PCPer on AMD Radeon Vega Frontier Edition Liquid Cooled

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
56,162
PCPer's Rowdy Ryan Shrout is back at it again this week with AMD's Vega Frontier Edition video card. This time he has put the liquid cooled version to the test. Do things end up any better this time than they did last time with the air cooled edition?

Which leads to the real question: what does this mean for the upcoming RX Vega product? While I know that testing the Frontier Edition in ONLY gaming is a bit of a faux pas, much of our interest was in using this product to predict what AMD is going bring to gamers later this month. It is apparent now that if the clocks are in the 1600 MHz range consistently, rather than dipping into 1400 MHz and below states as we found with the air cooler at stock, Vega in its current state can be competitive with the GeForce GTX 1080. That’s an upgrade over where it stood before – much closer to GTX 1070 performance.
 
So, a liquid-version-1500$ card, in order to reach the performance of a 1-year-older-600$ card (*GTX 1080 reference ) !!!
Waiting for RX Vega now (*the gaming edition of course :p!!! )
 
who is the target audience for this card with such a significant price increase and only a relatively small boost in performance?
 
PCPer said:
I did notice that AMD limits the maximum thermal temperature to 70C in Wattman, compared to 90C with the air-cooled card. When I asked AMD for the reason behind this change, they only stated that it was “because they can” due to the vastly improved cooling capability.
Click to expand...

What is the boiling point of the fluid used in the cooler? If the fluid were to vapourise the unit could explode. If it's distilled water then you don't want the temperature getting anywhere near 100, and there'll be considerable vapourisation at 90.
 
os2wiz said:
Frankly I am tired of hearing all the irrational roid rage against RX Vega , when no one has tested RX Vega. The only report on benchmarks of RX Vega is a week old report on an engineering sample that had Firestrike scores 10% faster than a 1080. WE do NOT know the age of the drivers it was using but I would guess that release performance will at least be 10% faster, more or less within 10% of a 1080 Ti. Anyone who knows about AMD driver development over the past 3-4 years know that they continue to improve driver performance AFTER release. So there is still a reasonable possibility that within 2 months the RX Vega flagship model will equal a 1080 Ti in performance. People can sneer all they want but logic and fact are on my side. The fools who project RX Vega performance based on Frontier Edition are just that. AMD said quite a while ago the Frontier Edition was not designed for gaming, but AI and game development. That is why those who choose to game on F.E. are having very erratic experience. The drivers are designed to be backwards compatible so many of the new features on the Vega chip are totally disabled. The driver level is Fiji level like the old R9 Fury I have in my current Ryzen system. When all is said and done RX Vega will be a lot closer to a 1080 Ti than to a 1080. Also since it has built in DX 12.1 support. It will be more capable in some new games yet to be released that support that standard. I also suspect the infinity fabric in the Vega gpu will perform slightly better on Ryzen systems than on Intel. So we will see, but I am sure I won't be disappointed.
Click to expand...

Sure thing budy.... Wait for RX VEGA (*Gaming Edition)!!!.
If you think you will see some tremendous difference you are way mistaken (*also, please keep in mind that the Gaming RX VEGA will have only 8 GB HBM2 memory, instead of 16).
It's like saying, that i've seen the review of the TITAN Xp, and thinking that its pure gaming version , the GTX 1080Ti, will be totally different!!!
 
Vega is a diaster. It's going to be expensive cause they had to use HBM2. It is going to run hot and performance is going to be around over a year old 1080. Fury 2.0.
 
Hmmm. I'm really confused about this FE Vega. I get that it's not really a gaming card but then it's not really a professional card without the certified drivers but it does seem to deliver pretty good for pro purposes. But it's so way off for gaming for the price that it does seem hard to envision RX Vega being as good or better for gaming at $500 or below and it's going to have to be probably below $500 to be competitive. We'll see soon enough.
 
