Vega FE - Placebo Mode

[DXMember]

1 hour ago, PCGuy_5960 said:

No it doesn't, a fully working 7980XE die costs way more than $30. 

Potassium

sure, but they don't have a fully working 7980XE, do they? (:

I was talking more about the quad cores

[Grinners]

10 hours ago, DXMember said:

sure, but they don't have a fully working 7980XE, do they? (:

I was talking more about the quad cores

I eagerly await the release of your quad core seen as you have identified such an arbitrage opportunity and you can produce them so much cheaper than Intel.

[DXMember]

4 hours ago, Grinners said:

I eagerly await the release of your quad core seen as you have identified such an arbitrage opportunity and you can produce them so much cheaper than Intel.

I can't, but Intel can, but they won't  do it. Why don't they do it? Because free market, fucking capitalist pigs.

[mr moose]

1 hour ago, DXMember said:

I can't, but Intel can, but they won't  do it. Why don't they do it? Because free market, fucking capitalist pigs.

 You can't.  Absolutely you can't. Hardly anyone can.  Who's best placed to do what Intel is right now? AMD is, but are they doing a bang up job? They have 1 line of CPU's worth buying right now after how many years?  You don't understand corporate economics, you think it is all about greed.  It's not a product you can mock up in a few months in your back yard and then forget about it once sold, Intel literally invests billions $$$ in R+D alone.  They are not going to make that back selling their products at a little over cost.

[Mr.Meerkat]

On 2017-7-8 at 5:23 AM, PCGuy_5960 said:

Can we all just agree that Vega FE was 100% rushed?

Nobody said it wasn't to be fair  

 

Anyway, I do hope AMD pulls a Hawaii (again), a card that was considerably slower than the top kepler (780ti) which after many years of driver updates is now almost a 980 perf wise however they will have to do it in like 3 months as after all, AMD is known for the finewine nature of their cards although in this instance, they don't have the time. 

Basically what I'm saying is hopefully the more mature drivers that comes out in the next few months is able to give Vega a nice big kick in performance  

[DXMember]

7 hours ago, mr moose said:

 You can't.  Absolutely you can't. Hardly anyone can.  Who's best placed to do what Intel is right now? AMD is, but are they doing a bang up job? They have 1 line of CPU's worth buying right now after how many years?  You don't understand corporate economics, you think it is all about greed.  It's not a product you can mock up in a few months in your back yard and then forget about it once sold, Intel literally invests billions $$$ in R+D alone.  They are not going to make that back selling their products at a little over cost.

they should consider maybe 10% perf over 6 years wasn't worth the billions upon billion $$$ in R+D after all?

[cj09beira]

7 hours ago, Mr.Meerkat said:

Nobody said it wasn't to be fair  

 

Anyway, I do hope AMD pulls a Hawaii (again), a card that was considerably slower than the top kepler (780ti) which after many years of driver updates is now almost a 980 perf wise however they will have to do it in like 3 months as after all, AMD is known for the finewine nature of their cards although in this instance, they don't have the time. 

Basically what I'm saying is hopefully the more mature drivers that comes out in the next few months is able to give Vega a nice big kick in performance  

because thee is a lot of stuff that seems to not be working, it might become faster rapidly 

[Valentyn]

Not too far behind the 1080 here, considering the FE is running at 1440Mhz, and doesn't come close to max boost of 1600Mhz. I was surprised by the low temps on it though; so wonder why it didn't boost higher.
 

 

[cj09beira]

2 minutes ago, Valentyn said:

Not too far behind the 1080 here, considering the FE is running at 1440Mhz, and doesn't come close to max boost of 1600Mhz. I was surprised by the low temps on it though; so wonder why it didn't boost higher.
 

 

might be power throttling 

[tom_w141]

Anyone still hoping a driver is going to pull out 40-50% performance is high AF lol.

[cj09beira]

11 minutes ago, tom_w141 said:

Anyone still hoping a driver is going to pull out 40-50% performance is high AF lol.

there might be some good gains, as vega should have much higher geometry horsepower, and the rasterizer will help too, both don't seen to be enabled right now

[Bananasplit_00]

On 2017-07-07 at 11:10 PM, tmcclelland455 said:

It's a brand new card so naturally the drivers aren't going to work as they should. Common sense, people.

and its AMD, which never launch cards with good drivers either lol. the R9 Fury is a perfect example and so is the entire re:live driver update.

