RX Vega Owners Thread

ccityinstaller said:
You have been a member here how long? I don't need to prove anything to you, there are plenty of members on this board that can vouch for me in a personal and professional capacity. In your rush to prove me wrong, you forgot a very important point in that wall of text. I am running a 75% PowerLimit, which combined with the lower voltage and the fact my card stays under 35C under load (and I believe its a very good sample just as JustReason has) allows my card to stay at 1750Mhz all day every day.

I've actually been able to sustain 1775Mhz clocks but I have to bump the voltage up the point it is useless for a less then 1% improvement. If you don't believe me, look at my FireStrike Ultra and Superposition scores, they are in VEGA Owner's thread. Then go find some results on the web showing any other VEGA significantly out performing my card. I will be here waiting ;).
Click to expand...

Wow, I have been a member here a whopping 3 years less than you 2007 vs 2010, that makes a big difference, and the fact that you started with that point, shows you are grasping and now have resorted to bs attacks to imply you are not full of crap. I did not disregard your power limit settings (I've had mine up to 142) all power limit does is allow it to go past the P7 states when needed, to use more power when needed up to 75%. So, when you under volt, you are using less power UNTIL you need to go above the P7 voltage limits and it allows it to use more power up to 75% more to reach higher core clocks (as long as a user doesn't not run into thermal throttling, which you won't). IT does not reduce power draw under full load. P7 power limit set to 1050 mv and +75 power limit is equivalent to stock settings of 1200 mv and +50 power limit. However, with it set to 1050 mv, you will substain lower clock speeds than if you where 1200 mv. You can sit and spin all day about the 7 years I have been a member here, it doesn't change the fact that I am calling bs on your numbers. You very well might be sustaining 1750Mhz core clock speeds, but no way in hell you are doing it at 286 watts (GPU only) or less under full load.
 
NWRMidnight said:
Wow, I have been a member here a whopping 3 years less than you 2007 vs 2010, that makes a big difference, and the fact that you started with that point, shows you are grasping and now have resorted to bs attacks to imply you are not full of crap. I did not disregard your power limit settings (I've had mine up to 142) all power limit does is allow it to go past the P7 states when needed, to use more power when needed up to 75%. So, when you under volt, you are using less power UNTIL you need to go above the P7 voltage limits and it allows it to use more power up to 75% more to reach higher core clocks (as long as a user doesn't not run into thermal throttling, which you won't). IT does not reduce power draw under full load. P7 power limit set to 1050 mv and +75 power limit is equivalent to stock settings of 1200 mv and +50 power limit. However, with it set to 1050 mv, you will substain lower clock speeds than if you where 1200 mv. You can sit and spin all day about the 7 years I have been a member here, it doesn't change the fact that I am calling bs on your numbers. You very well might be sustaining 1750Mhz core clock speeds, but no way in hell you are doing it at 286 watts (GPU only) or less under full load.
Click to expand...

Whatever you say ;). I'm going to Rising Storm 2. If you do not like what I have to say, then feel free to add me to your ignore list.
 
ecmaster76 said:
A)Vega works just like any other card. Better silicon, which he may have, could have plenty of headroom at lower voltages. Lower voltages uses less power so you wont hit your power limit even if you manage to max out the clocks that are stable at that voltage. Pascal usually never throttling max boost clocks is a perfect example of this concept in action
B)A UPS having stable voltage regulation has jack to do with how good a ammeter it has on the AC lines. They need one just good enough to provide a reasonable estimate of remaining battery runtime and what not so they wont spend too much on them
Click to expand...

Dude, you are off base on just about all of what you said. google vega 64 overclocking, read up on it, might want to read up on under volting as well YOU DON"T over clock VEGA 64 by controlling core speeds as most cards have in the past. you control over clocking by voltage, because of the way it dynamically controls core speeds which it does by voltage and thermal, P states are ONLY low limits of that state, meaning it can and it will go well above the p7 state core clock if it still has voltage head room, and thermal head room. Undervolting on Vega 64 will reduce your core clocks. If you are under water, thermal is taken out of the equation, which leaves you voltage and your silicon that determines Max core speeds. yes, he might have great silicon, but that isn't really my point. My point was, it is impossible for him to be running at 1750 Mhz sustained, and only drawing 286 watts or less at full load.


As for UPS, you are only thinking about AC power coming in, where most quality UPS, specially business class ones, are also line conditioners, and to maintain proper, clean steady power to the devices. They can't do this properly or be even reliable if they don't have reliable accurate readings on the load as well.
 
