[News] AMD Radeon Vega Frontier for 1,199 USD [Update: Benchmark]

[vario]

Seems like the Lunch will be provided at launch, 400W stove will be required to fry the schnitzel properly :-/ .
When i saw this 375W water cooled FE . I knew something bad is going to happen. I just knew.I had a weak hope that its some kinda miscommunication ,and thats the PCB`s max delivery power (for oc).But it seems were looking ar first single GPU 375-400W Card.The "winter heating" jokes are gonna be way worse than with bulldozer for f`cks sake.

[chew*]

I still may buy one. How bad can it be...i have 2 290x oced on water. 1 baseboard = less heat then 2 baseboards.

There was alot of hype behind BD...none with Llano.

No need to market what markets itself.

Not saying that is the case here...if i can get 1080 performance in the sub $500 range...im down for 1.

[SinOfLiberty]

Enthusiasts

[chew*]

Nothing says loving like cuddling up to your oven...er pc.

[vario]

They should make a dedicated Pc for the arctic circle, with Skylake-x and vega.
OVEN PC, an all in one solution for your computing/heating needs!

[BeepBeep2]

Not surprised.

Vega @ 300w TDP, I predicted to slot in between 1080 and Ti in performance. Frontier Edition isn't quite there, but it almost is, matching the 1080. RX Vega on RX drivers at 300w TDP on launch day @ Siggraph, I expect to do ~3-5% better than this on average which is a reasonable expectation given AMD's historical driver improvements.

This mythical 400w+ RX Vega, if it comes true, will be "Polaris Power Scandal" all over again... as I said, 75w+150w+150w= <400w. If AIB's introduce overclocked variants that do surpass 375w TDP, I expect them to have 8+8+6 pin PCI-E power. It would be hard for me to believe AMD would shoot itself in the foot again with how the Polaris power draw breaking spec was all over teh interwebz.

TDP wise, based on these results I expect a
<200w 1070 competitor (+~5% performance) <--- RX Vega Nano, maybe 56 CU
300w 1080 competitor (+~5% performance) and <--- RX Vega
375w Ti competitor (~-0-5% performance, priced <$600) for consumers. <--- RX Vega X

R9 Nano delivered 85-90% of Fury X performance at half the power consumption. That architecture is efficient enough for compute that they are still using it w/ similar TDP for Radeon Instinct MI8. I expect and have always expected something similar with Vega.

Do I expect all three variants to show up at the end of July? No. Probably announcements in July, one card launching, other ones "this summer".

The only scenarios which I forsee only one RX Vega SKU are if yields are in the toilet (possible) and / or HBM2 supply is so constrained that they can't keep more than one SKU in stock.


EDIT:

I still may buy one. How bad can it be...i have 2 290x oced on water. 1 baseboard = less heat then 2 baseboards.

Man, I miss my R9 290s, and those 90+ degree summer days before they were stolen. Rockin' 5770's (air) and 6970's (H2O) over here.

[zanzabar]

TDP wise, based on these results I expect a
<200w 1070 competitor (+~5% performance) <--- RX Vega Nano, maybe 56 CU
300w 1080 competitor (+~5% performance) and <--- RX Vega
375w Ti competitor (~-0-5% performance, priced <$600) for consumers. <--- RX Vega X

that would be really good scaling on the clock speed for the 64cu parts if it is only 75W to gain ~30% for 75W. if anything i would assume a lower shader part that is overclocked and then a full shader higher end one with the same clock speed and the nano like one with low cu and lower clocks.

[BeepBeep2]

that would be really good scaling on the clock speed for the 64cu parts if it is only 75W to gain ~30% for 75W. if anything i would assume a lower shader part that is overclocked and then a full shader higher end one with the same clock speed and the nano like one with low cu and lower clocks.

I predicted
300w +5% performance over 1080,
375w -0 to -5% performance under 1080 Ti

1080 Ti is on average about 25% faster than the 1080 in games.

~15% performance for 375w SKU (1382 typical clock to 1550 MHz+ typical clock) I believe can happen with 25% (300-375w) higher power consumption.

Something funny here is that both the 300w air cooled Vega FE and 375w water cooled Vega FE are rated for "peak" "13.1 TFLOPS" and 1600 MHz "peak" clock speed, but I am guessing the 300w version likes to stick around 1382 (or less) much more than the water cooled one.

Who knows what will actually happen though, I am curious.

[SinOfLiberty]

[chew*]

Well i always wanted to do one of those electric fireplace flame effects as a side window case mod. Now i will have a theme that is fitting.

[SinOfLiberty]

It is summah time!

AMD, good strategy, watch `em burn in hell

[SinOfLiberty]

dandandnn

[AliG]

Nothing says loving like cuddling up to your oven...er pc.