[JediFragger]

17 minutes ago, Bananasplit_00 said:

and its AMD, which never launch cards with good drivers either lol. the R9 Fury is a perfect example and so is the entire re:live driver update.

Plus let's not forget the AMD 480 launch, that awesome 970 competitor

[Bananasplit_00]

2 minutes ago, JediFragger said:

Plus let's not forget the AMD 480 launch, that awesome 970 competitor

hmm it wasent that bad was it? i dont remember any major things fixed beyond haveing it draw more then the rated current from the PCIe slot(which would have been safe tbh, they are massivly under rated for what they can actiually take)

[cj09beira]

1 minute ago, Bananasplit_00 said:

hmm it wasent that bad was it? i dont remember any major things fixed beyond haveing it draw more then the rated current from the PCIe slot(which would have been safe tbh, they are massivly under rated for what they can actiually take)

its lower because its limited by what a 24pin connector can deliver, so if you have multiple cards, its better to have lower pcie power per card or have added connectors just for pcie.

some overclockers have managed to blown their 24pin because of that in the past

[Bananasplit_00]

1 minute ago, cj09beira said:

its lower because its limited by what a 24pin connector can deliver, so if you have multiple cards, its better to have lower pcie power per card or have added connectors just for pcie.

some overclockers have managed to blown their 24pin because of that in the past

thats with 980TIs and other similar cards, mainly dual GPU ones. for a single GPU card at stock there was no way that it wouldent handle it, the only reason it was adressed was because it was outside of the PCIe spec by a few watts, it wasent a problem beyond that and i doubt there was a recent motherboard out there for consumers that couldent handle it(OEM boards dount count because they are asholes and i say recent as there are some older motherboards on PCIe 2.0 that have the power cut down)

[cj09beira]

21 minutes ago, Bananasplit_00 said:

thats with 980TIs and other similar cards, mainly dual GPU ones. for a single GPU card at stock there was no way that it wouldent handle it, the only reason it was adressed was because it was outside of the PCIe spec by a few watts, it wasent a problem beyond that and i doubt there was a recent motherboard out there for consumers that couldent handle it(OEM boards dount count because they are asholes and i say recent as there are some older motherboards on PCIe 2.0 that have the power cut down)

what i mean is that the pcie spec it self, is lower than it could be to help reducing the chances of multiple cards on the same motherboard blow the 24 pin connector, (2,3+ cards)

[Okjoek]

It honestly looks like AMD is in a bad spot with this one to me. It sucks because I don't want to see AMD fall deep behind on the GPU war right as they're getting their act together on the CPU front.

 

On one hand I don't care one bit about high-end hardware. But I hate to see people or groups who are never able to recover from a mistake they make. I hope that's not the case because it just seems like nVIDIA has so much better resources to the point where they can simply toy with AMD by launching their next product when it looks like AMD is even dreaming of catching up and that's not really the fair fight I like to see.

[Humbug]

1 hour ago, Valentyn said:

Not too far behind the 1080 here, considering the FE is running at 1440Mhz, and doesn't come close to max boost of 1600Mhz. I was surprised by the low temps on it though; so wonder why it didn't boost higher.

Well Vega is not a hot chip.

It's actually quite cool running considering the high TDP and blower coolers that we have got so far. People assume the chip stays hot when they see the TDP numbers.

 

The below test was done by Gamers Nexus to show what temperature various GPUs can maintain at a fixed noise level.

The coolers were not allowed to spin up past 40dBA.

 

 

This I is a very good performance from a blower cooler for a high TDP card.

It means that for whatever reason Vega is very efficient at dissipating it's heat...

[cj09beira]

1 minute ago, Okjoek said:

It honestly looks like AMD is in a bad spot with this one to me. It sucks because I don't want to see AMD fall deep behind on the GPU war right as they're getting their act together on the CPU front.

