CES AMD livestream
- Thread starter pax
- Start date
If the card has 2 8pin plugs it's likely a power sucker. examples: every card released with 2 8pinlol
both my 1080 ti and 2080 ti have 2 8pin plugs
both my 1080 ti and 2080 ti have 2 8pin plugs
You’re going to tell me both those cards don’t draw high wattage? Both your cards are also non-reference PCBs .. not a valid comparison.
pax
Well-known member
Simple auto undervolt now from wattman which tells me amd was simply lazy when it came out with its initial driver. Or they were rushing a part out and overvolted it to produce more working parts from a given batch of gpus.
Maybe they would have lost some of the production parts not working to lower volts but due to high BOM costs they couldnt stand to lose some fraction of the gpus to it. Im sure Nvidia did and does the same thing only they mass produced it and didnt care to lose some chips to low volts right off the bat.
Once amd gets off hbm and gets into the same ballpark in BOM as Nvidia they will be fine in the power to perf dept.
Nagorak
Well-known member
Is your Vega newer? My belief is that the initial production run required higher voltage to get acceptable yields, and later cards undervolt a lot better. My Sapphire Vega 64 Nitro+ can run at close to max speed with only 950mv (default is 1200 mv)--like a negligible performance drop for cutting voltage by 20%.
I think Vega would have been viewed a lot more favorably if it had released running at 950 or 1000 mv rather than 1200 mv. I hope that Vega VII is released at a more reasonable voltage, but I don't want to hold my breath.
I think Vega would have been viewed a lot more favorably if it had released running at 950 or 1000 mv rather than 1200 mv. I hope that Vega VII is released at a more reasonable voltage, but I don't want to hold my breath.
pax
Well-known member
Ya its a recent buy from last month but a reference model so not sure when it was made. Could be left over reference cooler or they kept on making them.
Core Volt is about 1.0... but core speed of 1560 with spots on 1600 was kept same. Core volt does fluctuate IIRC but max watts from hwinfo was 221.
Core Volt is about 1.0... but core speed of 1560 with spots on 1600 was kept same. Core volt does fluctuate IIRC but max watts from hwinfo was 221.
NWR_Midnight
New member
It has nothing to do with when it was made. It's all about Silicone Lottery. There are all kinds of people who have had similar results with the very first Vega 64's, yet there are other's who have little success. I've had mine running as low as 220 watts under full load at stock speeds, but I didn't put it under water to worry about how much power it draws, or to not overclock it. It pulls about 260 to 290 watts (game dependent) running with the core at 1650 Mhz under full load at 1.15 volts. (stock is 1530 Mhz average) , and memory at 1100 Mhz (stock is 945 Mhz)
I agree with you, however the statement was that all dual 8pin cards have been power suckers. Vega is definitely a power sucker unless you undervolt it yourself. At stock settings, it sucks the juice. The card has dual 8pin for a reason.
If you think the efficiency of the chips doesn't get better as the process becomes more and more mature, you crazy
. Yes, some people have had great results with early adoption Vega64's, that is silicone lottery - however the cards that have been made a year after will absolutely be more matured, efficient chips. Nature of the beast on that one.
NWR_Midnight
New member
If you think the efficiency of the chips doesn't get better as the process becomes more and more mature, you crazy
For efficiency to change, they have to make modifications to the process and the design aka revisions. Neither was done, as it was Go-Flo doing the chip manufacturing and we all know Go-Flo sucks, and we know AMD didn't make any new design change to the Vega 64 You don't just get more efficient chips by making more following the same identical process and design. Now, if you where comparing different versions.. As an example: Ryzen 1 vs Ryzen+, that is a demonstration of the process becoming more mature which is all in correlation to changes they did to the process and the design. Non of that was done with the Vega 64 on the reference boards, with no revisions being done that we know of.
We’ve seen this done before without official “revisions” being released. As the process becomes more mature (by doing it over and over) the amount of successful wafers produced gets higher - thus more selective with binning etc...
Nagorak
Well-known member
Yields often improve over time. Intel has some graphs of this with various process nodes. Obviously I can't say for certain this happened at GloFo, but I just assumed that it probably would.
To be clear, this is only speculation on my part. There's no practical way to actually prove it without having data to hundreds of cards produced at different points in time. I suppose AMD or GloFo would have data on the yields, but I don't expect they'll be telling us.
I'll agree with you that Silicon lottery is still going to play a part in any case.
NWR_Midnight
New member
More yields doesn't necessarily mean better effienciency or less power draw. It just means they get less faulty chips per wafer. The better binned chips are used in the liquid cooled versions of the card which are clocked higher, as well as higher voltages, hence more stable, which also have a higher power draw.