Rx 490 indirectly confirmed

badsykes said:
http://www.guru3d.com/articles-pages/asus-radeon-rx-470-strix-gaming-4gb-review,1.html

Seems guru3d is more or less confirming the rx 490 card:


"Vega 10 - AMD Radeon RX 490 8GB (enthusiast class performance)
Polaris 10 - AMD Radeon RX 480 4GB and 8GB (2014 high-end performance)
Polaris 10 - AMD Radeon RX 470 4GB and 8GB (mainstream performance)
Polaris 11 - AMD Radeon RX 460 2GB/4GB (entry level)"
Click to expand...

There I fixed it.

If its 490 it probably won't have hbm2 than? (Like fury's not any 3xx card..)
 
QUOTE=atomBomb;1337883522]There I fixed it.

If its 490 it probably won't have hbm2 than? (Like fury's not any 3xx card..)[/QUOTE]

badsykes said:
http://www.guru3d.com/articles-pages/asus-radeon-rx-470-strix-gaming-4gb-review,1.html

Seems guru3d is more or less confirming the rx 490 card:


"Vega 10 - AMD Radeon RX 490 8GB (enthusiast class performance)
Polaris 10 - AMD Radeon RX 480 4GB and 8GB ( 2014 Enthusiast performance)
Polaris 10 - AMD Radeon RX 470 4GB and 8GB (mainstream performance)
Polaris 11 - AMD Radeon RX 460 2GB/4GB (entry level)"
Click to expand...

atomBomb said:
There I fixed it.

If its 490 it probably won't have hbm2 than? (Like fury's not any 3xx card..)
Click to expand...

fixed that for YOU.

hardly matters, performance level has dropped an entire price class, just like every generation.

There is just no current Enthusiast level replacement.

so what?
 
The Mac said:
QUOTE=atomBomb;1337883522]There I fixed it.

If its 490 it probably won't have hbm2 than? (Like fury's not any 3xx card..)
Click to expand...





fixed that for YOU.

hardly matters, performance level has dropped an entire price class, just like every generation.

There is just no current Enthusiast level replacement.

so what?[/QUOTE]

GTX 970 is "enthusiast" now? Funny, I think the 980 and 980 Ti owners would disagree. I know I would, having owned each of those 3 cards over the last couple years.
 
atomBomb said:
GTX 970 is "enthusiast" now? Funny, I think the 980 and 980 Ti owners would disagree. I know I would, having owned each of those 3 cards over the last couple years.
Click to expand...


970/980 were enthusiast class WHEN THEY WERE RELEASED. There was nothing faster other than the titan x which was a Halo product, not enthusiast class.

480/1060 same performance as 970/980/390/390x for much cheaper...

im not sure where your argument is here, of course moving down in the product stack equates to prior generation higher performance.

Stating its "2014 performance" is silly, of course it is.

2014 r9-280 performance is now at the 460 level.

It been that way since the beginning for GPUS.
 
Treeckcold57 said:
Vega 10 is supposed to be a RX 490 series with GDDR5. While Vega 11 is supposed to be a ultra enthusiast with HBM2.
Click to expand...

Have to say I'm disappointed with that statement. Fury launched with HBM and AMD clearly stated that was the future for enthusiasts cards. If that's so why release Vega 10 with GDDR5X? I can only think that HBM2 isn't near mass production despite what Hynix say and AMD are hurting badly by NV dominating the high end so they have to offer something to counter it. If that's true I think it is a retrograde step from AMD.

If they can't get HBM2 out of the door we might as well forget about Navi as that's gonna just be a pipe dream. As usual AMD can't execute their product launches to save their fu**ing lives. Come on AMD step up to the plate and deliver for once god dammit! Luckily my Fury Pro should keep my freesync monitor going for some time even if I need to lower the settings. Never bothered me before. Just sad that, as usual, AMD are getting pi**ed all over again.

Of course they might be pinning everything on Zen and letting RTG wait in the wings. If so, brave or foolhardy we'll soon find out.
 
