Sell the 290 and get abit of cash while the mining craze is happening and put towards Vega. That's what I have done.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Poll: ** The AMD VEGA Thread **
- Thread starter TNA
- Start date
- Status
Sell the 290 and get abit of cash while the mining craze is happening and put towards Vega. That's what I have done.
Associate
- Joined
- 29 Jun 2016
- Posts
- 543
This is interesting:
http://wccftech.com/nvidia-volta-v100-gpu-tesla-v100-pci-express-announced/
I know it is Nvidia news and WCCFTech, however 14TFlops doesn't seem like much of a leap over the Vega's (Frontier) 13TFlops. Particularly so when Vega has had it's front end completely re-designed to remove the famed Fury X bottlenecks.
One thing to note, the article mentions V100 as having 5376 Cuda cores, which gives peak TFlops @ 1550MHz as 16.6TFlops, which is quite a bit above Vega. Remains to be seen what clocks they can achieve on a 21B transistor chip though
Obviously memory bandwidth may be Vega's biggest issue... Oh well, only 7? days to benches!
http://wccftech.com/nvidia-volta-v100-gpu-tesla-v100-pci-express-announced/
I know it is Nvidia news and WCCFTech, however 14TFlops doesn't seem like much of a leap over the Vega's (Frontier) 13TFlops. Particularly so when Vega has had it's front end completely re-designed to remove the famed Fury X bottlenecks.
One thing to note, the article mentions V100 as having 5376 Cuda cores, which gives peak TFlops @ 1550MHz as 16.6TFlops, which is quite a bit above Vega. Remains to be seen what clocks they can achieve on a 21B transistor chip though
Obviously memory bandwidth may be Vega's biggest issue... Oh well, only 7? days to benches!
This is interesting:
http://wccftech.com/nvidia-volta-v100-gpu-tesla-v100-pci-express-announced/
I know it is Nvidia news and WCCFTech, however 14TFlops doesn't seem like much of a leap over the Vega's (Frontier) 13TFlops. Particularly so when Vega has had it's front end completely re-designed to remove the famed Fury X bottlenecks.
One thing to note, the article mentions V100 as having 5376 Cuda cores, which gives peak TFlops @ 1550MHz as 16.6TFlops, which is quite a bit above Vega. Remains to be seen what clocks they can achieve on a 21B transistor chip though
Obviously memory bandwidth may be Vega's biggest issue... Oh well, only 7? days to benches!
That's the full chip with FP64 cores taking up room and power/heat.
The top gaming chip (either 2080 Ti or Titan Xv) will be the full 5376 cores, but with the 2688 FP64 cores stripped out.
This should run at 2 GHz (maybe a tad more, since the 12nm process may run higher clocks), so expect around 21-23 FP32 Tflops from the top Volta card once max-clocked.
AMD RX Chun-Li coming 2020.
Underboss
- Joined
- 20 Oct 2002
- Posts
- 33,778
- Location
- Oxfordshire / Bucks
I've actually been looking at buying one of these, although im quite happy with my ageing HD 7850, i could use the extra power
There is more than just FP64 cores that will be stripped, you also have 5376 INT 32 cores as well as all of the Tensor cores.
But even so, GP102 is already 470MM^2, so with ~50% more cores by going to 128 Cores per SM and 7SM per GPC, it will likely put GV102 at the reasonable yields limits of 550-600MM^2.
So they either fo with 6 SM per GPC to keep the size down and bump the base clocks. or they put it close to the limits and keep the same base clocks as now.
Permabanned
- Joined
- 12 Sep 2013
- Posts
- 9,221
- Location
- Knowhere
Well done
It's great how AMD do all the promoting to make the Vega name the buzzword around the tech industry, yet they're the only ones without an actual Vega product to sell.
What i just did today with my rx 480. Got 260 quid for it. Could perhaps get more if i had more patients and used the bay but it only took my 2 hours to sell it with a direct instant available bank transfer. Now i'll just wait for Vega while gaming on my old r9 290 which does a good enough job as placeholder. These mining peeps are crazy.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 31,179
In a way, the Star System.
Remember it's 12nm now, so smaller.
TSMC haven't said how much smaller exactly it is though. All they've said is "higher density, lower leakage, and lower cost" (than their 16nm)
I think it's reasonable to assume they can get it in the 550-600mm2 range, which should be fine since they've done chips that size before and the cost is lower per mm2 than 16nm.
I know this isn't much news, but Samsung gave a brief nod to vega during the EPYC conference.
Another thing to note is in the Anandtech breakdown of EYPC they said this: (http://www.anandtech.com/show/11551...7000-series-cpus-launched-and-epyc-analysis/2)
Does it potentially lend credence to the rumors about advantages with running Vega and Ryzen together? Will be interesting to see if anything comes from this.
Vega instinct confirmed for July release. 2.0Gbps HBM2 is MIA.
http://images.anandtech.com/doci/11562/1497994727970893916589.jpg
** Do Not Hotlink images **
http://www.anandtech.com/show/11562/amd-epyc-launch-event-live-blog-starts-4pm-et-
Another thing to note is in the Anandtech breakdown of EYPC they said this: (http://www.anandtech.com/show/11551...7000-series-cpus-launched-and-epyc-analysis/2)
Does it potentially lend credence to the rumors about advantages with running Vega and Ryzen together? Will be interesting to see if anything comes from this.
Vega instinct confirmed for July release. 2.0Gbps HBM2 is MIA.
http://images.anandtech.com/doci/11562/1497994727970893916589.jpg
** Do Not Hotlink images **
Last edited by a moderator:
Soldato
- Joined
- 26 Sep 2013
- Posts
- 10,756
- Location
- West End, Southampton
Wonder if I should sell both my 390's both with waterblocks and just use my 7950 for a few months.
I think the thing from Samsung is because they're supplying the HBM2 for the Apple version of Vega. The one in the iMac Pro with 400 GB/s bandwidth.
Also that Instinct MI25, that's passively cooled isn't it? If so that's 1501 MHz passively cooled, good sign?
I see where your coming from but they used the word "plus". Maybe i'm reading into it too much.
Yes and no. It relies on the fans in the server rack to draw air over the heat sink. Also the card is 1 slot high. where did you see clocks of 1501Mhz? and yes i think that is a good sign.
- Status