega to Feature 4 GB and 8 GB Memory

It looks like AMD is confident enough on its HBC (High-Bandwidth Cache) and HBCC (High-Bandwidth Cache Controller) technology, and other assorted improvements to overall Vega memory management, to consider 4 GB as enough memory for high-performance gaming and applications. On a Beijing tech summit, AMD announced that its RX Vega cards (the highest performers in their next generation product stack, which features rebrands of their RX 400 line series of cards to th new RX 500) will come in at 4 GB and 8 GB HBM 2 (512 GB/s) memory amounts. The HBCC looks to ensure that we don't see a repeat of AMD's Fury X video card, which featured first generation HBM (High-Bandwidth memory), at the time limited to 4 GB stacks. But lacking extensive memory management improvements meant that the Fury X sometimes struggled on memory-heavy workloads.

If the company's Vega architecture deep dive is anything to go by, they may be right: remember that AMD put out a graph showing how the memory allocation is almost twice as big as the actual amount of memory used - and its here, with smarter, improved memory management and allocation, that AMD is looking to make do with only 4 GB of video memory (which is still more than enough for most games, mind you). This could be a turn of the screw moment for all that "more is always better" philosophy.
Add your own comment

52 Comments on AMD's RX Vega to Feature 4 GB and 8 GB Memory

#1
RejZoR
The fact they dare to release RX Vega in 4GB configuration is telling me they are REALLY confident in HBC capability.

EDIT:
I'm just wondering how they are doing it. How they prioritize what goes into on-board memory and what goes into system RAM. Is it game dependent or is it fully on-the-fly, do they have to make game profiles, this is the stuff I'm wondering the most about. Because if they can achieve this fully on-the-fly without any profiles, just a really intelligent algorithms (maybe assisted with driver updatable algorithms on software level) that could be really sweet. It could mean the end of obsessive stacking of insane amounts of memory on graphic cards and also give them more cost flexibility since they could focus more of the production cost on actual GPU instead of on VRAM memory modules. They aren't cheap, especially not in HBM2 flavor.
Posted on Reply
#2
Nkd
rx vega will not be coming with 4gb of memory. You can bet your ass on that. There might be lower end models that in the 300 dollar range that might use 4gb. But top end models will sure be 8gb. No way AMD puts 4gb on enthusiast parts. They will just get chewed out by negative feedback regardless of how it performs. People will always be critical of it. There is a lot of room left for amd to have a few cards remember? They have nothing above rx480 that can be had for 200. Everything above is still open. You may get 299, 399 and 499 models.
Posted on Reply
#3
Nkd
Also can you post a link to source of video where they said it would be 4 gigs of cards for high end gaming? You seem to be assuming here what they mean. How do we not know that they might be talking about low end server parts here?
Posted on Reply
#4
theGryphon
Nkd said:
rx vega will not be coming with 4gb of memory. You can bet your ass on that. There might be lower end models that in the 300 dollar range that might use 4gb. But top end models will sure be 8gb. No way AMD puts 4gb on enthusiast parts. They will just get chewed out by negative feedback regardless of how it performs. People will always be critical of it. There is a lot of room left for amd to have a few cards remember? They have nothing above rx480 that can be had for 200. Everything above is still open. You may get 299, 399 and 499 models.
I believe cut down versions, definitely sub-300, will come in 4GB, while the top models will come with 8GB. Cut down versions would still be technically Vega-arch, so yeah, there will be some Vega parts with 4GB HBM2...
Posted on Reply
#5
RejZoR
Nkd said:
rx vega will not be coming with 4gb of memory. You can bet your ass on that. There might be lower end models that in the 300 dollar range that might use 4gb. But top end models will sure be 8gb. No way AMD puts 4gb on enthusiast parts. They will just get chewed out by negative feedback regardless of how it performs. People will always be critical of it. There is a lot of room left for amd to have a few cards remember? They have nothing above rx480 that can be had for 200. Everything above is still open. You may get 299, 399 and 499 models.
RX480 and RX580 might have 8GB, but they only have that and no more. RX Vega with HBC can address all the memory you have in the system. In my case that would be 16+ GB of always free RAM and 8GB on-board. Not even GTX 1080Ti or Titan X Pascal has that. Once it goe spast it's 11 or 12GB, it's done. Granted, it's a lot and you'll rarely hit that, but with RX Vega, you don't even have to worry about that.

