AMD Radeon Pro Vega Frontier Edition Unboxed, Benchmarked

[P4-630]

Don't kid yourself, I had a Corsair 850 psu and a pair of 7970's and it ran the benches.. well it completed them but the numbers were low. The psu was old and after installing a new psu the numbers went up

Guys, GPUs don't 'brown out' and get "slower" because of an inadequate PSU...

So what is it now!??
LOL!!

Should I buy a PSU with moar watts to get a better 3D Mark score??
Or no....

 

[PerfectWave]

At least we know that the air gpu is clocked at 1600 mhz. maybe it is the cpu that bottleneck the gpu.

 

[Nokiron]

At least we know that the gpu is clocked at 1600 mhz. maybe it is the cpu that bottleneck the gpu.

Not in 3Dmark, and not a 4790K.

What we do know is that the card probably throttles immensly because of thermals (Powerlimit as well?).

 

[NGreediaOrAMSlow]

Are you really serious? If that card pulls the maximum supposed "300W", he still has 250(!!)W for the CPU and rest of the components.

A system with a 4790K would pull 100W at most.

And I guess you forgot the power supply rating. Which is a two digit number representing the guarantee load. If branded PSU, then should be at least 80. Which means guarantee up to 80% of it's full capacity.

Non branded (generic) PSU are lower than that.

550*.80=440

He has a guarantee load up to 440 Watts. While the power may reach 550, pushing it may have side effects.

Also a K processor. Do you think overclocking it will still consume the same?

 

[OSdevr]

He's right guys

If the PSU wasn't enough, he/we'd know because the PC would have shut down/potentially damaged itself. It didn't shut down, ergo it sufficed.

(this isn't to say i agree with everything anyone's ever said about today's market and having a 500W PSU [don't], am just commenting on this specific scenario)

I've had personal experience tell me otherwise:

I had a i5-3570K (77w TDP, not overclocked) machine with a 600w Thermaltake PSU that worked well until I added a GTX 660 TI (150w). Everything was fine if a game wasn't graphically demanding but as soon as I played one that was the video would look really... odd and my sound would become scratchy and quickly cut out. I tried reinstalling drivers and such but no luck until I tried a larger power supply. To my great surprise that fixed everything.

EDIT: My system froze shortly after the 'symptoms' started and needed to be rebooted.

 

[Nokiron]

And I guess you forgot the power supply rating. Which is a two digit number representing the guarantee load. If branded PSU, then should be at least 80. Which means guarantee up to 80% of it's full capacity (550W).

Non branded (generic) PSU are lower than that.

No I didnt. Because you are wrong.

The efficiency is calculated from the wall, not internally. A 550W-rated powersupply with 80% efficiency pulls 660W from the wall but still delivers 550W to the components.

I've had personal experience tell me otherwise:

I had a i5-3570K (77w TDP, not overclocked) machine with a 600w Thermaltake PSU that worked well until I added a GTX 660 TI (150w). Everything was fine if a game wasn't graphically demanding but as soon as I played one that was the video would look really... odd and my sound would become scratchy and quickly cut out. I tried reinstalling drivers and such but no luck until I tried a larger power supply. To my great surprise that fixed everything.

That's a problem with the individual unit. Not the model.

Stop with the non-issue, seriously. It's not the power supply.

 

[Captain_Tom]

17k on fire strike? that is 1070 territory?

They showed it off gaming next to a Titan XP, and PCgamer said they were performing almost identically.

That would put it at least 60% stronger than the 1070 even just to be near the 1080 Ti.


Something fishy is going on here...

 

[fullinfusion]

So what is it now!??
LOL!!

Should I buy a PSU with moar watts to get a better 3D Mark score??
Or no....

No your missing the point.. think of your mod, how well does it perform at 4.2v vs 3.4v? Understand

My 850 I had was hitting over 1100w at the wall.. pictures are here on the site somewhere.. it didn't enjoy that pull and after awhile it just ran like it wasn't sure what to do..

 

[pat-roner]

well he only has a 550W PSU when AMD recommends 850W for this card.(bet 750 would be fine though)

750w are you bonkers?

 

[KainXS]

At least we know that the air gpu is clocked at 1600 mhz. maybe it is the cpu that bottleneck the gpu.