os2wiz said:
Frankly I am tired of hearing all the irrational roid rage against RX Vega , when no one has tested RX Vega. The only report on benchmarks of RX Vega is a week old report on an engineering sample that had Firestrike scores 10% faster than a 1080. WE do NOT know the age of the drivers it was using but I would guess that release performance will at least be 10% faster, more or less within 10% of a 1080 Ti. Anyone who knows about AMD driver development over the past 3-4 years know that they continue to improve driver performance AFTER release. So there is still a reasonable possibility that within 2 months the RX Vega flagship model will equal a 1080 Ti in performance. People can sneer all they want but logic and fact are on my side. Those who project RX Vega performance based on Frontier Edition are misguided. AMD said quite a while ago the Frontier Edition was not designed for gaming, but AI and game development. That is why those who choose to game on F.E. are having very erratic experience. The drivers are designed to be backwards compatible so many of the new features on the Vega chip are totally disabled. The driver level is Fiji level like the old R9 Fury I have in my current Ryzen system. When all is said and done RX Vega will be a lot closer to a 1080 Ti than to a 1080. Also since it has built in DX 12.1 support. It will be more capable in some new games yet to be released that support that standard. I also suspect the infinity fabric in the Vega gpu will perform slightly better on Ryzen systems than on Intel. So we will see, but I am sure I won't be disappointed.
Click to expand...

To each their own, but I disagree. It is nothing more than wishful thinking to think the Vega will see 1080 Ti level of performance at this point. Either way it is irrelevant. By the time these "magical" drivers that will make the card 30% faster (not happening) come along it will be competing with Volta and not the 1080 Ti. This card is too little and waaay too late.
 
Sold all my AMD stock today... Wasn't much, just 100 shares and I doubled my money but I just kinda think it's peaked. Ryzen is a solid CPU and had me encouraged but I ain't feeling too good about Vega. I know all we have is a lot if speculation based on FE but if the last 7 years or so have taught me anything is that when it comes to AMD, the pre-release speculation (not marketing hype) is usually pretty close to accurate.
 
os2wiz said:
Frankly I am tired of hearing all the irrational roid rage against RX Vega , when no one has tested RX Vega. The only report on benchmarks of RX Vega is a week old report on an engineering sample that had Firestrike scores 10% faster than a 1080. WE do NOT know the age of the drivers it was using but I would guess that release performance will at least be 10% faster, more or less within 10% of a 1080 Ti. Anyone who knows about AMD driver development over the past 3-4 years know that they continue to improve driver performance AFTER release. So there is still a reasonable possibility that within 2 months the RX Vega flagship model will equal a 1080 Ti in performance. People can sneer all they want but logic and fact are on my side. Those who project RX Vega performance based on Frontier Edition are misguided. AMD said quite a while ago the Frontier Edition was not designed for gaming, but AI and game development. That is why those who choose to game on F.E. are having very erratic experience. The drivers are designed to be backwards compatible so many of the new features on the Vega chip are totally disabled. The driver level is Fiji level like the old R9 Fury I have in my current Ryzen system. When all is said and done RX Vega will be a lot closer to a 1080 Ti than to a 1080. Also since it has built in DX 12.1 support. It will be more capable in some new games yet to be released that support that standard. I also suspect the infinity fabric in the Vega gpu will perform slightly better on Ryzen systems than on Intel. So we will see, but I am sure I won't be disappointed.
Click to expand...

over the top blind hope here ^

-- this is the first person I've read posts from here who was over the top promoting AMD being able to do something they simply can't. For all the hate the AMD supporters get -- this if the first example I've seen that plays to the overhype stereo type.

Most of us expected Vega gaming card to be on par with a 1080. Not at 1080TI (35% faster).
I'm hoping for Vega to do great things too, I owned a couple Fury X cards and really liked them in Crossfire!!! - but your scenario simply isn't happening.

Two months for equitable 1080TI speed based on driver updates? um. no.

The fine wine driver updates do improve performance, but in the HardOCP review of Nvidia vs AMD driver support and performance increases over several years - it's only slightly better than Nvidia's improvements.

I may still buy a Vega. I just don't have any expectation it'll be 1080 TI levels.
 
MacLeod said:
Sold all my AMD stock today... Wasn't much, just 100 shares and I doubled my money but I just kinda think it's peaked. Ryzen is a solid CPU and had me encouraged but I ain't feeling too good about Vega. I know all we have is a lot if speculation based on FE but if the last 7 years or so have taught me anything is that when it comes to AMD, the pre-release speculation (not marketing hype) is usually pretty close to accurate.
Click to expand...

Eh only 100 shares nothing to lose, but I would have held it till threadripper release and epyc played out a bit longer. Vega may not be a game changer anymore but threadripper and epyc are and they're probably more important to their core business now with RTG spun off. I suppose there's still time for Vega RX to shape up... not very confident but lol...
 
thesmokingman said:
Eh only 100 shares nothing to lose, but I would have held it till threadripper release and epyc played out a bit longer. Vega may not be a game changer anymore but threadripper and epyc are and they're probably more important to their core business now with RTG spun off. I suppose there's still time for Vega RX to shape up... not very confident but lol...
Click to expand...