You have to think outside of box for Vega. The old "more memory on graphic card is better" doesn't apply here anymore because it's not following old methods or ways of doing it.
Posted on Reply
#6
qubit
Overclocked quantum bit
Everything I'm seeing suggests that they're finally coming up to par with NVIDIA. Let's hope they actually achieve this and give them some decent competition for a change.
Posted on Reply
#7
RejZoR
I sure hope so :) All the stuff I've read about RX Vega sounds really exciting, especially HBC thingie. It's easy to just cram more of everything on graphic card. But coming out with such solutions, that's how we make huge leaps. They used to cram wider memory bus on cards. And with some ingenuity, we now went back from 512bit down to 384 and 256bit thanks to memory compression without sacrificing any speed. HBC is no different really. It's out of the box thinking. And it's why I'm so excited about Vega and why GTX 1080/1080Ti was like meh for me the entire time.
Posted on Reply
#8
Readlight
Why evry time when you mov your head in game texstures is loading slow on rx 460?
Posted on Reply
#9
INSTG8R
My Fury hasn't choked on anything yet at 1440. I still max out most anything. New stuff like Wildlands and ME:A have forced me down a notch tho.
Posted on Reply
#10
Filip Georgievski
HBC is a very good tech for the new VEGA GPUs.
The fact that they still use 4 to 8GBs on these cards tells us that these card will be VRAM efficient.
The 4GB models will aim 1440p and the 8GB models will aim 4K since there is no need for more.
Gaming engines also are becoming more efficient in VRAM usage as well so im guessing that AMD will target VRAM limits on these cards to cut down on price, but give us 512bit bus and HBC + HBM Memory to be fast and efficient in modern AAA titles.
Im hoping for AMD to be better competitve so Nvidia will cut down on price and ALL FANBOYS TO BE HAPPY WITH BETTER AND CHEAPER GPUS.
Posted on Reply
#11
RejZoR
Texture streaming in modern engines has solved a lot of constraints. Essentially, that's HBC right there already, on a game level. Game knows it has 20GB of textures, but for the player position, it only requires 4GB of them. And as you move through level, it's preparing, parsing and unloading them as you move around and textures change. It may fluctuate depending on player location.

Problem with texture streaming is that you're essentially doing a VRAM+HDD/VRAM+SSD instead of something a lot faster. And Vega's HBC with VRAM+RAM (+SSD) could certainly address that far better in same way how CPU addresses it's memory hierarchically. L1 cache is VRAM. L2 is RAM. L3 can be SSD. Because texture streaming still causes hitching, stuttering and framerate lag when doing texture streaming the way current game engines do it (VRAM+HDD) because it's parsing textures from really slow medium.
Posted on Reply
#12
HD64G
INSTG8R said:
My Fury hasn't choked on anything yet at 1440. I still max out most anything. New stuff like Wildlands and ME:A have forced me down a notch tho.
Those 2 games will run much better after 2-3 months after being patched out properly. It always happens with big titles.
Posted on Reply
#13
Bytales
My 1080 SLI is going to be replaced by two DUAL VEGA Card with 8GB per GPU, hopefully theill do just that. So im sittin tight waiting for that.
Posted on Reply
#14
cryohellinc
First they promised 16, now max 8, fascinating. And "better allocation" sounds fishy.
Posted on Reply
#15
3rold
4GB. This has to be a joke. What year is it? 2008?
Posted on Reply
#16
ADHDGAMING
RejZoR said:
The fact they dare to release RX Vega in 4GB configuration is telling me they are REALLY confident in HBC capability.

