Overclock.net banner
2441 - 2460 of 13155 Posts

·
Registered
Joined
·
2,534 Posts
For those that really have to have the most out of the card they're going to have to shunt it but its been shown already you're increasing the power draw big time for a small percentage gain in actual performance. I don't know if I will shunt mod it anymore.
 

·
Banned
Joined
·
1,191 Posts
All the basic reference cards are the same as the FE, just the bios is different (and cooler.) All the cards, including the ref are way over engineered. What is limiting perf is heat, but mainly voltage. Under water, there will be little difference between ref and custom cards too. (the custom cards might be better when moded and under LN. Maybe.)
gamers nexus has a bunch videos if you want to nerd out. Here is the one for the FE https://www.youtube.com/watch?v=-zsWoef9MqQ short version == overkill.


RE: RMA to Nvidia. Don't worry about it. I've mauled many a card I bought directly from nvidia and I've never had a problem getting a replacement. That said, they have all been Titan cards (X, X(P), Xp, etc.)
That is why i normally get nvidia cards directly from them. The RMA dance from EVGA and Asus takes to long. Also, you can't resell most big brands of video cards (ie, used cards) on amazon anymore. But you still can for Nvidia. I'll soon have 3 Titan Vs up for sale there.
My suggestion is always keep some backup card around so when you blowup your card, you can still use your PC while waiting for a replacement.

Yeah, nO. To anyone seeing me quoting this guy's post and wondering, get an A variety which is on paper from NVIDIA the one able to be over clocked at this point as the none A varieties have been stated by NVIDIA themselves to be binned crappier cards. Companies that are paying NVIDIA the extra cash up front to get the A variety are then able to over clock and cool the cards with custom better cooling solutions and then able to push the cards as better over clockers etc. in order to get more money.

This launch is such a farce, it's unreal. They're acting like MSM and lying about everything possible piece of information to cause confusion expecting to get away with bold face lying. While at the same time trying to be honest with the retailers BUYING said A variety versions in order to sell and make a profit. This whole mess is so over the top hyped and straight up bold faced lied about, it's unreal.
 

·
4k Enthusiast
Joined
·
3,964 Posts
For those that really have to have the most out of the card they're going to have to shunt it but its been shown already you're increasing the power draw big time for a small percentage gain in actual performance. I don't know if I will shunt mod it anymore.
Everything I'm seeing is that the percentage performance gains are really awful compared to the percentage power increase above 300W. I'm not even sure custom BIOS is worth the risk this time around.
 

·
NOP sled
Joined
·
93 Posts
I can't say with certainty but GPU artifacting has never been the same for me as GDDR artifacting. GPU artifacting usually ends up being popping or flashing textures while GDDR artifacting ends up looking like weird symbols or missing textures for instance.

For instance this is a picture someone put up of their artifacting. I would be guessing GDDR6 artifacting over GPU artifacting on that instance.
Mine started like that, totally "space invaders" for the symbols.

I got an RMA request in, hoping this isn't some Nvidia shenanigans where I wait 5 weeks for an email. LOL

Now I get this. Reminds me of when my Radeon 9700 Pro bit the dust. Although it was very pink. This is pretty tame with the white. Mind you less psychoactives these days.
 

Attachments

·
Registered
Joined
·
1,269 Posts
very cool how Bitspower shipped my WB for 2080ti strix OC edition , now i need to find the card somewhere in the world
Ditching the Palit card and the Phanteks block?
 

·
Registered
Joined
·
1,269 Posts
yeah it was a dead card that is too hot (65c peak if i'm going 4k unlocked frame rate) and doesn't OC at all although it runs games just fine , reference PCB cards are not on my radar at all now
It does sound like there was something wrong. Although it's probably not because it's a reference PCB otherwise the same issue would be popping up everywhere as 90% of the 2080Ti's currently available seem to be reference PCB haha. Do you happen to know if the Palit card has a 300A or non-A GPU?

Hope the Strix gives you a better experience than the Palit card. It looks like it's got one hell of a power delivery on it, it's a shame that it's limited to 325W with the standard BIOS though. That's lower than some reference designs, but there might be a custom BIOS out there.
 