I watched it drop to $11 a while back and hover for a while and got a little nervous. With what Vega is looking like it's gonna be I got a feeling it would drop a little more. I am by no means an investing whiz... Not by any stretch of the imagination but with the bleak Vega outlook and AMD not paying dividends (I'm a big fan of dividend stocks) I decided to move on.....which pretty much guarantees it doubles in the next few days haha!
 
Quartz-1 said:
What is the boiling point of the fluid used in the cooler? If the fluid were to vapourise the unit could explode. If it's distilled water then you don't want the temperature getting anywhere near 100, and there'll be considerable vapourisation at 90.
Click to expand...

You can't vaporize a liquid like that. Flash vaporization only occurs on large boilers where the operator let the water level get too low. This is a constant volume under pressure, and while the pressure may raise, flash vaporization can't happen under those conditions.

And just about all AIO water cooling solutions add some type of glycol which not only inhibits rust, reduces mold/bacteria growth, but raises the boiling point.
 
Archaea said:
over the top blind hope here ^

-- this is the first person I've read posts from here who was over the top promoting AMD being able to do something they simply can't. For all the hate the AMD supporters get -- this if the first example I've seen that plays to the overhype stereo type.

Most of us expected Vega gaming card to be on par with a 1080. Not at 1080TI (35% faster).
I'm hoping for Vega to do great things too, I owned a couple Fury X cards and really liked them in Crossfire!!! - but your scenario simply isn't happening.

Two months for equitable 1080TI speed based on driver updates? um. no.

The fine wine driver updates do improve performance, but in the HardOCP review of Nvidia vs AMD driver support and performance increases over several years - it's only slightly better than Nvidia's improvements.

I may still buy a Vega. I just don't have any expectation it'll be 1080 TI levels.
Click to expand...

Agree, overhyping like that hurts AMD image, but at least in this forum, most of us have reasonable expectation where performance will land.
 
MacLeod said:
I watched it drop to $11 a while back and hover for a while and got a little nervous. With what Vega is looking like it's gonna be I got a feeling it would drop a little more. I am by no means an investing whiz... Not by any stretch of the imagination but with the bleak Vega outlook and AMD not paying dividends (I'm a big fan of dividend stocks) I decided to move on.....which pretty much guarantees it doubles in the next few days haha!
Click to expand...

Hey as long you making money, you win!
 
Is it just me or do all of those graphs look disappointing?

Even if drivers could make up the difference in performance between Vega and the 1080 Ti (most likely not), you're still looking at a card that consumes way more power and requires liquid cooling.
 
From what I know , the Vega RX is based on the same GPU as the founders edition cards, so, even with optimized drivers I can't see it giving any more than 10% extra performance, which might take it to within a whisker of GTX1080 performance. here's hoping !!
If that turns out to be the case, as long as the price is right, I would buy one, and thats what AMD needs to do to make it competitive. They did it with the Ryzen and it has made Intel jump to it to be competitive !!
 
Sith'ari said:
Sure thing budy.... Wait for RX VEGA (*Gaming Edition)!!!.
If you think you will see some tremendous difference you are way mistaken (*also, please keep in mind that the Gaming RX VEGA will have only 8 GB HBM2 memory, instead of 16).
It's like saying, that i've seen the review of the TITAN Xp, and thinking that its pure gaming version , the GTX 1080Ti, will be totally different!!!
Click to expand...

You will be sadly mistaken and HBM2 has much greater bandwith than GDDR5X so 8GB is more than adequate even for 4K like I have at home now on an R9 Fury.with 4K HBM1. Different drivers most advanced features disabled, Fiji level drivers on Frontier Edition = ignorance on your limited logic processors in your brain.
 
os2wiz said:
You will be sadly mistaken and HBM2 has much greater bandwith than GDDR5X so 8GB is more than adequate even for 4K like I have at home now on an R9 Fury.with 4K HBM1. Different drivers most advanced features disabled, Fiji level drivers on Frontier Edition = ignorance on your limited logic processors in your brain.
Click to expand...

GREAT !!! Everything is OK then !!! Keep waiting for the Revolution!!!
(*after all NAVI is coming in few years,.....or a decade;))
 
DigitalGriffin said:
You can't vaporize a liquid like that. Flash vaporization only occurs on large boilers where the operator let the water level get too low. This is a constant volume under pressure, and while the pressure may raise, flash vaporization can't happen under those conditions.