Last edited:
ccityinstaller said:
Whatever you say ;). I'm going to Rising Storm 2. If you do not like what I have to say, then feel free to add me to your ignore list.
Click to expand...


I really don't care. I just don't like false claims being made. And the fact that i showed you how wrong you are, you can't even man up and admit it. Your loss, not mine. Enjoy your game.
 
There is a HUGE variation on what Vega dies can do for voltage and power usage, so this cross talk is kinda moot. We already saw that based on reviewers and that is a small sample, expect it to be even more varied with general users.
 
NWRMidnight said:
I really don't care. I just don't like false claims being made. And the fact that i showed you how wrong you are, you can't even man up and admit it. Your loss, not mine. Enjoy your game.
Click to expand...
Actually you are not right. Not sure where you are getting your info but the power limit is not entirely based on volts, but TDP, as in Watts and temp, Likely VRMs, amps/current, and other power algorithms.

Case of fact: In fire strike Using just stock settings and Balanced profile, my LC Vega 64 will stay in Pstate6. If all I do is reduce the voltage 20mV, it will go to Pstate7 (which is 1750mHz) and run about 1720mHz. So based on your theory what happened shouldn't have happened. But the fact is voltage only plays a part as far as the TDP and amp/current algorithms used for power target %. So in fact reducing the voltage and maintaining stability of the clock can result higher sustained clocks yet reduced power usage.
 
JustReason said:
Actually you are not right. Not sure where you are getting your info but the power limit is not entirely based on volts, but TDP, as in Watts and temp, Likely VRMs, amps/current, and other power algorithms.

Case of fact: In fire strike Using just stock settings and Balanced profile, my LC Vega 64 will stay in Pstate6. If all I do is reduce the voltage 20mV, it will go to Pstate7 (which is 1750mHz) and run about 1720mHz. So based on your theory what happened shouldn't have happened. But the fact is voltage only plays a part as far as the TDP and amp/current algorithms used for power target %. So in fact reducing the voltage and maintaining stability of the clock can result higher sustained clocks yet reduced power usage.
Click to expand...


It can but it really depends on the chip too. Voltage vs stability has a lot to do with the silicon (architecture) and node.

And we know Vega has a wide bell curve when it comes to voltages......
 
NWRMidnight said:
I really don't care. I just don't like false claims being made. And the fact that i showed you how wrong you are, you can't even man up and admit it. Your loss, not mine. Enjoy your game.
Click to expand...
Undervolting, 56 CU vice 64 of Vega 64, water cooling the whole card at below 35c, good die -> Very much possible I would say. A Vega 56 I would think would have a better chance at sustaining 1750mhz at the right conditions. My Vega 64 LC which does not like to be undervolted much can sustain over 1700mhz with an aggressive Power Limit. Now I would be looking at over 500w from the wall with a +20 power limit. This argument reminds me of the many I had with my Nano where folks just don't believe or don't understand what I was saying (maybe my fault), in that I could sustain FuryX speeds in virtually every game and it would not throttle etc. - Just realize some variations, even huge ones do occur. Speaking of the Nano, it is working with two Gigabyte 1070 G1's ( Each having 3 fans) in a mining rig, open style case - guess which card is the loudest/noisiest of the bunch? The Nano is the most quiet of the three with a single fan :D - go figure - yes all three indicate up to 100% usage at the same time mining.
 
NWRMidnight said:
<snip> My point was, it is impossible for him to be running at 1750 Mhz sustained, and only drawing 286 watts or less at full load.


As for UPS, you are only thinking about AC power coming in, where most quality UPS, specially business class ones, are also line conditioners, and to maintain proper, clean steady power to the devices. They can't do this properly or be even reliable if they don't have reliable accurate readings on the load as well.
Click to expand...

I don't know that its impossible by any means. He may not be 100% accurate since ~50w for fans+pump sounds kinda high to me. He may also be overestimating the CPU wattage. So lets say 325w for Vega. If he really lucked out its doable.
I'm talking about both ends of the UPS. At the end of the day its just constant voltage/frequency AC power supply. Amperage changes in response to load to keep the waveform as close to ideal as possible. Actual amperage measurements only get used for run time measurements, etc. Obviously their will be protections from overcurrent in the form of a fuse or whatnot. Protection from overheating will be temperature controlled; a unit with good ventilation in cool ambient air will be able to handle more current than a unit under less ideal conditions
 
JustReason said:
Actually you are not right. Not sure where you are getting your info but the power limit is not entirely based on volts, but TDP, as in Watts and temp, Likely VRMs, amps/current, and other power algorithms.

