AMD Vega Discussion Thread

[TheoneandonlyMrK]

Nope. The green led indicator was for all fury owners.

It seems really hard for some people to resist the urge to curse and attack. Not you Xzibit, the one above you. Being civil and keep discussion on matter is hard i guess.

Keep it on topic, your having a laugh your power save led ,r9,390s what the fecks on topic about that, its just your other moaning about Amd thread.

 

[Caring1]

Somehow we are both yawning .....

Reading some threads has that affect on me.

 

[Kanan]

Chill is actually a pretty good feature that runs very well. Just not in every game. It should run perfectly fine for the games it supports and decrease power consumption by a lot.

 

[m1dg3t]

Keep it on topic, your having a laugh your power save led ,r9,390s what the fecks on topic about that, its just your other moaning about Amd thread.

Exactly, he just wants to piss & moan about AMD, it's hilarious.

Lets see if this post gets reported/deleted by some special snowflake/s

Tools.

 

[Kanan]

Some more info about "Radeon Chill", most people don't seem to know it - maybe slightly off topic, but anyway:

Starting at about 13:30 it's mind boggling how he *halves* power consumption from 120 W to 60 W just by activating Chill. And the game still runs perfectly fine. It works so well, that it even reacts to the speed of the mouse pointer. As you move the pointer faster, it increases FPS to match its speed.

 

[R4E3960FURYX]

Waiting for more info about AMD RX Vega family at Computex Taipei 2017.

Maybe there are many interesting details for discuss.

 

[xkm1948]

More "possible" VEGA benchmark coming. It is the same guy as last time: 枫叶之血

This is the same card as the one for Time spy ~5900 last time. 1200MHz core and 700MHz HBM2

Again take this with whatever exotic salt you prefer. And don't turn this into feces throwing contest again.


http://www.3dmark.com/fs/12296284

Comparing the score with my 1105MHz FuryX

Can this be the actual new Nano??

 

[R4E3960FURYX]

Yep, this is RX Vega Nano 8GB HBM2 Core Clock 1200MHz 410 MHz HBM2.
Cost around 359-399$ arrive on June 2nd.

Technically R9 Fury X gain at most bandwidth @600MHz HBM1 OC (1200MHz Efficiency)
If you set R9 Fury X Core 1200MHz , R9 Fury X could easily perform a lot better than
RX Vega Nano 8GB HBM2.

 

[FordGT90Concept]

Vega Nano only being 1.1-10.7% faster than Fury X? I hope not.

 

[uuuaaaaaa]

More "possible" VEGA benchmark coming. It is the same guy as last time: 枫叶之血

This is the same card as the one for Time spy ~5900 last time. 1200MHz core and 700MHz HBM2

Again take this with whatever exotic salt you prefer. And don't turn this into feces throwing contest again.


http://www.3dmark.com/fs/12296284

Comparing the score with my 1105MHz FuryX

View attachment 87616

Can this be the actual new Nano??

I saw this earlier and I got disappointed... Fury X performs almost identically at 1100MHz (at least in fs)...

 

[Caring1]

If this is the "baby vega" then we have some good stuff to look forward to from it's bigger siblings.

 

[HD64G]

Does anyone know if this latest score is from a 4096-shader GPU? Might be a cut one. A little patience is needed now. We all might learn much more in a week or so.

 

[uuuaaaaaa]

Does anyone know if this latest score is from a 4096-shader GPU? Might be a cut one. A little patience is needed now. We all might learn much more in a week or so.

Not sure, because ipc seems to be lower than Fiji XT...

 

[RejZoR]

On what planet would that make any kind of sense?

 

[R4E3960FURYX]

Maybe uncle raja going to sell the peanut with salt Vega based GPU for us.

Why not combine 2304 x 2 = 4,608 shader 288TMU 64ROP with 16GB GDDR5 as " RX490/RX590 " better?

