surge from PSU failing may have damaged riser. Ive had a riser cause that issue before. Easy quick check is swap 2 cards and see if problem follows
2 Likes
That is such a good freaking idea. Iām going to replace the riser and try it out tomorrow. Thank you.
This info is true but just a little clarification to avoid confusions.
Sata and Molex cables usually are exactly the same and can handle a lot more, the power limit is not in the cable is in the connector.
I use molex connectors for my risers, but there is not a standard in how many watts they can handle thats why people always recommend using only pcie.
And of course i will never use SATA
2 Likes
Bro, howās your Vega 56 Red Devil - Hynix? Would you mind to share stable setting and straps? Itās been a headache to manage this cardsā¦ in ETH preferably. In few hours, TRM detected dead on the card.
Grateful for your help.
Cheersā¦
best i can do for my hynix red devil vega is 46.83 MHS (stable) with settings
CORE : 1000, VDD : 850, MEM : 910
straps :
amdmemtweak --CL 20 --RC 36 --RP 13 --WR 11 --CWL 7 --FAW 12 --RAS 23 --REF 65535 --RFC 164 --RTP 6 --RRDL 4 --RRDS 3 --WTRL 4 --WTRS 4 --RCDRD 19 --RCDWR 11
if the card crash after a few hours usually i add more VDD (increment 10)
also donāt have the patience for this card 
I end up selling it to my friend
my friend settings :
CORE : 920, VDD : 800, MEM : 907
straps :
amdmemtweak --CL 19 --RC 36 --RP 13 --WR 11 --CWL 7 --FAW 12 --RAS 23 --REF 65535 --RFC 249 --RTP 6 --RRDL 4 --RRDS 3 --WTRL 4 --WTRS 4 --RCDRD 19 --RCDWR 11
achieve 46 MHS 94 watt
thanks broā¦ will try that settingā¦
Careful with that.
I always learned that you needed to power a single card from the same source, so both risers and cardās power connectors from the same PSU.
If you use 2 different power sources units, they might give (slight) differences that might ruin your cards.
Donāt know how much of it is true, but I havenāt taken any risk.
I thought it was even merely 45W.
Anyway, not a single SATA in my rigs due to that reason!
Also only few Molex; most are 6/8 pin PCIE connectors.
Maybe it has a different hiveos version or kernel?
Hi bro, not to nag, but this is an ERGO specific topic; there are already 3 other topics with a lot of Vega info for ETH:
1 Like
Ya I would avoid molex but even a 1300W EVGA PSU only has 6 PCIE 8 pins. I would need 2 8 pins for 4 risers so I would only be able to power 4 cards for a 1300W PSU. Havent ever heard of issues with molex but no one can seem to find a spec that definitively says it can handle XXX power. Since people ābelieveā it is 145W I just do one molex per riser and even if they are off I am set. I have a 50% margin of safety at max draw so I am ok with that.
Also, you power the riser and GPU with the same PSU mainly so if one goes out you are not causing a voltage differential at the riser to GPU interface. Also important that the two PSU are connected to same phase outlet. Breakers use two out of phase 120V inputs and that is how you can get 240V (american standard of course). This means different breakers are out of phase and have 240V potential to each other. I use a riserless mobo and would fry something if one of the two PSU were out of phase or one off and one on, cant adhere to the standard. I use a common ground that shuts them both off if one isnt on. Need a PhD to do this stuff i swear
1 Like
Im inclined to believe you as they always failed at the connector, where resistance will be the highest, but I am not sure how a connector could have less capacity than the wires they support. Iām not educated enough to argue one way or another, just not sure how functionally that would be.
that set up is on a rig I bought locally, Ive thought about rewiring it but that guy built it that way in 2018 and its still running fine. I donāt like that set up because I usually unplug the server PSU when I am changing cards on it and yesterday I booted the rig with the server PSU unplugged which gave me a scare.
