If this is a Instinct of Pro then that bodes well for the RX Vega cards since the pro cards are usually lower clocked. A 1600MHz Fiji would be between a 1080 and 1080Ti and Vega is supposed to be much improved architecture. If this leak is true then even the low end Vega should be 1600MHz or more which is exciting.
-
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.
AMD VEGA confirmed for 2017 H1
- Thread starter Doom112
- Start date
- Status
If this is a Instinct of Pro then that bodes well for the RX Vega cards since the pro cards are usually lower clocked. A 1600MHz Fiji would be between a 1080 and 1080Ti and Vega is supposed to be much improved architecture. If this leak is true then even the low end Vega should be 1600MHz or more which is exciting.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 31,179
Yup, the costing for 16GB of (scarce!) HBM2 would be pretty insane...
Not scarce, missing! SK Hynix are only showing 4GB 1.6GBps modules in their catalogue so far, which means Instinct/Radeon Pro needs 4 Modules per interposer to get 16GB.
The Vega die shot shown, only shows 2 modules.
However let's have a pause for thought and something intriguing.
AMD said that Vega and Ryzen would make an ecosystem if both are together.
The Ryzen cpu is SOC. Which means that would have direct access to the gpu and the HBM2.
If the 16gb HBM2 cache can be accessed by the CPU SOC, that would affect also the infinity fabric, greatly improving the perf on both components.
AMD said that Vega and Ryzen would make an ecosystem if both are together.
The Ryzen cpu is SOC. Which means that would have direct access to the gpu and the HBM2.
If the 16gb HBM2 cache can be accessed by the CPU SOC, that would affect also the infinity fabric, greatly improving the perf on both components.
apparently the Pro Duo only shows specs for a single GPU in this test. 64CU and 4GB of memory.
https://compubench.com/device.jsp?b...pi=cl&D=AMD+Radeon(TM)+Pro+Duo&testgroup=info
But due to the memory size it is likely to be an Mi25 part.
Those are what they publicly have available, if AMD have a contract with them for 2Gbp/s HBM2 for 8 and 4GB stacks, then they are unlikely to show on their public product portfolio.
https://compubench.com/device.jsp?b...pi=cl&D=AMD+Radeon(TM)+Pro+Duo&testgroup=info
But due to the memory size it is likely to be an Mi25 part.
Not scarce, missing! SK Hynix are only showing 4GB 1.6GBps modules in their catalogue so far, which means Instinct/Radeon Pro needs 4 Modules per interposer to get 16GB.
The Vega die shot shown, only shows 2 modules.
Those are what they publicly have available, if AMD have a contract with them for 2Gbp/s HBM2 for 8 and 4GB stacks, then they are unlikely to show on their public product portfolio.
Me neither. Don't want a blower type anyway. If I won the lottery my RIG would have SLI 1080Tis in it. (even though im not a fan of multi GPU, if money is no object you know).
Soldato
- Joined
- 18 May 2010
- Posts
- 23,010
- Location
- London
If there was some sort of 1+1=3 factor to be gained from having Ryzen and Vega combo then I am sold!!
HSA need to be levaraged by software developers, the heterogeneous computing is more likely to come to fruition in the pro market rather than desktop
Problem is 8GB Hynix modules weren't listed for production until end of the year.
There's also the possibility than AMD has opted to go to Samsung for supply larger modules for Instinct; similar to what NVIDIA has been doing for Tesla.
Even the Tesla V100 only has 900GB/s Bandwidth, which if they were using SK Hynix, would be 1024GB/s.
If that's the case I don't see why they wouldn't bother slapping 4 modules of Samsung 4GB on there for Instinct cards. The cost of 4 modules isn't an issue for Enterprise; and still leaves the original design for 2 4GB modules of the consumer gaming cards.
I do hope the lack of 2.0Gbps Hynix modules on the catalogue does mean AMD has first rights to it all, as without it they can't meat the Vega 10 specification of 512GB/s bandwidth using two modules.
Their Catalogue no longer lists 2.0Gbps ( To Be Determined ), and 1.6Gbps moved to Q1 2017 for availability.
https://videocardz.com/65649/sk-hynix-updates-memory-product-catalog-hbm2-available-in-q1-2017
Here's the link directly to the SK Hynix PDF on what they have available.
https://www.skhynix.com/static/filedata/fileDownload.do?seq=374
between glofo and hynx, AMD cannot catch a break
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 31,179
between glofo and hynx, AMD cannot catch a break
****ed either way
It really does feel that way. if SK Hynix lived up to their production ETAs we'd probably have Vega already; never mind GloFo.
I just hope the top Vega card we get isn't a compromise because of their partner's failing to live up to promises.
At least the 1600Mhz on a possible enterprise part is promising a bit; as those are usually clocked lower thn high end gaming equivalents.
Permabanned
- Joined
- 12 Sep 2013
- Posts
- 9,221
- Location
- Knowhere
Definitely.
http://wccftech.com/amd-radeon-rx-v...lineup-includes-3-skus-priced-at-599-499-399/
$599 - Radeon RX Vega Nova
$499 - Radeon RX Vega Eclipse
$399 - Radeon RX Vega Core
$599 - Radeon RX Vega Nova
$499 - Radeon RX Vega Eclipse
$399 - Radeon RX Vega Core
Why wouldn't they? Hynix were originally advertising 2.0GBps and only more recently dropped the 2,0 for 1.6Gbps chips. AMD undoubtedly do have a contract to course HBM2 form Hynix but that doesn't stop Hynix selling GBM2 to other parties The only 2 liekly scenarios is that Hynix simply can't make 2.0Gbps chips yet, or the volume is so small they can't supply AMD and other parties sufficiently. IF Hynix had good supply they absolutely would be marketing 2.0Gbps chips and selling them to other custoemrs.
WCCF links to TPU and TPU quote tracks back to reddit post.
Names don't make sense, they don't make a range.
Chevy Vega, Chevy Nova, Mitsubishi Eclipse and the whole Intel Core range
No.
And prices leave a gap between Core and RX580
Guess we'll hear something concrete tomorrow
- Status