And just about all AIO water cooling solutions add some type of glycol which not only inhibits rust, reduces mold/bacteria growth, but raises the boiling point.
Click to expand...

Yup. While a chip can get hot enough to heat its own volume to 100c, there is not enough energy involved to heat the die, IHS, coldplate and the working fluid to that kind of temp. Coolant temps in an AIO *might* hit 50c. Even 60c would require far greater amounts of power...or the builder forgetting to connect the rad fans. Even then...I doubt it.

Pump failure can be sufficient to cause local boiling, and usually manifests as a bunch of gunk oozing out of the pump/coldplate as the additives crystalize after coming out of solution. But you have to be not paying attention for a long time to get to that point.
 
os2wiz said:
You will be sadly mistaken and HBM2 has much greater bandwith than GDDR5X so 8GB is more than adequate even for 4K like I have at home now on an R9 Fury.with 4K HBM1. Different drivers most advanced features disabled, Fiji level drivers on Frontier Edition = ignorance on your limited logic processors in your brain.
Click to expand...
Vega FE has 483 GB/s of memory bandwidth. 1080 Ti has 484 GB/s on GDDR5X. Where I come from 484 > 483. The Titan Xp has 547 GB/s. GDDR6 will be up to 14 GT/s. That equates to 672 GB/s on a 384-bit bus like used on the 1080 Ti.

Should we wait for HBM3? :ROFLMAO:
 
Armenius said:
Vega FE has 483 GB/s of memory bandwidth. 1080 Ti has 484 GB/s on GDDR5X. Where I come from 484 > 483.
Click to expand...

And that's on stock clocks. Pretty sure every 1080 Ti will take a +300 offset. Most will take +400. That pushes bandwidth up to ~500GB/s.

That's assuming the offset applies to the quad-pumped rate, and not the actual clock rate. If it's clock rate, then the bandwidth would be 530-550 GB/s which seems a little high.
 
Armenius said:
Vega FE has 483 GB/s of memory bandwidth. 1080 Ti has 484 GB/s on GDDR5X. Where I come from 484 > 483. The Titan Xp has 547 GB/s. GDDR6 will be up to 14 GT/s. That equates to 672 GB/s on a 384-bit bus like used on the 1080 Ti.

Should we wait for HBM3? :ROFLMAO:
Click to expand...
Bro please. HBM2>GDDR5X ok? :)
 
Armenius said:
Vega FE has 483 GB/s of memory bandwidth. 1080 Ti has 484 GB/s on GDDR5X. Where I come from 484 > 483. The Titan Xp has 547 GB/s. GDDR6 will be up to 14 GT/s. That equates to 672 GB/s on a 384-bit bus like used on the 1080 Ti.

Should we wait for HBM3? :ROFLMAO:
Click to expand...

You forgot that Fury's HBM1 has more bandwidth than Vega's HBM2.

#waitforvega navi
 
Kardonxt said:
Sounds like a repeat of Fury. Can't say I'm shocked.
Click to expand...
AMD needs to stop building gigantic GPUs, it’s never panned our well for them. Nvidia learned that lesson with the GTX 480.
 
os2wiz said:
You will be sadly mistaken and HBM2 has much greater bandwith than GDDR5X so 8GB is more than adequate even for 4K like I have at home now on an R9 Fury.with 4K HBM1. Different drivers most advanced features disabled, Fiji level drivers on Frontier Edition = ignorance on your limited logic processors in your brain.
Click to expand...

Are you really going to call other people names while being wrong on every fact-based statement you try to make?

Much greater bandwith huh? Bandwith what?

You don't even bother to google the fanboy crap you eat up so readily.

Did you do this crap last round too? Was the Fury an engineering sample on old drivers which was going to come out on top of the 980 by 30% as soon as the magic drivers came and unlocked the real performance, too?

I mean an Fury runs games fine, and I'm sure the Vega will run games fine too. But coming up with this Jack and the Beanstalk fairy tale crap to convince yourself it's the fastest card and that it just needs better drivers to make it the fastest thing EVARRR is like.. the definition of fanboyism. While telling other people their "logical processors in their brains are failing"? Seriously?
 
Last edited:
I had a silly thought of the gaming version of the card doing to the gpu what threadripper did to ryzen.


If they used that fabric to link 4 gpu's in it and had each gpu render 1/4th of the screen for 200fps in 4k, and sold it for $1899, I would buy it :eek:
 
Last edited:
Azphira said:
I had a silly thought of the gaming version of the card doing to the gpu what threadripper did to ryzen.


