AMD Radeon RX Vega Put Through 3DMark

[B-Real]

Gamers but quiet, quiet cards don't double the heat output of their competitor. You know how I know this? Sales figures for the last 10 years.

If performance per watt didnt matter amd wouldn't have been only 28% of the market share when the 290/390 was selling.

What I saw was that the AIB models of 290, 390 produced not much more heat than the 970 980 models. Same with the 480/580-1060. But ok, you tell that people are buying cards that have better performance/watt. I just don't accept it.

 

[R00kie]

Who cares about power consumption when we speak about a 400-500-700$ card?

People who care about having a cool and quiet card in their PC. Nobody wants a card that either heats your loins up to a crisp, or takes half of your slots in a case.

What I saw was that the AIB models of 290, 390 produced not much more heat than the 970 980 models. Same with the 480/580-1060. But ok, you tell that people are buying cards that have better performance/watt. I just don't accept it.

Sorry, but that was not the case at all, and I will be living proof of this. I bought an MSI R9 290 with a TFIV cooler, and that thing was still hot as the sun. Then I finally decided to get a GTX 970 just because it was a lot cooler and ate a lot less power.

 

[B-Real]

Gamers but quiet, quiet cards don't double the heat output of their competitor. You know how I know this? Sales figures for the last 10 years.

If performance per watt didnt matter amd wouldn't have been only 28% of the market share when the 290/390 was selling.

Well, you say that gamers chose after performance/watt. I say that at the 390 times, 970 was at the same price, delivered same performance, but consuming less power (without tuning AMD cards). I say if one manufacturer delivers a GPU that is significantly cheaper than the other while performing nearly the same altough consuming more power, people will buy the former one.

 

[Daven]

In line with everything we've seen until now.
RX Vega == Vega FE Gaming mode.
As it should be.

Not quite. Again the 1080 was overclocked to a whopping 1924 MHz. I don't believe the Vega FE Gaming Mode was compared to the 1080 clocked this high.

 

[B-Real]

People who care about having a cool and quiet card in their PC. Nobody wants a card that either heats your loins up to a crisp, or takes half of your slots in a case.



Sorry, but that was not the case at all, and I will be living proof of this. I bought an MSI R9 290 with a TFIV cooler, and that thing was still hot as the sun. Then I finally decided to get a GTX 970 just because it was a lot cooler and ate a lot less power.

Yep, it was true for the 290. ~+10 Celsius compared to the 970. Then check the 390: https://www.techpowerup.com/reviews/Sapphire/R9_390_Nitro/27.html Nearly same temps as the 970 with nearly the same fan noise. Still, the AIB versions of 290 producing a bit above 70 Celsius while the reference getting 94 Celsius (about 20 Celsius cooler) and making about 15dB louder noise is absolutely not a bad result, given their TDP.

 

[R00kie]

Yep, it was true for the 290. ~+10 Celsius compared to the 970. Then check the 390: https://www.techpowerup.com/reviews/Sapphire/R9_390_Nitro/27.html Nearly same temps as the 970 with nearly the same fan noise.

Yet with far greater power consumption, same performance and slightly higher price tag.

 

[Frick]

Well, you say that gamers chose after performance/watt. I say that at the 390 times, 970 was at the same price, delivered same performance, but consuming less power (without tuning AMD cards). I say if one manufacturer delivers a GPU that is significantly cheaper than the other while performing nearly the same altough consuming more power, people will buy the former one.

It wasn't as bad a difference as this. People started caring when Fermi came out. As cdawall said, some do not care, but in general people care more (which is a good thing).

 

[B-Real]

Yet with far greater power consumption, same performance and slightly higher price tag.

That's why I said earlier that if 2 cards come out with the same price tag and same performance, what people can consider? Performance/Watt ratio and drivers.

 

[EarthDog]

The Geforce GTX 1080 was overclocked to 1924 MHz which is well above stock/turbo/max clocks of 1607/1733/1800 MHz. This is pretty much the best possible score a GTX 1080 can have overclocked almost to the max.

If the RX Vega at 1630 MHz is stock or even turbo, they will be in a good position. I calculate the card will be halfway between a stock GTX 1080 and a stock GTX 1080 Ti.

Rumored power consumption of 375W still suuuuuuuuuuuuuuuuuuuccccckkkkkkkssss.

Hey mark... actual boost clocks go around 100Mhz over whatever is listed. Its likely a simple factory overclocked card. Most 1080's will do 2000 MHz overclocked manually. Not to mention, 3Dmark isn't good at reading clockspeeds in the first place... so its a tough call.

 

[R00kie]

That's why I said earlier that if 2 cards come out with the same price tag and same performance, what people can consider? Performance/Watt ratio and drivers.

Which is why RX Vega is not going to be a go to card this late to the party.

