• 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 **

On or off the hype train?

  • (off) Train has derailed

    Votes: 207 39.2%

  • (on) Overcrowding, standing room only

    Votes: 100 18.9%

  • (never ever got on) Chinese escalator

    Votes: 221 41.9%
  • Total voters
    528
Status
Not open for further replies.
Chuk_Chuk said:
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.
Click to expand...

Hmmm I didn't notice that bit in that way at first. But it does seem to imply some advantage doesn't it?

I wonder if it just has something to do with Infinity fabric helping Vega specifically for compute tasks? Or if it's something more fundamental like using HBM to boost the speed of the infinity fabric interconnect?

There was speculation about that already, since infinity fabric is scaled off RAM on AM4. Could they make it faster using HBM in some configuration?
 
AllBodies said:
Hmmm I didn't notice that bit in that way at first. But it does seem to imply some advantage doesn't it?

I wonder if it just has something to do with Infinity fabric helping Vega specifically for compute tasks? Or if it's something more fundamental like using HBM to boost the speed of the infinity fabric interconnect?

There was speculation about that already, since infinity fabric is scaled off RAM on AM4. Could they make it faster using HBM in some configuration?
Click to expand...
Anandtech, mentions that with EPYC, IF connects two CCX dies together across multiple CPUs.

Socket-to-socket communication is designed at the die level, rather than going through a singular interface. One die in each processor is linked to the same die in the other processor, meaning that for the worst-case scenario data has to make two hops to reach a core or memory controller on the other side of the system. Each link has a bidirectional 37.9 GB/s bandwidth, which is only slightly less than the intra-socket communication bandwidth, although we would expect socket-to-socket to have a slightly higher latency based on distance. AMD has not shared latency numbers at this time.

http://images.anandtech.com/doci/11...ss_and_analysts_06_19_2017-page-073_575px.jpg
Click to expand...

The key thing is they mention giving access to the memory controller on the Chip with two hops. I'm wondering if:
1. Using IF AMD could essentially tie together a compute unit to a CCX core. So that it can receive instructions direct from CPU at a hardware level. If possible there are quite a lot interesting things they could pull off.
2. Using IF with HBCC allows data fetch requests to be sent in a more direct manner. So send an instruction to the CPU which send an instruction to the memory controller to send data to the GPU. I'm assuming that currently there would need to be a software layer to make such a request possible?
 
TonyTurbo78 said:
Wonder if I should sell both my 390's both with waterblocks and just use my 7950 for a few months.
Click to expand...
Dont know how much you would be able to get for them. Polaris seems to be all the rage right now when it comes to mining and i only sold cause i could get enough so if vega is turning out to be to expensive/unimpressive i could just buy a brand new rx 580 without loosing money.
 
Gotta hand it to AMD, the Instinct Vega cards looks really nice!

1dv1pOraSt2p86mbXHhYQQ.png
 
AllBodies said:
Someone on the AMD reddit has bought the water-cooled Vega Frontier edition, and says they'll test a ton of things then report back:

https://www.reddit.com/r/Amd/comments/6if94c/ill_let_you_guys_know_how_the_rx_vega_might_stack/
Click to expand...

Interesting!

Release date is listed as 1st July for both of the Frontier cards.

Specs listed are the same as the previous leak. Looks like it was legit.

Air TDP 300W
Liquid TDP 375W

Air & Liquid
16GB
4096 Stream Processing Units
FP32: 13.1 TFLOPS
FP16: 26.2 TFLOPS
 
Last edited:
five8five said:
Interesting!

Release date is listed as 1st July for both of the Frontier cards.

Specs listed are the same as the previous leak. Looks like it was legit.

Air TDP 300W
Liquid TDP 375W

Air & Liquid
16GB
4096 Stream Processing Units
FP32: 13.1 TFLOPS
FP16: 26.2 TFLOPS
Click to expand...

I'm really perplexed with the common TFLOPS number for both air/liquid.

These numbers fully-match the air-cooled MI25, so it makes sense that the air-cooled Vega would have them. But shouldn't the liquid one clock a bit higher?
 
akarypid said:
I'm really perplexed with the common TFLOPS number for both air/liquid.

These numbers fully-match the air-cooled MI25, so it makes sense that the air-cooled Vega would have them. But shouldn't the liquid one clock a bit higher?
Click to expand...
I agree. At least it matches the other info leaked so looks legit. Is it the fastest card or is it the slower one? :)

The other things the 75W TDP difference.
 
Not sure if there has been a UK price listed on the forum?

Page is now gone but it had the Air Cooled FE Version at £1139.99. Release date unavailable.
 
Last edited:
akarypid said:
I'm really perplexed with the common TFLOPS number for both air/liquid.

These numbers fully-match the air-cooled MI25, so it makes sense that the air-cooled Vega would have them. But shouldn't the liquid one clock a bit higher?
Click to expand...

The MI25 card has been officially announced as 12.3 Tflops FP32, so 1501 MHz.

13.1 Tflops is 1600 Mhz.

And the TDP difference is weird between the water and air cooled founders editions. I can only assume the 1600 Mhz clock is for the air cooled one, and they just haven't listed the water cooled one.

And so the 13.1 Tflops figure is a "From 13.1 Tflops".
 
AllBodies said:
The MI25 card has been officially announced as 12.3 Tflops FP32, so 1501 MHz.

13.1 Tflops is 1600 Mhz.

And the TDP difference is weird between the water and air cooled founders editions. I can only assume the 1600 Mhz clock is for the air cooled one, and they just haven't listed the water cooled one.

And so the 13.1 Tflops figure is a "From 13.1 Tflops".
Click to expand...

I don't think so. The air cooled one is up to 13,1Tflops which will be Boost clock. The MI25 is designed to hold the clocks, but only manages to get 1500 mhz at 300W TDP. The Clocks of the Air Fe will probably often fall under 1600mhz and be in the range of ~1500mhz, while the water cooled version will run 1600mhz all the time.
 
akarypid said:
I'm really perplexed with the common TFLOPS number for both air/liquid.

These numbers fully-match the air-cooled MI25, so it makes sense that the air-cooled Vega would have them. But shouldn't the liquid one clock a bit higher?
Click to expand...

The MI25 is 1 TFLOPS less.
Also the 375W on the watercooled version only means it has unlocked TDP as it can take abuse with that cooler.
Raja also stated there are 8+6 and 8+8 pin versions.
 
Samwell said:
I don't think so. The air cooled one is up to 13,1Tflops which will be Boost clock. The MI25 is designed to hold the clocks, but only manages to get 1500 mhz at 300W TDP. The Clocks of the Air Fe will probably often fall under 1600mhz and be in the range of ~1500mhz, while the water cooled version will run 1600mhz all the time.
Click to expand...

The MI25 is a passively cooled card, i think its obvious even the FE will have higher clocks not to mention AIB cards
 
Status
Not open for further replies.
Back
Top Bottom