Once again .. memory bandwidth is not an issue .. HBM is ahead of it's time and it isn't necessary for an enthusiast card.

GDDR5X is pushing out tons of bandwidth. It's not a big deal if Vega 10 doesn't have HBM.
 
You make it sound like AMD didn't make it clear from the get go that Vega was a q4 16 q1 17 product. Except that they have stated it since they announced their road map. Until we see a card hbm or not is speculation.

Sent from my VS990 using Tapatalk
 
I don't understand the obsession with HBM/HBM2 at all. I'm only interested in it because it makes more room for the core .. besides that, memory bandwidth is fine as it is with GDDR5X, at least for a bit. I'm in no rush for HBM or HBM2.
 
When Fury came out GDDR5X didn't exist yet. There was some question as to whether GDDR5 could ever reach those speeds. Thus, HBM made more sense.

Not to mention, HBM is hugely important for AMD, because it's a potential game changer for their APUs. Remember that not everything AMD does in graphics is focused only on their discrete cards.

Another thing to consider is that HBM has lower power use than GDDR5/X.
 
Vega 10 could have two stacks of 4 hi HBM2 memory running at ~400gb/sec or ~500gb/sec (depending which speed HBM2 memory used) with a 2048bit bus giving 8GB of memory space. Would be cheaper to make then 4 stacks of 2hi or 4hi(16gb memory). Faster HBM2 version would place the memory bandwidth a little bit faster than the Titan X Pascal. HBM2 is versatile. One stack of 4hi could be used for 4gb for lower end cards with a 1024bit bus, 250mb/s bandwidth.

Vega 10 maybe will have two stacks of HBM2 memory 4Hi 8GB, 500mb/s
Vega 11 maybe will have four stacks of HBM2 memory 4Hi 16GB, 1000mb/s

Vega is ready, now it is a matter of HBM2 and Interposer production.

I am not sure we will see a consumer version of Vega 11 for a while. Maybe a Pro-Consumer card later. If Vega 10 competes well against the Titan X Pascal then what will Vega 11 be like with 50% greater shaders? Double the memory bandwidth?

I am not sure where the 490 and Fury line will fall if there is going to be a Fury line (I hope so). Nvidia has 1070/1080, Vega 10? and TitanXP/1080Ti? Vega 11?
 
Nunz said:
I don't understand the obsession with HBM/HBM2 at all. I'm only interested in it because it makes more room for the core .. besides that, memory bandwidth is fine as it is with GDDR5X, at least for a bit. I'm in no rush for HBM or HBM2.
Click to expand...

Aside from offering alot more performance, power savings, and a smaller on die memory controller, its a pretty pedestrian upgrade.

The power savings are more than a few watts also, it was the only reason why the fury was able to to stay somewhat competitive in the power consumption area compared to the 980ti

While nvidia possibly doesnt need these advantages now, amd sure does.
 
Higgy10 said:
Aside from offering alot more performance, power savings, and a smaller on die memory controller, its a pretty pedestrian upgrade.

The power savings are more than a few watts also, it was the only reason why the fury was able to to stay somewhat competitive in the power consumption area compared to the 980ti

While nvidia possibly doesnt need these advantages now, amd sure does.
Click to expand...

A lot more performance? Compared to what? GDDR5X is pushing out more than enough bandwidth to satisfy 4K. Power savings are there and I agree, but we're talking ~30 watts? Nothing that makes me feel like it's absolutely necessary.

We still need more raw power. GDDR5X on a 384bit bus is more than enough to satisfy 4K. The only reason I'm even interested in HBM/HBM2 is because it allows more wattage for the core -- aka more power.
 
Nunz said:
A lot more performance? Compared to what? GDDR5X is pushing out more than enough bandwidth to satisfy 4K. Power savings are there and I agree, but we're talking ~30 watts? Nothing that makes me feel like it's absolutely necessary.

We still need more raw power. GDDR5X on a 384bit bus is more than enough to satisfy 4K. The only reason I'm even interested in HBM/HBM2 is because it allows more wattage for the core -- aka more power.
Click to expand...