 

[r9]

If its cheaper compared to 1080, this will sell very well.

I'm sure they will ask more and play the gsync cheaper monitors card.

 

[Daven]

Hey mark... actual boost clocks go around 100Mhz over whatever is listed. Its likely a simple factory overclocked card. Most 1080's will do 2000 MHz overclocked manually. Not to mention, 3Dmark isn't good at reading clockspeeds in the first place... so its a tough call.

To put things into perspective, here is the TP review of the Gigabyte GTX 1080 Aorus Xtreme Edition:
https://www.techpowerup.com/reviews/Gigabyte/GTX_1080_Aorus_Xtreme_Edition/

The card is clocked at 1759+ MHz. Even with another 100 MHz turbo, the card would not beat Vega in the 3dMark benchmark (not that means much as we need real world gaming benchmarks). Also the peak power consumption of the Gigabyte card was 243W. I'm sure the power would be even more at 1924 MHz.

Vega is equivalent to a very high, manually overclocked 1080 GTX. That's all I'm saying.

 

[cdawall]

What I saw was that the AIB models of 290, 390 produced not much more heat than the 970 980 models. Same with the 480/580-1060. But ok, you tell that people are buying cards that have better performance/watt. I just don't accept it.

Accept what you want. Name another reason why anything released post 290 was garbage for sales figures? Mining is the only reason Q2 2016 numbers moved.

 

[HD64G]

That's why I said earlier that if 2 cards come out with the same price tag and same performance, what people can consider? Performance/Watt ratio and drivers.

Freesync vs G-Sync cost is a good factor for a gamer's decision also. And new features for last and next gen games alongside with the better performance on last graphic APIs as DX12 and Vulcan should be considered as well. If Vega RX manages to be on par with factory oced 1080s for the same price, it will be sold easily. Some request 1080 performance for 1070 price which is suicidal for any enterprice to do such stupidity to itself. Let's hope now (not very possible judging by AMD's past new arch GPU releases) for reviews on mature drivers with all the features enabled to get a clear picture of where RX Vega will stand.

 

[chief-gunney]

Sounds ok, will give me the performance I'm looking for as an upgrade from the RX 480. I'll be putting a water block on the one with the best vrm's. I hope the pricing is good.
Ive got a 1440p Freesync monitor so no Nvidia ever again for me.

From what I can surmise, Vega is a stepping stone to Navi which will see a dual vega on infinity fabric which the OS sees as single gpu. Nvidia are worried about that and I believe they're working on doing something similiar.

 

[Imsochobo]

Not really unexpected, based on the performance of polaris. Most educated guesses put the VEGA chip at 1080 level at best. The fact that it is consistently faster then a 1070 is a bit better then expected. Unfortunately, it also confirmed that AMD completely threw this GPU generation to nvidia for anything other then low end cards, and does not bode well for volta VS navi.

Of course, that assumes this is the full 4096 core part. Based on FE's performance, I have little reason to believe otherwise, but still.
At which point volta will be out and dominating what is left of VEGA.

AMD needs to deliver at launch, not a year later.

true, but it may have additional 10% left in it until actual launch, unless above is soo true.
if it has 10% more perf than we expect then it may have an actual chance to sell.

I really really want to get away from crapvidia linux drivers, but I still want to game at 4K so crossing my finger but I won't buy it unless it surpasses GTX1080 by a good margin

 

[Dimi]

Not quite. Again the 1080 was overclocked to a whopping 1924 MHz. I don't believe the Vega FE Gaming Mode was compared to the 1080 clocked this high.

You couldn't be more wrong. First off, the Vega was probably clocked to the absolute max. Secondly, a 1080 can go a lot higher than a measely 1924 mhz. Thats probably just a weak boost without ANY OC at all. Most 1080's will go to around 2025-2050 comfortably.

 

[schrank251]

"This puts its performance either on-par or below that of the GeForce GTX 1080..."

No.

The Geforce GTX 1080 was overclocked to 1924 MHz which is well above stock/turbo/max clocks of 1607/1733/1800 MHz. This is pretty much the best possible score a GTX 1080 can have overclocked almost to the max.

If the RX Vega at 1630 MHz is stock or even turbo, they will be in a good position. I calculate the card will be halfway between a stock GTX 1080 and a stock GTX 1080 Ti.

Rumored power consumption of 375W still suuuuuuuuuuuuuuuuuuuccccckkkkkkkssss.