You know, I was planning on buying a kamado grill this summer. Maybe I can consolidate costs...

And just get a Vega GPU instead

[Dimitriman]

More benchmarks surfacing:

https://www.techpowerup.com/234775/n...dition-surface

Not much more to say about it though, only that in a few games it comes closer to the 1080Ti, but mostly its around 1080 level as per previous tests.

[SinOfLiberty]

Damn, son 1080 ti lightning is Pile Of Crp/ NV really does not allow anyone to break any time limits.

End of July, that is.

[chew*]

Sometimes the cup is half full.

Best news i have heard so far regarding vega.

The irony is i can probably sell my 2 290x on Powercolor LCS bios with full cover aqua computer waterblocks to miners and break even without shelling a dime out of pocket for vega other than...a new waterblock.

Since i had 2 290x...im already power supplied up lol.

"Apparently, there was some testing done on mining software as well, and performance is reported as disappointing (as in, "very low".) This probably speaks to the HBC (High Bandwidth Cache) and HBCC (high Bandwidth Cache Controller), which probably will require a lot of fine tuning from mining software (remember the GTX 1080 is generally poor in mining workloads compared to the GTX 1070 because of the higher latency of its GDDR5X memory implementation.) Perhaps these news come as a relief, however, since availability of RX Vega cards will likely be limited without miners taking up the whole supply."

[Aleki]

The owner of that card was running 17.1.1 Drivers..
Those drivers are six months old

https://www.techpowerup.com/234737/a...ed-benchmarked

PS: Sorry for the large quote, was looking to hide, but this forum does not support spoiler tags

[Hell Hound]

Help I can get a 980 today for $220 @ micro,should I wait for vega.I'm looking @ power to perf/ratio.
2 gtx980's/fury nano's seems to be where vega will land but for how much.

[BenchZowner]

The owner of that card was running 17.1.1 Drivers..
Those drivers are six months old

That is not true.
The drivers are the latest ones AMD has made available for Vega and they are like 1 1/2 month old.

[Tim]

https://www.pcper.com/reviews/Graphi...Cooled-Review/

Ouch. AMD blown their own foot off releasing Frontier first. Even if Vega is better, having the Frontier perform worse then an 1080 is really not great! And that power level!!! Sheesh.

Vega RX better bring something a heck of a lot better, but it's very much starting to look like a disaster.

[zanzabar]

https://www.pcper.com/reviews/Graphi...Cooled-Review/

Ouch. AMD blown their own foot off releasing Frontier first. Even if Vega is better, having the Frontier perform worse then an 1080 is really not great! And that power level!!! Sheesh.

Vega RX better bring something a heck of a lot better, but it's very much starting to look like a disaster.

how would rx vega be better. they might pick up some performance from drivers but on the hardware side the only thing they could fix is using full spec hbm. the clock speed wont come up unless they have a TSMC fabed part since the samsung process is garbage.

[Kai Robinson]

Someone mentioned that only 2 out of the 10 architectural changes for Vega are currently active in the drivers, the tile based rasteriser isn't supported yet and basically it's running in Fiji fallback mode for a lot of stuff.

[Tim]

Someone mentioned that only 2 out of the 10 architectural changes for Vega are currently active in the drivers, the tile based rasteriser isn't supported yet and basically it's running in Fiji fallback mode for a lot of stuff.

They just launched a high end professional card that needs to run at such a high level in the professional market and people say 80+% of it isn't working in drivers?

I'm not sure whether to cry or laugh at that!

[BenchZowner]

Someone mentioned that only 2 out of the 10 architectural changes for Vega are currently active in the drivers, the tile based rasteriser isn't supported yet and basically it's running in Fiji fallback mode for a lot of stuff.

It's not that much different from Fiji you know.

Also most of the "old drivers" "info" is bs and wishful thinking.

TBR isn't working properly, now that can be anything from not ready yet or AMD failing to get it to work properly.

Either way, with every feature-set "activated" and "optimized" drivers, and perhaps increased clock speeds, it can potentially get 20-15% close to the 1080 Ti on average.

[Aleki]

That is not true.
The drivers are the latest ones AMD has made available for Vega and they are like 1 1/2 month old.

My bad, read the statement in the review after you pointed it out; GPU-z reports driver as 17.1.1

They just launched a high end professional card that needs to run at such a high level in the professional market and people say 80+% of it isn't working in drivers?

I'm not sure whether to cry or laugh at that!

I think the way they saw it was, it's a $1k workhorse card that can easily manage itself against similar nVidia Pro cards. Even if 80% of the features are disabled, it's a pretty compelling card for its price(in the workspace environment); the extra sauce to be unlocked in the future is just a bonus.

Despite their gaming performance, I strongly believe the RX cards will still sell well, if not for GPU mining alone