Case of fact: In fire strike Using just stock settings and Balanced profile, my LC Vega 64 will stay in Pstate6. If all I do is reduce the voltage 20mV, it will go to Pstate7 (which is 1750mHz) and run about 1720mHz. So based on your theory what happened shouldn't have happened. But the fact is voltage only plays a part as far as the TDP and amp/current algorithms used for power target %. So in fact reducing the voltage and maintaining stability of the clock can result higher sustained clocks yet reduced power usage.
Click to expand...

No, I am not wrong on how the power limit works. I already stated if it has the thermal head room, and does not get throttle due to temperature it increase the core clock speeds until you hit the max voltage set in the p7 state + power limit, it hits the thermal limits, or crash because your chip can't handle the speeds.. Power limit has NOTHING to do with TDP, unless you hit the TDP ceiling of card, which means it will throttle before you ever hit the voltage limits.

Your own "case of fact" shows you are not hitting your p7 state as you claim. IF you where, you would be hitting 1750 Mhz, NOT 1720 Mhz.. the 1750 Mhz you have is the MINIMUM it would run if you where indeed hitting the P7 state. (If you have the voltage head room, and thermal head room, it will clock much higher) Not sure how you figure 1720Mhz is hitting the P7 state when you haven't even hit the lowest core clock it will run in that state.
 
ecmaster76 said:
I don't know that its impossible by any means. He may not be 100% accurate since ~50w for fans+pump sounds kinda high to me. He may also be overestimating the CPU wattage. So lets say 325w for Vega. If he really lucked out its doable.
I'm talking about both ends of the UPS. At the end of the day its just constant voltage/frequency AC power supply. Amperage changes in response to load to keep the waveform as close to ideal as possible. Actual amperage measurements only get used for run time measurements, etc. Obviously their will be protections from overcurrent in the form of a fuse or whatnot. Protection from overheating will be temperature controlled; a unit with good ventilation in cool ambient air will be able to handle more current than a unit under less ideal conditions
Click to expand...

His cpu is over clocked to 5+ Ghz, I doubt he is over estimating what it pulls. Also, he is supposedly getting near 25% increase speeds while only increasing power draw by 30% with his Vega 64. If it was an Intel cpu, I could by it, but we are talking about AMD Vega, the most power hungry card on the planet.. there is no way in hell his power draw is scaling that good. also, even if he over estimates his fans/pump, he does not figure in motherboard/memory/hard drive power draw, so it is a wash in that sense.
 
NWRMidnight said:
No, I am not wrong on how the power limit works. I already stated if it has the thermal head room, and does not get throttle due to temperature it increase the core clock speeds until you hit the max voltage set in the p7 state + power limit, it hits the thermal limits, or crash because your chip can't handle the speeds.. Power limit has NOTHING to do with TDP, unless you hit the TDP ceiling of card, which means it will throttle before you ever hit the voltage limits.

Your own "case of fact" shows you are not hitting your p7 state as you claim. IF you where, you would be hitting 1750 Mhz, NOT 1720 Mhz.. the 1750 Mhz you have is the MINIMUM it would run if you where indeed hitting the P7 state. (If you have the voltage head room, and thermal head room, it will clock much higher) Not sure how you figure 1720Mhz is hitting the P7 state when you haven't even hit the lowest core clock it will run in that state.
Click to expand...
That was one example and mine clocks to 1800+ on its own with the clock set to 1772. Voltage has not a thing to do with it other than how it plays into TDP and VRMs. There are plenty of examples of this behavior all over the net eg: people undervolting and maintaining higher clocks, even in some reviews. You are basically assessing that all Vega64s require exactly 1.2V to run their boost clock of 1750 when in FACT they do not as I have stated when looking at all the results users have posted.

Besides my FS example does in fact hit Pstate7. Real easy to tell as Pstate6 has lower voltage settings and mine is using the setting I have for Pstate7. The same applies to Superposition. These are supposed to hammer your GPUS so I would not expect extreme clocking behavior as I get in other games.

At any rate his power usage seems very realistic given the setting he is using and my own findings recently allow for his to be believable.
 
How many of you guys are using aftermarket water cooling? Any threads online about waterblocks? I'm not coming up with much with my google skillz....
 
NWRMidnight said:
His cpu is over clocked to 5+ Ghz, I doubt he is over estimating what it pulls. Also, he is supposedly getting near 25% increase speeds while only increasing power draw by 30% with his Vega 64. If it was an Intel cpu, I could by it, but we are talking about AMD Vega, the most power hungry card on the planet.. there is no way in hell his power draw is scaling that good. also, even if he over estimates his fans/pump, he does not figure in motherboard/memory/hard drive power draw, so it is a wash in that sense.
Click to expand...