·
Premium Member
Joined
·
6,020 Posts
Received my 2080 Ti FE today. Just powered it up and to my horror, one of the fans is defective. What I can only describe as a constant brushing sound as if the bearing is catching on something.

You can really tell Nvidia are pushing these out as fast as possible. No QC.

Returning it for a refund. Simply unnacceptable for a £1100 card.
 

·
Registered
Joined
·
855 Posts
What are y'alls clocks at idle? Only noticed yesterday I always sit at 1350/7000 with everything at default settings.

Edit: duh, just remembered I had turned "max performance" mode in nVidia control panel.
 

·
Enthusiast
Joined
·
361 Posts
Discussion Starter · #2,458 ·


MSI GeForce RTX 2080 Ti Gaming X Trio - Mini Review

Had my Gaming X Trio for a while now, but got stuck playing Call of Duty so didn't test it until now.

The good news is, card is extremely quiet (also got a 0 RPM mode) thanks to its massive cooler, and it looks absolutely amazing, only card that might come close would be the AORUS.

The bad news, it can barely overclock because of the ridiculous power limit. You are paying for a custom PCB with more power phases (1) and an extra 6-pin power connector, but they aren't helping anything at all.

Default MSI Gaming X Trio BIOS is at a comfortable 300W, with the possibility of 330W increasing the power limit to 110%. (Founders Edition goes up to 320W, EVGA XC up to 338W, Gigabyte Gaming OC up to 366W and Galax OC up to 380W)

To properly stress the card for my purposes (taking Screenshots in 5K to 8K resolution), I raised every setting in Call of Duty: Black Ops 4 to the max (except Texture Quality to avoid hitting the VRAM ceiling) and set the resolution to 5120 x 2880 (5K), joined a custom TDM game and wrote down the performance numbers after 3 minutes (26c Ambient).

#1 Stock, Out of Box (300W)
72c with the fan speed at 41/50% (1380/1350rpm).
0.987-1.000v / 1860-1875 MHz

Card has a boost clock of 1755 MHz and it managed to reach 1875 MHz, but this is far from the 2000-2100 MHz many reviewers took their Founders Edition during overclocking. Need to point out 5K resolution puts considerably more stress on the card, so it can't maintain as high of a core clock, if I put it back to 1440p it would go a lot higher, so not really comparable but still.

#2 Power Limit 110% (330W)
73c with the fan speed at 42/52% (1440/1410rpm).
1.018-1.031v / 1890-1905 MHz

So raising the Power Limit by 30W increased the voltage by 0.031v (3.1%), which in turn increased the core clock by 30 MHz (1.6%).

How does temperature affect it? Since many consider 73c to be high.

#3 Power Limit 110% (330W) & Fan Speed 100%
56c with the fan speed at 100/100% (3380/2620rpm).
1.025v / 1935 MHz

Fans spinning at almost twice the RPM (85%+) lowered the temperature by 17c (30%), both the core clock and voltage became a more solid straight line at 1.025v / 1935 MHz.

So conclusion here is that out of box, installing MSI Afterburner, increasing the power limit and fan speed, the fastest the card will possibly go is 1935 MHz.

Anything above 1935 MHz is then considered an overclock. But before getting into that, I was curious how much voltage the GPU actually needed for this speed, because out of box they are always (to a varying degree) using a much higher voltage than required.

#4 Power Limit 110% (330W) & Fan Speed 100% & Manual Voltage/Frequency Curve Editor set to 0.900v/1935MHz
51c with the fan speed at 100/100% (3380/2620rpm).
0.900v / 1935 MHz

Temperature went down by 5c and it was only using ~85% of the power limit (TDP). I was impressed but not for long, as it was not a usable voltage curve. As expected when you idle in Windows the temperature goes down, and because of it the gpu clock goes up, so starting a game at 0.900v/1935MHz the GPU is then running close to 2000MHz, at only 0.900v it is very unstable and crashes fast, before the gpu has time to heat up and clock down to 1935 MHz.

To solve this early crash (startup) I had to up the voltage from 0.900v to 0.950v, it then managed to remain stable at the initial low temperature/higher clock.

But what if.. 0.900v would have worked at the initial startup, since 51c is almost watercooling temperature, what if the fan speed is lowered to a quiet level?

