You have probably missed this
Why? HBCC should be possible with GDDR5x as well and hence the SSG SKUs.
EKWB is releasing a full cover waterblock for RX Vega:
https://www.ekwb.com/news/ek-is-releasing-full-cover-amd-radeon-vega-water-blocks/
I got in touch with them and they confirmed to me that an EK Fluid Gaming kit equivalent to the A240G (that is for GTX 1080/Ti) is coming soon for Vega too (it's their affordable full alluminium custom-loop-in-a-box)
No- its a full custom loop in a kit using aluminum parts (so they aren't to be mixed with any other copper or brass WC parts)so is that an AIO-type deal? I'd love to pair something like that with a Vega 56 if it is never released with water, but I'm not going to have a full water system at any point.
...if the mining rumors are true, though, I will likely never see one of these things. FFS.
not sure what you're looking atWhere are aluminium parts used? I can see all parts made of cooper.
not sure what you're looking at
from EK's page "Speaking with gamers around the world has given us a very clear picture that building a PC is a wonderful experience. Once you do it yourself and see how easy it is, you become addicted. But like all great things, building a PC requires a certain budget and usually, this budget is limited to a certain hardware price range. Custom liquid cooling tends to be expensive and only a small number of people can afford it. We wanted to change that! We found a way to make it all much more affordable with new aluminum technology that brings great performance for almost half the price of our core enthusiast lineup."
https://www.ekwb.com/news/ek-fluid-gaming-sets-new-standard-for-water-cooling/
Technical specifications
- Warranty:
2 years- Cold Plate Material:
Aluminum alloy- Radiator Material:
Aluminum alloy- Fitting Material:
Aluminum alloy
- Reservoir material:
Plastic- Socket support:
LGA-115x, LGA-2011(-3), LGA-2066, AM4- Tubing Dimensions:
10/13mm (ID/OD) - Clear
Veradun mentions the standard copper block and then mentions the aluminum kit coming with a Vega block....the next few posts are talking about the kit and krteq then follows with the aluminum question. I suppose he could be talking about the regular block but the context led me to believe it was about the kit since thats what the previous posts were about.That is an entirely different part than the one being talked about.
Using HBM2 was a mistake. Nvidia was smart to go with GDDR5X.
I don't think that going with GDDR5X fixes any of Vega's problems. At best maybe it launches sooner, but I think the rumors of HBM shortages were overblown. NVidia was shipping cards with HBM2 almost a full year before AMD.
It seems clear that Vega has had other issues holding it back rather than just waiting for memory.
AMD also needs HBM off the ground at some point for their APUs. It isn't magically going to get good without investment.
so is that an AIO-type deal? I'd love to pair something like that with a Vega 56 if it is never released with water, but I'm not going to have a full water system at any point.
...if the mining rumors are true, though, I will likely never see one of these things. FFS.
I'm already starting to see people saying Navi is going to be the first true architecture under Raja Koduri.People were hyping Vega around the time Fury X was released?
Every forum and website?
Are you really serious or just practicing your hyperbole?
In the first link you can see the full cover wb, that is copper or nickel plated copper.
In the following part I was speaking about the low cost aluminium kit, that includes a full cover wb for the video card and a wb for the CPU, pump/tank combo, tubes etc (https://www.ekfluidgaming.com/ek-kit-a240g) and it is sold for 250$ atm (geforce version) and it is coming for Vega too (confirmed directly to me).
I'm going to try and get a V56 for me at the first opportunity of a preorder. If I can, next step will be this kit. It costs a third of a copper loop, and it is good enough, and I will probably extend it with a second (aluminium) rad later just for the sake of it :>
Talking about something is not the same as hyping. Trying to explore possibilities is not hyping. A few individuals going to extremes is that person hyping. Try not to generalize.I'm already starting to see people saying Navi is going to be the first true architecture under Raja Koduri.
Even if that's true, it doesn't guarantee Navi is going to be the return to form.
Navi is starting to be hyped and its due in 2019?
AFAIK, HBCC works by using fine grained data transfers from main memory to the HBM2 cache while simultaneously feeding the GPU. This needs low latency and good bandwidth to work well. I remember reading that GDDR5X is very good in overall bandwidth but rather poor with small random transfers. First seen in the mining differences between GTX1070 and GTX1080.Why? HBCC should be possible with GDDR5x as well and hence the SSG SKUs.
The reason they needed HBM is to overcome architectural shortcomings as they need significantly more bandwidth then Nvidia cards. That worked ok for Fury which has quad stacked HBM1 for 512GB/s.
I suspect Vega was designed expecting quad stacked HBM2, run efficiently we'd then be having say 800+ Gb/s. Due to problems with HBM2 (probably cost) that wasn't possible so instead we end up with dual stacks and less bandwidth then a Fury. In addition to make up for lack of stacks it's been clocked too high ruining energy efficiency.
That said if they'd known they'd not have HBM I bet they would have spent more time sorting out their weak compression - HBM has just been a bit of a crutch meaning they could ignore that for now and work on other things.
This is wrong.The reason they needed HBM is to overcome architectural shortcomings as they need significantly more bandwidth then Nvidia cards. That worked ok for Fury which has quad stacked HBM1 for 512GB/s.
I suspect Vega was designed expecting quad stacked HBM2, run efficiently we'd then be having say 800+ Gb/s. Due to problems with HBM2 (probably cost) that wasn't possible so instead we end up with dual stacks and less bandwidth then a Fury. In addition to make up for lack of stacks it's been clocked too high ruining energy efficiency.
That said if they'd known they'd not have HBM I bet they would have spent more time sorting out their weak compression - HBM has just been a bit of a crutch meaning they could ignore that for now and work on other things.
Fury needs all 512GB/s to perform. Vega is fury at significantly higher clocks hence it needs a proportional increase in bandwidth to perform as well as Fury clock for clock.Why would they need 800GB/s???
That doesn't make any sense to me...
This is wrong.
HBM2 was always supposed to double HBM1 clocks thus doubling bandwidth/stack and lowering voltage simultaneously.
You don't design for 4 stacks and just change to 2 at the last minute. Each stack has its own PHY which has to be integrated with everything else, unless you're saying Vega has 4 HBM2 memory controllers but only using 2.
The present HBM2 stacks are clocked below what the spec specifies. Could be simply too much resistance with the micro-bumps bonding techniques presently used, thus higher voltage needed leading to higher temps.
I don't think it makes sense to have a HBCC without HBM memory, and HBCC appears to open a world of benefits to be exploited.
AMD really needs Samsung's HBM2 production.an interesting point was made by an executive of SK Hynix during a recent conference call: the company’s customers are willing to pay up to 2.5 times more for HBM2 versus what they paid for HBM1.
Is it per stack, or per GB?So they're paying 2.5X more per stack...
Which means 0.6x per GB.