If they used that fabric to link 4 gpu's in it and had each gpu render 1/4th of the screen for 200fps in 4k, and sold it for $1899, I would buy it :eek:
Click to expand...

That's a premise behind nvidias white paper and Navi. It is more than possible theoretically, but there's some scheduling and memory issues to deal with and heat density issues as well
 
Drakensoul said:
Are you really going to call other people names while being wrong on every fact-based statement you try to make?

Much greater bandwith huh? Bandwith what?

You don't even bother to google the fanboy crap you eat up so readily.

Did you do this crap last round too? Was the Fury an engineering sample on old drivers which was going to come out on top of the 980 by 30% as soon as the magic drivers came and unlocked the real performance, too?

I mean an Fury runs games fine, and I'm sure the Vega will run games fine too. But coming up with this Jack and the Beanstalk fairy tale crap to convince yourself it's the fastest card and that it just needs better drivers to make it the fastest thing EVARRR is like.. the definition of fanboyism. While telling other people their "logical processors in their brains are failing"? Seriously?
Click to expand...

Wait a minute I just got my Fury card last year from Newegg. I had been on a R9 290. I never was in the debate between Fury and or Fury X and 980 Ti. So you've got the wrong man. I never said RX Vega is the fastest card, your second fairy tale. The fact is you guys ahve prejudged RX Vega with no firm basiis to do so. Using Frontier Edition to make your point is totally unscioentific. Only release drivers and game performance can really give an idea what RX vega will do. Then add 10% on top of that for AMD has since Fury (Fiji) consistently improved drivers after release date. All I am saying is that in 60 days from now RX Vega will be in the competitive range with a 1080 Ti perhaps 5 to 10% slower at most.
 
The HBM2 on Vega cards is also over volted and overclocked. 1.35V vs 1.20V etc. Even at stock and 1.2V GDDR5X/GDDR6 is as efficient as HBM2 (Source Hynix).

So now HBM2 is also a problem on the power consumption side.

The only real benefit of HBM2 is ECC and smaller footprint. Everything else it loses in or at best ties in. Nvidia didn't keep HBM2 to P100 and V100 without reason. And dont start on the HBCC PR BS, Polaris got that one too with GDDR5.
 
os2wiz said:
Frankly I am tired of hearing all the irrational roid rage against RX Vega , when no one has tested RX Vega. The only report on benchmarks of RX Vega is a week old report on an engineering sample that had Firestrike scores 10% faster than a 1080. WE do NOT know the age of the drivers it was using but I would guess that release performance will at least be 10% faster, more or less within 10% of a 1080 Ti. Anyone who knows about AMD driver development over the past 3-4 years know that they continue to improve driver performance AFTER release. So there is still a reasonable possibility that within 2 months the RX Vega flagship model will equal a 1080 Ti in performance. People can sneer all they want but logic and fact are on my side. Those who project RX Vega performance based on Frontier Edition are misguided. AMD said quite a while ago the Frontier Edition was not designed for gaming, but AI and game development. That is why those who choose to game on F.E. are having very erratic experience. The drivers are designed to be backwards compatible so many of the new features on the Vega chip are totally disabled. The driver level is Fiji level like the old R9 Fury I have in my current Ryzen system. When all is said and done RX Vega will be a lot closer to a 1080 Ti than to a 1080. Also since it has built in DX 12.1 support. It will be more capable in some new games yet to be released that support that standard. I also suspect the infinity fabric in the Vega gpu will perform slightly better on Ryzen systems than on Intel. So we will see, but I am sure I won't be disappointed.
Click to expand...

So you're telling me in a few weeks we're going to see magical drivers that makes this hardware 30-40% faster? Hot damn! I also have a bridge to sell you if interested.
 
In the tests it seems to just hit an imposed performance ceiling, no matter how much its clocked. Doesnt look good in its current form, but again we are two weeks out of RX Vega, looking forward to those reviews!
 
right now AMD is having road tour to show case Vega. as i was expected AMD did a blind test between RX Vega vs nvidia card (something similar to what they did with PCWorld). though as usual there is no FPS counter. Nvidia setup is using Gsync while AMD setup is using Freesync. rather than performance AMD is touting the whole cost building AMD system is cheaper including Freesync monitor. but it seems AMD indeed admit the Nvidia system is using GTX1080 (non Ti).
 
You must log in or register to reply here.
Back
Top