The ones I build that have a extra server PSU have risers and GPUs on same PSU, plus I use a different breakout board that has a cable connecting the Server PSU to the Mobo so it turns on and off with the main PSU.
1 Like
Anyone knows if AMD driver version makes any difference on hashrate?
based on my experience, no. Just stability but thats really only for windows
1 Like
Came back here to mention - I really think Vegaās are such solid mining cards right now. Used to get 5700ās too, but they are stuck really just to ETH at 57MH but 150W wall. Vega does 54MHs at 150W as well, not a huge difference in performance. I also have a hell of a time getting 5700ās stable, very fickle. But push come to shove and vega can swap to ERGO and do so very well. Few other cards are viable in two major coins. Even 3000 series nvidia have hash limiters making swapping between ETH and alts not as efficient, albeit they are better assets should crypto cease to exist.
totally agree, Im building a 5700 rig just to sell and its taken me about a week to get 4 5700s stable. Vegas are great, two coins they are good on now, and hard enough to learn that you can buy them for $500-600. Iāve bought 13 used vegas locally in the past 6 months.
I did get my hands on a FE 3080 last weekend, from the store for $750 (camped out in line over night lol), and that is way better but no way I could get 13 of those for that price.
1 Like
alsoā¦ been working on dropping wattage a little bit, to make room for extra cards.
You can run Samsung at 846 mem and still push 182-183 with effective core at 1300.
At 846 mem I can also run 64s flashed to 56 with this strap:
amdmemtweak --RC 36 --RP 11 --WR 14 --CWL 8 --FAW 12 --RAS 20 --REF 65535 --RFC 248 --RTP 4 --RRDL 6 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 10
846 memory auto sets SOC at 847 so that, compared to 847 mem & 860 SOC, let me turn down vdd 6-12 on some cards.
*on WoolyPooly so rejected = āStale Sharesā
4&6 are bad cards they run effective core at -70 so just tuned them down to drop wattage.
Hey everyone, I have been lurking for a good while and finally decided to post to see if I can get some help tuning in these Vega cards. One is a power color Vega 56 red dragon with hynix memory and the other was a power color Vega 64 red dragon that I reflashed to a 56 bios. Im running both of these in hiveos and using team red miner. I canāt get the hashrate over 170 on either card. I have tried multiple oc settings and multiple straps as well. What I have posted in the pic is where I have got them to run best so far. Any advice on getting these tuned up for some more hasrate and less power would be greatly appreciated!
Vega 56 hynix
Core 1300
Core voltage 990
Mem clock 847
Soc freq 860
Strap is amdmemtweak --CL 21 --RC 36 --RP 13 --WR 13 --CWL 7 --FAW 12 --RAS 23 --REF 65535 --RFC 248 --RTP 6 --RRDL 5 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 20 --RCDWR 12
Vega 64 flashed to 56 Samsung mem
Core 1300
Core voltage 990
Mem clock 847
Soc freq 860
Strap amdmemtweak --CL 21 --RC 36 --RP 13 --WR 13 --CWL 7 --FAW 12 --RAS 23 --REF 65535 --RFC 248 --RTP 6 --RRDL 5 --RRDS 3 --WTRL 9 --WTRS 4 --RCDRD 20 --RCDWR 14
your Samsung is at its power limit so is essentially throttling, try dropping vdd to like 900 or less to reduce below the 185 power limit. Then can turn up core until you like the hashrate.
So if the Samsung is at is power limit would the hynix be as well?
No I think its same bios you can run command amd-info from your browser to see the power limit but that bios has a 185 limit.
if you run that command it will also show you your effective core which im guessing is like 1240 on the Hynix, might just need 40-60 more core to hit 180.
You could also probably drop the RCDRD in that strap my Hynix needs 14 RCDRD on ETH but runs fine at 11 on ERGO.
I dropped the RCDRD on the hynix to 11 and now itās at 182mh with all the same oc settings as previous. I also changed the Samsung card to 1325 and 900mv and there was no change. Think I should go more on the core still?