 

On one hand I don't care one bit about high-end hardware. But I hate to see people or groups who are never able to recover from a mistake they make. I hope that's not the case because it just seems like nVIDIA has so much better resources to the point where they can simply toy with AMD by launching their next product when it looks like AMD is even dreaming of catching up and that's not really the fair fight I like to see.

well, vega will probably help them enter the server space because of its compute abilities, which for amd is probably a much more important market as its a huge market, that and all ryzen skus will help amd to have larger R&D for 2 gens from now (as the next one is probably reaching end of development soon)

[Valentyn]

Just now, Humbug said:

Well Vega is not a hot chip.

It's actually quite cool running considering the high TDP and blower coolers that we have got so far. People assume the chip stays hot when they see the TDP numbers.

 

The below test was done by Gamers Nexus to show what temperature various GPUs can maintain at a fixed noise level.

The coolers were allowed to spin up past 40dBA.

 

 

This I is a very good performance from a blower cooler for a high TDP card.

It means that for whatever reason Vega is very efficient at dissipating it's heat...

I saw that test alright. The Frontier Edition blower is as good as the MSI Armor was a surprise!

Beats the pants off the NV Reference; although it seems the concentrated heat from HBM+GPU is an issue under some loads.

 

PcPer seeeing it settle at 83 degrees; similar to NV reference; but FE still not able to clock above 1440Mhz.

So it seems there's more to the boosting issue than just temps.

 

Also saw this posted on Overclockers.co.uk. If that power usage numbers for 16GB HBM2 is true that's a nice extra 75W 8GB RX Vega can spend on GPU frequency.

 

https://disqus.com/by/klaudiuszkaczmarzyk/comments/

 

[Humbug]

1 hour ago, cj09beira said:

there might be some good gains, as vega should have much higher geometry horsepower, and the rasterizer will help too, both don't seen to be enabled right now

1 hour ago, tom_w141 said:

Anyone still hoping a driver is going to pull out 40-50% performance is high AF lol.

I Don't really see people expecting that anymore. That would make it ahead of a 1080ti.

 

I think now people are just expecting Rx Vega to come out and beat the GTX 1080 with it's gaming drivers, at a slightly lower price.

[cj09beira]

2 minutes ago, Humbug said:

Well Vega is not a hot chip.

It's actually quite cool running considering the high TDP and blower coolers that we have got so far. People assume the chip stays hot when they see the TDP numbers.

 

The below test was done by Gamers Nexus to show what temperature various GPUs can maintain at a fixed noise level.

The coolers were allowed to spin up past 40dBA.

 

 

This I is a very good performance from a blower cooler for a high TDP card.

It means that for whatever reason Vega is very efficient at dissipating it's heat...

amd has reduced its CU/mm^2 which should help

 

2 minutes ago, Valentyn said:

I saw that test alright. The Frontier Edition blower is as good as the MSI Armor was a surprise!

Beats the pants off the NV Reference; although it seems the concentrated heat from HBM+GPU is an issue under some loads.

 

PcPer seeeing it settle at 83 degrees; similar to NV reference; but FE still not able to clock above 1440Mhz.

So it seems there's more to the boosting issue than just temps.

 

Also saw this posted on Overclockers.co.uk. If that power usage numbers for 16GB HBM2 is true that's a nice extra 75W 8GB RX Vega can spend on GPU frequency.

 

https://disqus.com/by/klaudiuszkaczmarzyk/comments/

 

hbm does not use 150w ahaha,  

[Valentyn]

6 minutes ago, cj09beira said:

amd has reduced its CU/mm^2 which should help

 

hbm does not use 150w ahaha,  

Unless it's "overclocked" and overvolted to get to 1.89Gbps; since Hynix are still only listed 1.6Gbps HBM2.

Buildzoid did that PCB teardown of the Frontier Edition and was concerned the HBM2 will be overvolted from what he saw.

 

Still lots to wait and see for.

[cj09beira]

according to buildzoid the card many times reaches its 300w limit.

 hbm2 consumption is super low 20-40w (made up numbers)

there are reasons to believe the tiled based rasterizer reduces power consumption as it reduces the amount of work the gpu has to do by not drawing certain triangles much earlier in the pipeline, but that has to be tested latter