#5 Power Limit 110% (330W) & Fan Speed 45% & Manual Voltage/Frequency Curve Editor set to 0.900v/1935MHz
66c with the fan speed at 45/45% (1530/1220rpm).
0.900v / 1935 MHz

I have yet to mention the noise level, but 35% fan speed is literally inaudible over (quieter than) my 4x 120mm 850 rpm radiator fans, though at only 35% the card reached closer to 75c, which I thought was a bit much as at around 67c the gpu clocked down from 1935MHz to 1920MHz. So I upped it to 45% which is still almost inaudible, it really is super quiet! And that's while maxing the card in a 2018 game at 5K resolution, only 66c and almost dead silent, I think I'm safe in saying this is one of the quietest cards (but remember it is also the longest at 327mm, does not fit in many cases).

After finding out the card runs 1935W, it requires 0.950v to remain stable in lower resolution/temperature and that it can do it all with 45% fan speed, this is the result.

#6 Power Limit 110% (330W) & Fan Speed 45% & Manual Voltage/Frequency Curve Editor set to 0.950v/1935MHz
69c with the fan speed at 45/45% (1530/1220rpm).
0.950v / 1935 MHz

It went up 3c and from 85% power limit to 92-96% by incrasing the voltage from 0.900v to 0.950v.

Out of Box with Increased Power Limit & Manual 100% Fan Speed:
56c, 100% Fan Speed, 1.025v / 1935 MHz

Optimized with Increased Power Limit & Manual 45% Fan Speed & Voltage/Frequency Curve Editor:
69c, 45% Fan Speed, 0.950v / 1935 MHz (Almost inaudible and lower power usage, same performance)

Finally overclocking, this is where things get depressing.. this is out of box overclocking, max it could do was +90 without running into the same problem as earlier when the temperature was low and it clocker higher and crashed when starting a game.

#7 Power Limit 100% (300W) & Core Clock +90MHz
76c with the fan speed at 45/45%, slightly higher ambient.
0.968-0.981v / 1920-1935 MHz

Raising the power limit by 10%.

#8 Power Limit 110% (330W) & Core Clock +90MHz
77c with the fan speed at 45/45%, slightly higher ambient.
0.987-1.000v / 1950-1965 MHz

So that was it, it managed to go from 1935 MHz to 1965 MHz (+1.5%) with +90 on the core clock. Note that this was with a low fan speed, to keep it quiet, it would have gone higher with fans at 100% but I could never stand that noise.

Out of Box:
1875 MHz

Out of Box with Power Target 110%:
1905 MHz (+30 MHz)

Out of Box with Power Target 110% and Fans at 100%:
1935 MHz (+30 MHz)

Overclocking Core Clock +90 with Default Power Target and Fans on 45%:
1935 MHz

Overclocking Core Clock +90 with Power Target 110% and Fans on 45%:
1965 MHz (+30 MHz)

So from first to last, a 90 MHz increase. Realistically just installing afterburner you can get it to 1935 MHz in seconds. Then it's only a 30 MHz increase.

This means trying to overclock is almost pointless, 30 MHz (1.5%) is likely only a single FPS.

It was interesting though seeing how it scales with temperature, going full fan speed raised the voltage and core clock speed by 30 MHz, so using a waterblock to cool the GPU definitely helps, and you could probably get at least 30 MHz more (1.5%) but this is with a 330W BIOS, many cards stop at just 290W.

(Rough temperatures, just to show the steps I noticed)
Above 50c - 1.042v
Above 60c - 1.031v
Above 70c - 1.024v


Higher voltage and lower temperature means the core clock goes up, how much a waterblock helps I will find out, but not anytime soon. Also I was not able to adjust the voltage up, only down under 1.000v, was able to set it to 1.250v in Afterburner but it never actually went above 1.024v.

Cause of this low overclock is why? What prevented me from going above 1965 MHz? Was it power limit, temperature or voltage?

Here is where both the expected and unexpected revealed itself, from the looks of the results above and the fact that power limit usually were at 100-105%, you would assume temperature is the problem, and lack of voltage control.. but that is not the case at all.

After two other members here confirmed it was safe to flash BIOS on the Gaming X Trio, I decided to give it a go.