EDIT:
I'm just wondering how they are doing it. How they prioritize what goes into on-board memory and what goes into system RAM. Is it game dependent or is it fully on-the-fly, do they have to make game profiles, this is the stuff I'm wondering the most about. Because if they can achieve this fully on-the-fly without any profiles, just a really intelligent algorithms (maybe assisted with driver updatable algorithms on software level) that could be really sweet. It could mean the end of obsessive stacking of insane amounts of memory on graphic cards and also give them more cost flexibility since they could focus more of the production cost on actual GPU instead of on VRAM memory modules. They aren't cheap, especially not in HBM2 flavor.
They dont they are leaving it up to the AIBs this time when it comes to the reference cards. If they want the reference Models but want theirs to be 4GB models only, they can do that. My guess would be that we wont see many AIBs wanting rebadged Reference models running 4GB, Mind you it is HBM but the average consumer will see 4GB as 4GB not the 6.5 or 7GB equivilent of GDDR5.
Posted on Reply
#17
ADHDGAMING
INSTG8R said:
My Fury hasn't choked on anything yet at 1440. I still max out most anything. New stuff like Wildlands and ME:A have forced me down a notch tho.
Same, Fury X is a highly slept on BEAST. 3 Monitors each 1440p, games running DX12 or anywhere close to well optimized just work like a dream
Posted on Reply
#18
RejZoR
3rold said:
4GB. This has to be a joke. What year is it? 2008?
Apparently your mentality is stuck in 2008 because you clearly don't understand High Bandwidth Cache and it's meaning to VRAM capacity. Go and watch AMD's presentation of Deus Ex Mankind Divided running maxed out on RX Vega with memory artificially limited to only 2GB VRAM (and unlimited outside of RX Vega).

ADHDGAMING said:
They dont they are leaving it up to the AIBs this time when it comes to the reference cards. If they want the reference Models but want theirs to be 4GB models only, they can do that. My guess would be that we wont see many AIBs wanting rebadged Reference models running 4GB, Mind you it is HBM but the average consumer will see 4GB as 4GB not the 6.5 or 7GB equivilent of GDDR5.
They are not leaving anything up to AIB's. AIB's have no control over the GPU and since HBM2 is on the same interposer as GPU which then gets shipped to AIB's, they can only stick this interposer to their PCB. They can't do anything else. What AIB's can do is to fiddle with power delivery, PCB design and cooling. As for average consumers, average consumers don't buy 700€ graphic cards. They buy RX460 with 16GB VRAM for 200€. Large majority of those who will buy RX Vega know what HBC is and what benefit it brings even at only 8GB VRAM capacity.
Posted on Reply
#19
GhostRyder
Well Vega sounds very interesting, but we really need to see it and its been kinda a drag without a high end GPU on the table.
Posted on Reply
#20
JMccovery
cryohellinc said:
First they promised 16, now max 8, fascinating. And "better allocation" sounds fishy.
When did AMD promise a consumer GPU with 16GB of HBM2?

I know it was said that the increased number of layers per stack will allow them to release HBM2 products with 16GB and 32GB, but I never saw AMD promise that those capacities would be on consumer products.
Posted on Reply
#21
Captain_Tom
Nkd said:
rx vega will not be coming with 4gb of memory. You can bet your ass on that.
The top end ones likely won't (Ones that compete with the 1080 Ti).


However there would be no problem at all if the one meant to compete with the 1070 - 1080 did. With their new tech 4GB = 8GB. Literally.
Posted on Reply
#22
Captain_Tom
JMccovery said:
When did AMD promise a consumer GPU with 16GB of HBM2?

I know it was said that the increased number of layers per stack will allow them to release HBM2 products with 16GB and 32GB, but I never saw AMD promise that those capacities would be on consumer products.
Thank you. Some people just can't read lol.


I wouldn't be surprised if someone like SAPPHIRE released a toxic edition with 16GB of VRAM, but 8GB standard would be more than enough.
Posted on Reply
#23
FordGT90Concept
"I go fast!1!11!1!"
If they're serious about 4 GiB, I'm going to cry in an insane manner. There's already benchmarks where Fury X is clearly memory limited. I can understand not offering 16 GiB but 4 GiB is asinine.
Posted on Reply
#24
Captain_Tom
FordGT90Concept said:
If they're serious about 4 GiB, I'm going to cry in an insane manner. There's already benchmarks where Fury X is clearly memory limited. I can understand not offering 16 GiB but 4 GiB is asinine.
Have you paid any attention to what's going on?


AMD claims to have new tech that allows for half as much memory to be used as normal. If that's true, then a 4GB model is more than viable. In fact I would say this kinda confirms that AMD is confident it will work. They wouldn't do it otherwise...
Posted on Reply
#25
Fluffmeister
It'll just be nice when they release the fucking thing.

I've already bought the candles for the GTX 1080's first birthday.
Posted on Reply
Add your own comment