No, i have the same GTX 1080 and even at stock it boots up higher than that in the table. Firestrike shows the peak GPU clock, at stock mine boots up to about 2 GHz but average is about 1950 MHz, resulting in a 23k graphics score.
Here is my worst stock result.
http://www.3dmark.com/fs/11885095
Overclocked i get 25k graphics score.
http://www.3dmark.com/fs/12177709

 

[EarthDog]

To put things into perspective, here is the TP review of the Gigabyte GTX 1080 Aorus Xtreme Edition:
https://www.techpowerup.com/reviews/Gigabyte/GTX_1080_Aorus_Xtreme_Edition/

The card is clocked at 1759+ MHz. Even with another 100 MHz turbo, the card would not beat Vega in the 3dMark benchmark (not that means much as we need real world gaming benchmarks). Also the peak power consumption of the Gigabyte card was 243W. I'm sure the power would be even more at 1924 MHz.

Vega is equivalent to a very high, manually overclocked 1080 GTX. That's all I'm saying.

We haven't seen a manually overclocked 1080 compared to it yet. As I said, that is likely a factory overclocked card.

Lesson on GPUz and boost: There is baseclock, boost, and actual clocks. The 1759 Mhz is the BASE CLOCK for that card. The BOOST clock is typically around 100 MHz above that. The BOOST clock is the MINIMUM boost, not the ACTUAL clocks. YOu can take what GPUz reads and generally add 100 MHz or so for ACTUAL clocks.

As far as the power rating, that is with the clocks stock, around 2000 MHz for the AORUS. Here is a snip from the article you linked. Manual overclocking had it going over 2100 MHz...default, was 2038. Youi can bet your arse the Vega will show the same increase in power when AIB's have their factory versions out too raising the already high values of 300/375W.

Maximum overclock of our sample is 1470 MHz on the memory (15% overclock) and +88 MHz to the GPU's base clock, which increases maximum Boost from 2038 MHz to 2126 MHz (4% overclock).

You have to know how to read GPUz and how boost works bud. I agree with your end point mostly (performs like a factory OC 1080), but not how you managed to get there as that was just plain off.

 

[PowerPC]

Just release it already! Nobody will care, if it's priced right. Only a handful of people are ever interested in $800 cards, so who cares it doesn't beat 1080 ti? It's such a petty argument. The only factor that ever matters with AMD will be price/performance, as always. It's great that there is a $500.000 Lamborghini on the market, but most people will still go out and buy the more affordable Subaru.

 

[Dimi]

To put things into perspective, here is the TP review of the Gigabyte GTX 1080 Aorus Xtreme Edition:
https://www.techpowerup.com/reviews/Gigabyte/GTX_1080_Aorus_Xtreme_Edition/

The card is clocked at 1759+ MHz. Even with another 100 MHz turbo, the card would not beat Vega in the 3dMark benchmark (not that means much as we need real world gaming benchmarks). Also the peak power consumption of the Gigabyte card was 243W. I'm sure the power would be even more at 1924 MHz.

Vega is equivalent to a very high, manually overclocked 1080 GTX. That's all I'm saying.

Did you even read the review? You link a review of a card that OC's to 2126 MHz according to the review yet you think this card clocks at a max of 1859 mhz..........

 

[cdawall]

To put things into perspective, here is the TP review of the Gigabyte GTX 1080 Aorus Xtreme Edition:
https://www.techpowerup.com/reviews/Gigabyte/GTX_1080_Aorus_Xtreme_Edition/

The card is clocked at 1759+ MHz. Even with another 100 MHz turbo, the card would not beat Vega in the 3dMark benchmark (not that means much as we need real world gaming benchmarks). Also the peak power consumption of the Gigabyte card was 243W. I'm sure the power would be even more at 1924 MHz.

Vega is equivalent to a very high, manually overclocked 1080 GTX. That's all I'm saying.

Did you even read that review?

Maximum overclock of our sample is 1470 MHz on the memory (15% overclock) and +88 MHz to the GPU's base clock, which increases maximum Boost from 2038 MHz to 2126 MHz (4% overclock).

Stock boost was 2038

 

[dozenfury]

Wattage I'm not as concerned about, I have dozens of appliances that use more than the 120w difference between the Vega and 1080. The monthly power cost difference is relatively minimal unless you are mining at 100% 24x7. What I'm more concerned with is heat, since that heats up all of the other components in my PC and even to a small degree the room temp. Heat also is going to likely hurt oc headroom, which has been an issue on recent AMD cards.

If the benches are accurate, I'd put a fair price on a RX Vega at $399. That would beat the 1070 on pure performance/price (I paid $399 for my 1070 new), but also offsets the higher power and other minor AMD cons. The NV platform is much more mature now having been out a year, oc's well, gets quicker driver updates than AMD, and will probably run cooler. Any more than $399 and I'd happily buy a 1070 for the same price, a 1080 in the low $500's, or go $700 for a much faster 1080ti. That's all excluding the mining price shenanigans which should eventually wind down.

 

[DRDNA]

I was really looking forward to this release and I thought it would be a pinch better.

 

[Durvelle27]

This could possibly be the lower end Vega considering there will be 3 different Skus