#9 GALAX BIOS & Power Limit 126% (380W) & Core Clock +90MHz
78c with the fan speed at 45/45%. (Disregard temperature as you can adjust the voltage down from 1.043v using Voltage/Frequency Curve Editor)
1.043v / 1965 MHz
 
 
Click SHOW for Graphs!


Okay so it worked fine so far, with the increased power limit the voltage instantly went up to a rock solid 1.043v never deviating, same with the core clock, a completely straight line of 1965 MHz.

So what happened when I increased the +90 core clock further?.. GPU shot up to 2040 MHz! Another 75 MHz by typing in a random number which I have now forgotten, with a hot (75c) silent card (1500rpm), now that is impressive! It did crash soon after, which is most likely because of the high temperature (unstable), water cooled it would surely go even further than 2050 MHz, that I am sure of.

Final conclusion is that power limit is the main problem, even on a quiet high temperature air cooled card. Changing BIOS is not only for water cooled cards is what I am saying, there's several % of performance to gain.

I feel for anyone who is stuck with 290W power limit BIOS or a water cooled Founders Edition that can not flash BIOS (yet), it really restricts overclocking of any kind.

Of course realize there is a lot more testing to be done on the 380W BIOS but I feel I should wait until I have a water cooled card, to compare it on air versus water preferably the same day, so temperature, game patches, drivers are all the same.

Also there is no way I am keeping the MSI Gaming X Trio (It is a visually stunning & quiet card but other than that it has nothing going for it at all), the best cards are as usual the cheapest reference PCB ones, because almost every card on air will hit the power limit without replacing the BIOS, which has to be done if you're pushing the card on water so it does not really matter which card you buy, as long as it is reference PCB (so most waterblocks fit, unless you want a specific block). There are 3 notable cards that has exceptionally high power limit out of box, and those are Gigabyte Gaming OC at 366W but does not allow for a reference waterblock even though it is a reference PCB card, they added a FAN/RGB connector in a stupid location, then the EVGA FTW3 Ultra which is a huge card with custom PCB at 373W and last, surprisingly, the Galax OC reference PCB card with 380W BIOS (I can not confirm it ships with that BIOS yet but according to Tom's Hardware their sample had it).

TL;DR Nvidia has the RTX 2080 Ti in a straight jacket, 320W on the Founders Edition is a travesty.
 

·
Gaming Overclocker
Joined
·
2,063 Posts


MSI GeForce RTX 2080 Ti Gaming X Trio - Mini Review

Had my Gaming X Trio for a while now, but got stuck playing Call of Duty so didn't test it until now.

The good news is, card is extremely quiet (also got a 0 RPM mode) thanks to its massive cooler, and it looks absolutely amazing, only card that might come close would be the AORUS.

The bad news, it can barely overclock because of the ridiculous power limit. You are paying for a custom PCB with more power phases (1) and an extra 6-pin power connector, but they aren't helping anything at all.

Default MSI Gaming X Trio BIOS is at a comfortable 300W, with the possibility of 330W increasing the power limit to 110%. (Founders Edition goes up to 320W, EVGA XC up to 338W, Gigabyte Gaming OC up to 366W and Galax OC up to 380W)

To properly stress the card for my purposes (taking Screenshots in 5K to 8K resolution), I raised every setting in Call of Duty: Black Ops 4 to the max (except Texture Quality to avoid hitting the VRAM ceiling) and set the resolution to 5120 x 2880 (5K), joined a custom TDM game and wrote down the performance numbers after 3 minutes (26c Ambient).

#1 Stock, Out of Box (300W)
72c with the fan speed at 41/50% (1380/1350rpm).
0.987-1.000v / 1860-1875 MHz

Card has a boost clock of 1755 MHz and it managed to reach 1875 MHz, but this is far from the 2000-2100 MHz many reviewers took their Founders Edition during overclocking. Need to point out 5K resolution puts considerably more stress on the card, so it can't maintain as high of a core clock, if I put it back to 1440p it would go a lot higher, so not really comparable but still.

#2 Power Limit 110% (330W)
73c with the fan speed at 42/52% (1440/1410rpm).
1.018-1.031v / 1890-1905 MHz

So raising the Power Limit by 30W increased the voltage by 0.031v (3.1%), which in turn increased the core clock by 30 MHz (1.6%).