Then what is the power draw, voltages and speed of you Vega 56 / 64?
 
Smitty2k1 said:
I would also be curious.
Click to expand...

I am using the EK copper+aectal block with some IC Diamond TIM...I usually run CL Liquid Metal but with the cost of this card and the fact that the HBM and GPU are on the interposer I did not want to risk the LM running somehow. My temps are super cold, but I have a ton of RAD space and the ambient temperature of my room is 67~69F. I've left this system looping Prime95 and the GPU fully loaded for over a day and I've never seen it go above 35C core, with the hotspot peaking 5-6C higher for brief moments.
 
ccityinstaller said:
I am using the EK copper+aectal block with some IC Diamond TIM...I usually run CL Liquid Metal but with the cost of this card and the fact that the HBM and GPU are on the interposer I did not want to risk the LM running somehow. My temps are super cold, but I have a ton of RAD space and the ambient temperature of my room is 67~69F. I've left this system looping Prime95 and the GPU fully loaded for over a day and I've never seen it go above 35C core, with the hotspot peaking 5-6C higher for brief moments.
Click to expand...

I've read a few posts that clocks don't go much higher, even though temps have plummeted (Just like most processes today....). Have you found that to be true? Going water only makes sense if you want lower temps/quiet?

BTW. Thanks for the info!
 
Last edited:
DogChainX said:
I've read a few posts that clocks don't go much higher, even though temps have plummeted (Just like most processes today....). Have you found that to be true? Going water only makes sense if you want lower temps/quiet?

BTW. Thanks for the info!
Click to expand...

I can run 1800Mhz all day sustained if I keep my voltage @ 1.2V (which is the stock P7 state voltage) and a 65% Power Limit but Power draw is insane for the small return. I run 1750Mhz all day and it chews through every game I have, and FreeSync is just icing on the cake.

I've been a PC Gamer (3D wise) since I was 12 and I have never had such a smooth game experience. AMD has done an incredible job driver wise.
 
Any more info on this?

http://www.tomshardware.com/reviews/sapphire-rx-vega-64-nitro,5388-3.html

"In some games, we noticed that Sapphire's Radeon RX Vega 64 Nitro+ was actually slower than AMD's reference Vega 64 card, even though its clock rates are higher. This is not a testing error, nor a specific problem with the model we're testing. Other partner cards have been observed to suffer the same issue, with performance varying based on driver version.

Since this did not occur in all games, or to the same extent, we are guessing that optimizations made in AMD's driver since Vega 64's launch do not apply to third-party cards, since they're not explicitly treated as RX Vega in the driver. It's almost like there's a Radeon fall-back mode. Hopefully the company solves this issue sooner than later. Until then, just know our test results are correct, and there are times when Sapphire's Nitro+ board under-performs the AMD model."
 
I'm not surprised. I've actually got worse performance when overclocking sometimes. In this case, my Vega won't OC at all, so I just set a decent fan profile at leave it. But with my GTX 1080, I was able to OC significantly and, while stable, performance went down.
 
cybereality said:
I'm not surprised. I've actually got worse performance when overclocking sometimes. In this case, my Vega won't OC at all, so I just set a decent fan profile at leave it. But with my GTX 1080, I was able to OC significantly and, while stable, performance went down.
Click to expand...


Most likely the card is throttling. Also there are clock domains on these chips for the memory bus and rest of the chip, also cache thrashing could occur if the frequencies don't match up within the domains..... Lots a of things can happen with overclocking.
 
razor1 said:
Most likely the card is throttling. Also there are clock domains on these chips for the memory bus and rest of the chip, also cache thrashing could occur if the frequencies don't match up within the domains..... Lots a of things can happen with overclocking.
Click to expand...

I'm a convert to GPU water... High sustained overclocks, no noise, even with the base closed-loop solution.
 
razor1 said:
Used to do all that, just swap out hardware too fast now lol, getting to old and lazy it for it lol.
Click to expand...

I think I am done swapping out- the only remaining bug up my ass is that the Linux desktop I just built has a Freesync monitor, so I might want to add an AMD card, but...

It's that new Cryorig Taku, which can take a full-size GPU, but you might not want to put one in, especially an AMD space-heater...

Also, I hear AMD Linux support is pretty iffy (I do have a W10 install for it though).
 