The weak point of HBM usages was overclock range from stock clock only 5-10% capable.
However switching to GDDR5 usage overclock range could at least 15-25% capable from stock both GPU/VRAM that make sense for all AMD fans.

According to Radeon Pro 32GB GDDR5 model price tag 999$. AMD could setting 499-549$ price tag for
RX 490/590 within 4,608 shader 288TMU 64 ROP 16GB GDDR5 easily.

Only 1.5GHz base 1.7GHz boost may wipe out all nvidia counter parts at all.

RX Vega maybe pointless.

 

[RejZoR]

They could go multi-GPU, make smaller cheaper GPU's and stack them. Then just ensure profiles for CrossfireX are reliably delivered to gamers. It would be a win-win. Coz it's cheaper to make small GPU and stack them together than one massive GPU.

I know they are planning that for Navi afaik, but that's still so far away. Plus, crossfire scalling is excellent, always has been lately.

 

[Vya Domus]

Vega is already kind of pointless to AMD , but not from that perspective. They've already missed all the windows for a proper launch against the high-end 10 series. That's why they are so silent about it , not because it sucks or it's a monster , it's simply not that important anymore to them.

Vega unfortunately still looks more of a prototype than a serious product line because of that. Although , thankfully , it is a step up from the Fury X's strategy.

 

[RejZoR]

AMD will certainly have to narrow the release time gap when countering NVIDIA. Current cycle is totally off and it's hurting their sales for sure when NVIDIA is always the first one to come with brand new tech.

 

[john_]

If this is a Vega chip with 4096 cores, then probably the chip is ready for months. They just delay it because they have an ongoing debate about what gone wrong and if there is really any point in releasing this to the market.
At 14nm they better come with a 6000 cores chip in the high end, and 1.4GHz frequency at *only* 250W TDP, or we can only hope Ryzen to produce enough income(from their last financial report this gets a huge question mark) for them to become competitive in 1-3 years from now.

 

[R4E3960FURYX]

If those benchmarks leak around the webs were RX Vega 8GB HBM 2 base products. It was big mistake of RTG teams.

High bandwidth cache may not efficiently for graphics solution when we compare with those Intel optane memory act as system L4 cache with system HDD.

Bandwidth of DDR4 pools with HBM2 may become stuttering when realtime gaming run.

 

[FordGT90Concept]

Remember this?
http://www.tweaktown.com/news/52253...0-cf-beats-titan-leaked-benchmarks/index.html

RX 480 only got 17536 points at 1,949 MHz, a 54% overclock (ridiculous). 67DF:C4 is like the actual Polaris 10 chip so what the hell is that 67DF:C7? If history repeats, the Timespy, FireStrike, and Compubench that were leaked are probably the Nano version of Vega. I wouldn't be surprised if 1200 MHz is its peak. In other words, Vega does not impress.

 

[xkm1948]

My prediction based upon the latest 3 yrs of AMD GPU release suggests that VEGA might not meet my expectations of a 80Ti killer.

I would suggest you be more cautious GT90, some extreme fanboys might be super insulted right now.

 

[FordGT90Concept]

 

[RejZoR]

Sorry, but a GPU released 2 years later with roughly the same performance as R9 Fury X just makes absolutely no sense at all. Maybe AMD isn't the top dog when it comes to engineering (or marketing) at the moment, but they still aren't that stupid. They had a projection prototype for sure and if that showed essentially the same performance as R9 Fury X they'd scrap the whole thing back then already. And if they for some reason are just going with it despite those findings, then AMD has some serious issues. And I don't think they do.

I mean, lets be realistic, when you change so many things in the rendering pipeline of the graphic card and still somehow yield same performance as R9 Fury X, things just don't add up for the leaks.

Not to mention, if they wanted to keep 3DMark scores 100% private, they could run the thing offline and still get the numbers. So, this must be some sort of baiting or something...

 

[64K]

We will beat you like a red-headed step child if you are wrong.