How does temperature affect it? Since many consider 73c to be high.

#3 Power Limit 110% (330W) & Fan Speed 100%
56c with the fan speed at 100/100% (3380/2620rpm).
1.025v / 1935 MHz

Fans spinning at almost twice the RPM (85%+) lowered the temperature by 17c (30%), both the core clock and voltage became a more solid straight line at 1.025v / 1935 MHz.

So conclusion here is that out of box, installing MSI Afterburner, increasing the power limit and fan speed, the fastest the card will possibly go is 1935 MHz.

Anything above 1935 MHz is then considered an overclock. But before getting into that, I was curious how much voltage the GPU actually needed for this speed, because out of box they are always (to a varying degree) using a much higher voltage than required.

#4 Power Limit 110% (330W) & Fan Speed 100% & Manual Voltage/Frequency Curve Editor set to 0.900v/1935MHz
51c with the fan speed at 100/100% (3380/2620rpm).
0.900v / 1935 MHz

Temperature went down by 5c and it was only using ~85% of the power limit (TDP). I was impressed but not for long, as it was not a usable voltage curve. As expected when you idle in Windows the temperature goes down, and because of it the gpu clock goes up, so starting a game at 0.900v/1935MHz the GPU is then running close to 2000MHz, at only 0.900v it is very unstable and crashes fast, before the gpu has time to heat up and clock down to 1935 MHz.

To solve this early crash (startup) I had to up the voltage from 0.900v to 0.950v, it then managed to remain stable at the initial low temperature/higher clock.

But what if.. 0.900v would have worked at the initial startup, since 51c is almost watercooling temperature, what if the fan speed is lowered to a quiet level?

#5 Power Limit 110% (330W) & Fan Speed 45% & Manual Voltage/Frequency Curve Editor set to 0.900v/1935MHz
66c with the fan speed at 45/45% (1530/1220rpm).
0.900v / 1935 MHz

I have yet to mention the noise level, but 35% fan speed is literally inaudible over (quieter than) my 4x 120mm 850 rpm radiator fans, though at only 35% the card reached closer to 75c, which I thought was a bit much as at around 67c the gpu clocked down from 1935MHz to 1920MHz. So I upped it to 45% which is still almost inaudible, it really is super quiet! And that's while maxing the card in a 2018 game at 5K resolution, only 66c and almost dead silent, I think I'm safe in saying this is one of the quietest cards (but remember it is also the longest at 327mm, does not fit in many cases).

After finding out the card runs 1935W, it requires 0.950v to remain stable in lower resolution/temperature and that it can do it all with 45% fan speed, this is the result.

#6 Power Limit 110% (330W) & Fan Speed 45% & Manual Voltage/Frequency Curve Editor set to 0.950v/1935MHz
69c with the fan speed at 45/45% (1530/1220rpm).
0.950v / 1935 MHz

It went up 3c and from 85% power limit to 92-96% by incrasing the voltage from 0.900v to 0.950v.

Out of Box with Increased Power Limit & Manual 100% Fan Speed:
56c, 100% Fan Speed, 1.025v / 1935 MHz

Optimized with Increased Power Limit & Manual 45% Fan Speed & Voltage/Frequency Curve Editor:
69c, 45% Fan Speed, 0.950v / 1935 MHz (Almost inaudible and lower power usage, same performance)

Finally overclocking, this is where things get depressing.. this is out of box overclocking, max it could do was +90 without running into the same problem as earlier when the temperature was low and it clocker higher and crashed when starting a game.

#7 Power Limit 100% (300W) & Core Clock +90MHz
76c with the fan speed at 45/45%, slightly higher ambient.
0.968-0.981v / 1920-1935 MHz

Raising the power limit by 10%.

#8 Power Limit 110% (330W) & Core Clock +90MHz
77c with the fan speed at 45/45%, slightly higher ambient.
0.987-1.000v / 1950-1965 MHz

So that was it, it managed to go from 1935 MHz to 1965 MHz (+1.5%) with +90 on the core clock. Note that this was with a low fan speed, to keep it quiet, it would have gone higher with fans at 100% but I could never stand that noise.