IdiotInCharge said:
I think I am done swapping out- the only remaining bug up my ass is that the Linux desktop I just built has a Freesync monitor, so I might want to add an AMD card, but...

It's that new Cryorig Taku, which can take a full-size GPU, but you might not want to put one in, especially an AMD space-heater...

Also, I hear AMD Linux support is pretty iffy (I do have a W10 install for it though).
Click to expand...


AMD's Linux support definitely improved quite a bit but yeah its still behind nV's. Personally would stick with nV for Linux. But it really depends on what you are using it for, gaming or 3d work, stick with nV, Open CL compute work AMD might be a better option.
 
Nvidia does have better performance, but AMD (from what I've read) has the better open-source driver if you don't wish to run closed-source drivers on your machine.
 
Speed wise it doesn't seem to offer much over a Ref 56, but power and thermals is in a whole different class.
Yet OC'ing leaves much to be desired. :(

http://www.bit-tech.net/reviews/tech/graphics/sapphire-radeon-rx-vega-56-nitro-review/1/

We found the overclocking results a little disappointing. With so many power plugs and VRMs available, we were hoping the power-hungry Vega chip would eat it up and soar into massive frequencies, but raw clock speed is evidently a limiting factor too – we couldn't get the core much past 1,650MHz. It's not an efficient overclocker, either, with system power consumption now reaching nearly 500W. For reference, GTX 1080 Tis, which far outclass the performance here, have our system comfortably under 400W. Fan speed and temperatures also go up, but not too drastically – the massive cooler can still handle itself well. To be fair, performance did improve by between seven and nine percent, which was enough to eclipse a reference RX Vega 64 card and, in one instance, a GTX 1080 OC SKU.
Click to expand...
 
So I finally got around to flashing my RX Vega 64 to the LC version. It seems that the LC version has a 1752 frequency cap which my air cooled card with an EK water block doesn't seem to like at all. Anyone have a profile that I can use to try some cool stuff with my card? Not sure how to export the thing as it seems to be a text file. Maybe upload to a Google Drive or MEGA?

It still idles at 22c and caps at 35c at least during the winter time here in North Carolina which is a very mild winter compared to most other areas in the world.
 
I hate to bring this back from the dead. But I finally got my Vega 64 LC up and have been tuning it this week. I have been beating my dick off trying to get this thing stable. 50% power in benchmarks doesn't get me and HUGE performance increase from just 25%.

P7 states
1740/1190 25%
1730/1180 25%
Net the same score in superposition, but use about 50 watts less. LMAO


I game at 1440p, 144hz freesync. I had to turn all my games up to high/ultra so it will stay under 144hz. When I play overwatch, I can keep power at 0% and it still hits 144fps on ultra. This thing is an animal.
 

Attachments

  • 4.JPG
    4.JPG
    118.5 KB · Views: 0
THUMPer said:
I hate to bring this back from the dead. But I finally got my Vega 64 LC up and have been tuning it this week. I have been beating my dick off trying to get this thing stable. 50% power in benchmarks doesn't get me and HUGE performance increase from just 25%.

P7 states
1740/1190 25%
1730/1180 25%
Net the same score in superposition, but use about 50 watts less. LMAO


I game at 1440p, 144hz freesync. I had to turn all my games up to high/ultra so it will stay under 144hz. When I play overwatch, I can keep power at 0% and it still hits 144fps on ultra. This thing is an animal.
Click to expand...
For curiosity's sake, can you monitor what frequencies it is actually at during load?
 
THUMPer said:
I hate to bring this back from the dead. But I finally got my Vega 64 LC up and have been tuning it this week. I have been beating my dick off trying to get this thing stable. 50% power in benchmarks doesn't get me and HUGE performance increase from just 25%.

P7 states
1740/1190 25%
1730/1180 25%
Net the same score in superposition, but use about 50 watts less. LMAO


I game at 1440p, 144hz freesync. I had to turn all my games up to high/ultra so it will stay under 144hz. When I play overwatch, I can keep power at 0% and it still hits 144fps on ultra. This thing is an animal.
Click to expand...

If you want efficiency, then use these...1650 Core, 900mV core, 1100Mhz HBM, 850mV "memory voltage" with PL set at 0...That will give you ~1080 speed under 180W...Personally I run my VEGAs balls to the wall when I game, just because I can and they are insanely fast @ 1750Mhz/1125HBM coupled with FreeSync.

Which LCD do you have? I am looking for a new one..That new Acer 4K 144Hz HDR 4K just announced looks tasty, but it is not FreeSync 2....
 
You must log in or register to reply here.
Back
Top