Do the math on a 2048 or 4096 bit bus @ 500-1000mhz and compare to the ddr5x. Now whether or not its needed is another story of course but theoretical performance of hbm is considerably more.

Also i think its alot more than 30watts tbh, think about how much juice hawaii sucked at like ~2800 shaders vs fiji with ~4000 and a 275w tdp. I cant find the aritcle anymore but i believe it was something around 70w power savings roughly.

Also im pretty sure the on die memory controller is alot simpler and smaller thus giving more transistors/mm for more rops, cores etc.
 
Nunz said:
Once again .. memory bandwidth is not an issue .. HBM is ahead of it's time and it isn't necessary for an enthusiast card.

GDDR5X is pushing out tons of bandwidth. It's not a big deal if Vega 10 doesn't have HBM.
Click to expand...

OK Nuntz why did emperor Huang talk up last year that HBM2 was the big breakthrough for Pascal. See http://wccftech.com/nvidia-pascal-hbm2-16gb-1-tbs-2016/

GDDR5X is just GDDR5 on steroids and to be fair to Micron they saw an opportunity in the market because of lack of availability of HBM and stepped in with GDDR5X. It was a god send for Nvidia. They need to sell cards and if they couldn't get HBM they needed something else and hey presto GDDR5X came along. The 1080 is a stop gap card until they can get a card out with HBM2. Both vendors are betting their future on moving to HBM so it has to be a big deal.

Remember NVidia are refreshing their enthusiasts cards whereas AMD are refreshing their mid-range cards where GDDR5 makes sense. GDDR5 is just being resurrected to meet a short term need and is not something, in the high end, either company is committed to continuing.
 
HBM2 cost should come down and may not be that much more expensive then GDR5x. The card can be smaller, less power, simpler board offsetting some of the costs.

HBM2 stacks are very versatile in how hi and how many giving a broad range of memory bandwidth and capacity. This should allow a whole range of cards from HPC, ProConsumer, Professional down to mainstream using HBM2. I would not be too surprise that Vega 10/11 starts the 5xx series starting next year with the high end and working down over time to the lower end. The 490 may never exist in other words.
 
badsykes said:
I posted an article that is about HBM2 from tweaktown .... It looks objective...

http://www.rage3d.com/board/showthread.php?p=1337885562#post1337885562


enjoy
Click to expand...

Bandwidth requirements for AMD is higher than Nvidia since Nvidia is doing tile rendering with Maxwell and Pascal. Unless AMD changes how Vega renders they will need more memory bandwidth.

So for AMD to use DDR5x they would need most likely need a wider bus to keep up with Nvidia. HBM2 would solve the issue in the end and not have to really redesign GCN, drivers etc.

Since Nvidia prices went up dramatically this will probably help AMD with HBM2 on the high end allowing them to sell competitive products at around the same price. $1299 for a Titan XP does give ground for a high end Fury if it performs similar or better. Same with the 1080.

AMD just needs to figure out how to make a profit, a good profit with Vega. It may end up a whole new series starting on the high end working down as time progresses. Polaris may end up as a place holder, to get 14nm production going - make some money - get Zen out - then get Vega 5xx series out/FuryRage etc.
 
The Mac said:
970/980 were enthusiast class WHEN THEY WERE RELEASED. There was nothing faster other than the titan x which was a Halo product, not enthusiast class.

480/1060 same performance as 970/980/390/390x for much cheaper...

im not sure where your argument is here, of course moving down in the product stack equates to prior generation higher performance.

Stating its "2014 performance" is silly, of course it is.

2014 r9-280 performance is now at the 460 level.

It been that way since the beginning for GPUS.
Click to expand...

The argument is that as of today, when we are having this discussion, the 970 and 980 are not "enthusiast class" cards. They are no longer being manufactured or sold. Today's "enthusiast class" cards would be the 1070 and the 1080, and we all know they are way beyond the 480 and the 1060 in terms of performance.
 
You must log in or register to reply here.
Back
Top