AMD flat out says the peak clock is 1600MHz and the typical is 1382MHz but they never state the base clock. The reference RX 480 was clocked at 1266MHz but you had had to change the power limit so it would not power throttle and change the temp so it would not thermal throttle. So we don't know what clock its at really.

I've had personal experience tell me otherwise:

I had a i5-3570K (77w TDP, not overclocked) machine with a 600w Thermaltake PSU that worked well until I added a GTX 660 TI (150w). Everything was fine if a game wasn't graphically demanding but as soon as I played one that was the video would look really... odd and my sound would become scratchy and quickly cut out. I tried reinstalling drivers and such but no luck until I tried a larger power supply. To my great surprise that fixed everything.

EDIT: My system froze shortly after the 'symptoms' started and needed to be rebooted.

That does not sound right, I used my 3770k overclocked with an RX 480@ 1.3V and an 750Ti. Maybe that PSU you had was bad or a cheap PSU.

 

[Deleted member 172152]

Wasn't even running 1600mhz like it should. If it was running at something like 1400mhz that would mean a 1600mhz score of about 19.5k at least. Besides, we don't know what support's like, we don't know how well rx vega will do and most importantly, firestrike isn't an actual game. Lots of if's, dunno's and a guy doing the benchmarks that doesn't understand wattman. Great, now we still don't know how good rx vega is.

 

[Captain_Tom]

Wasn't even running 1600mhz like it should. If it was running at something like 1400mhz that would mean a 1600mhz score of about 19.5k at least. Besides, we don't know what support's like, we don't know how well rx vega will do and most importantly, firestrike isn't an actual game. Lots of if's, dunno's and a guy doing the benchmarks that doesn't understand wattman. Great, now we still don't know how good rx vega is.

Why aren't they using the Ultra 4K Firestrike bench? 1080p is almost completely irrelevant in GPU's this expensive.

 

[NGreediaOrAMSlow]

Are you really serious? If that card pulls the maximum supposed "300W", he still has 250(!!)W for the CPU and rest of the components.

A system with a 4790K would pull 100W at most.

100 Watts? More in the 120, 140oc according to this.
http://www.tomshardware.com/reviews/core-i7-4790k-devils-canyon-overclock-performance,3845-9.html

 

[Nokiron]

100 Watts? More in the 120, 140oc according to this.
http://www.tomshardware.com/reviews/core-i7-4790k-devils-canyon-overclock-performance,3845-9.html

Does not matter, even with that rating.

You are completely wrong with the power-supply rating by the way. You have misunderstood it completely.

 

[okidna]

Looking good for AMD If the promises that "RX Vega will actually be faster than Frontier version" is turned out to be true.

And I guess you forgot the power supply rating. Which is a two digit number representing the guarantee load. If branded PSU, then should be at least 80. Which means guarantee up to 80% of it's full capacity.

Non branded (generic) PSU are lower than that.

550*.80=440

Oh wow....

That's NOT what efficiency rating mean. To put it simple : Efficiency = (output power / input power)

More explanation : efficiency is a ratio between the output power (DC power generated by PSU) divided by input power (AC power needed by the PSU from electricity socket). The number will never be 1 (or 100%) because there's always certain amount of power lost during the AC to DC conversion. The higher the efficiency rating = less AC power consumed from the socket, and vice versa.

Simple example : you have a 550W, 85% rated efficiency (all load level for easy example) PSU, and your PC uses 350W of DC power. So, you actually use = (350W / 85%) = 411.76W of AC power from the socket. And when your PC actually uses 550W of DC power, you will pull 647W of AC power from the socket.

 

[Supercrit]

IMO that looks damn stupid!

I think I'm like a bird or something, I love shiny transparent things like gems etc.

And I guess you forgot the power supply rating. Which is a two digit number representing the guarantee load. If branded PSU, then should be at least 80. Which means guarantee up to 80% of it's full capacity.

Non branded (generic) PSU are lower than that.

550*.80=440

He has a guarantee load up to 440 Watts. While the power may reach 550, pushing it may have side effects.

Also a K processor. Do you think overclocking it will still consume the same?

It's the other way around, if the PSU manufacturer is not a dirty liar. 550w PSU should be able to output this amount to the components while pulling 550/0.8=687.5 from the wall.