Out of Box:
1875 MHz

Out of Box with Power Target 110%:
1905 MHz (+30 MHz)

Out of Box with Power Target 110% and Fans at 100%:
1935 MHz (+30 MHz)

Overclocking Core Clock +90 with Default Power Target and Fans on 45%:
1935 MHz

Overclocking Core Clock +90 with Power Target 110% and Fans on 45%:
1965 MHz (+30 MHz)

So from first to last, a 90 MHz increase. Realistically just installing afterburner you can get it to 1935 MHz in seconds. Then it's only a 30 MHz increase.

This means trying to overclock is almost pointless, 30 MHz (1.5%) is likely only a single FPS.

It was interesting though seeing how it scales with temperature, going full fan speed raised the voltage and core clock speed by 30 MHz, so using a waterblock to cool the GPU definitely helps, and you could probably get at least 30 MHz more (1.5%) but this is with a 330W BIOS, many cards stop at just 290W.

(Rough temperatures, just to show the steps I noticed)
Above 50c - 1.042v
Above 60c - 1.031v
Above 70c - 1.024v


Higher voltage and lower temperature means the core clock goes up, how much a waterblock helps I will find out, but not anytime soon. Also I was not able to adjust the voltage up, only down under 1.000v, was able to set it to 1.250v in Afterburner but it never actually went above 1.024v.

Cause of this low overclock is why? What prevented me from going above 1965 MHz? Was it power limit, temperature or voltage?

Here is where both the expected and unexpected revealed itself, from the looks of the results above and the fact that power limit usually were at 100-105%, you would assume temperature is the problem, and lack of voltage control.. but that is not the case at all.

After two other members here confirmed it was safe to flash BIOS on the Gaming X Trio, I decided to give it a go.

#9 GALAX BIOS & Power Limit 126% (380W) & Core Clock +90MHz
78c with the fan speed at 45/45%. (Disregard temperature as you can adjust the voltage down from 1.043v using Voltage/Frequency Curve Editor)
1.043v / 1965 MHz
 
 
Click SHOW for Graphs!


Okay so it worked fine so far, with the increased power limit the voltage instantly went up to a rock solid 1.043v never deviating, same with the core clock, a completely straight line of 1965 MHz.

So what happened when I increased the +90 core clock further?.. GPU shot up to 2040 MHz! Another 75 MHz by typing in a random number which I have now forgotten, with a hot (75c) silent card (1500rpm), now that is impressive! It did crash soon after, which is most likely because of the high temperature (unstable), water cooled it would surely go even further than 2050 MHz, that I am sure of.

Final conclusion is that power limit is the main problem, even on a quiet high temperature air cooled card. Changing BIOS is not only for water cooled cards is what I am saying, there's several % of performance to gain.

I feel for anyone who is stuck with 290W power limit BIOS or a water cooled Founders Edition that can not flash BIOS (yet), it really restricts overclocking of any kind.

Of course realize there is a lot more testing to be done on the 380W BIOS but I feel I should wait until I have a water cooled card, to compare it on air versus water preferably the same day, so temperature, game patches, drivers are all the same.

Also there is no way I am keeping the MSI Gaming X Trio (It is a visually stunning & quiet card but other than that it has nothing going for it at all), the best cards are as usual the cheapest reference PCB ones, because almost every card on air will hit the power limit without replacing the BIOS, which has to be done if you're pushing the card on water so it does not really matter which card you buy, as long as it is reference PCB (so most waterblocks fit, unless you want a specific block). There are 3 notable cards that has exceptionally high power limit out of box, and those are Gigabyte Gaming OC at 366W but does not allow for a reference waterblock even though it is a reference PCB card, they added a FAN/RGB connector in a stupid location, then the EVGA FTW3 Ultra which is a huge card with custom PCB at 373W and last, surprisingly, the Galax OC reference PCB card with 380W BIOS (I can not confirm it ships with that BIOS yet but according to Tom's Hardware their sample had it).

TL;DR Nvidia has the RTX 2080 Ti in a straight jacket, 320W on the Founders Edition is a travesty.
Excellent review mate, congrats! :specool:
Even if I was looking for some more sustained OC freq (with the 380W bios at least).
 
2441 - 2460 of 13155 Posts
Top