 

[uuuaaaaaa]

Do you have a link to the Disqus article?

Thanks,
Miles.

It's the wccftech comment section on an article.

 

[Ruru]

Well, a new product and drivers are probably far from "ready" yet..

 

[xkm1948]

In case people forgot, this is the leaked version of VEGA running at 1200MHz core

Pay attention to the graphic score, 17801. I plot a little chart using all the data points available. It looks like a pretty good linear regression line to me.


When I have time tonight I will do a series of frequency vs GPU score plot for my FuryX.

 

[EarthDog]

If one extrapolates data from extrapolated data to form another extrapolated data point, how can we go wrong????????????????

 

[OSdevr]

That does not sound right, I used my 3770k overclocked with an RX 480@ 1.3V and an 750Ti. Maybe that PSU you had was bad or a cheap PSU.

I just cracked it open to see if anything was amiss. Hardly a thorough examination, but no signs of damage. I suppose if there was it wouldn't be working at all though.

Either way, it looks like a GPU can act strangely without sufficient power and not immediately crash the system.

 

[EarthDog]

And I guess you forgot the power supply rating. Which is a two digit number representing the guarantee load. If branded PSU, then should be at least 80. Which means guarantee up to 80% of it's full capacity.

Non branded (generic) PSU are lower than that.

550*.80=440

He has a guarantee load up to 440 Watts. While the power may reach 550, pushing it may have side effects.

Also a K processor. Do you think overclocking it will still consume the same?

Sir... that is NOT how it works... the efficiency rating is how much power the PSU is pulling FROM THE WALL. It DOES NOT take away from the label rating!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

So, if I was pulling an actual 500W load on a 90% efficient PSU, I would be pulling 550W FROM THE WALL.

Does that help clear things up????

Pascall GPUs will power throttle (IE run lower/disabled boost) if not fed enough juice. It's entirely possible that vega has the same ability.

They do (NVIDA GPUs....or Vega...........or are we hoping something sticks??) Have any links to this? It is something I never heard of before...

Either way, it looks like a GPU can act strangely without sufficient power and not immediately crash the system.

Or, just keep on thinking that is going to happen...


Let's get our heads out of our...err, the sand, can we?

 

[KainXS]

I just cracked it open to see if anything was amiss. Hardly a thorough examination, but no signs of damage. I suppose if there was it wouldn't be working at all though.

Either way, it looks like a GPU can act strangely without sufficient power and not immediately crash the system.

Normally when a GPU does not get enough power from the PSU it crashes the system or cuts off from its protection but with that system and 600W PSU, the PSU must have been bad or cheaply made. Even with this Vega card, a 550W PSU is not what I would recommend but if his PSU is a quality one(RM550 or something) it should run and the PSU should not be the limiting factor here. Thermaltake years ago made some really shifty PSU's on their low end, maybe that was your problem.

 

[R-T-B]

Don't kid yourself, I had a Corsair 850 psu and a pair of 7970's and it ran the benches.. well it completed them but the numbers were low. The psu was old and after installing a new psu the numbers went up

I'm sorry, but I'm pretty skeptical that that's what's going on here, or even was in your case. Low voltages / under delivered power don't cause low benchmarks, they cause instabilities, crashes, or even hardware damage. Most likely outcome of overloading a PSU is actually a hard shutdown, as it won't "not get enough juice." PSU's are a pull technology, they will attempt to deliver the requested wattage, or if built well, shutdown when unable. Whether they do it with acceptable ripple or without an electrical fire is another matter.

 

[Deleted member 172152]

In case people forgot, this is the leaked version of VEGA running at 1200MHz core

Pay attention to the graphic score, 17801. I plot a little chart using all the data points available. It looks like a pretty good linear regression line to me.


When I have time tonight I will do a series of frequency vs GPU score plot for my FuryX.

View attachment 89593

Still a 1080 score (just look at the graphics) in a dx11 synthetic bench with a dx12 optimized card with workstation optimizations as well! Not exactly representative of rx vega and the drivers will get some improvements, but it's still a decent score and a pretty good card for a gaming prosumer. New benches soon apparently and we should get a few more professional reviews once reviewers get their cards, which they had to buy with their own money.