videocardz AMD Radeon R9 290X Memory Bus: 512-bit

[3DVagabond]

I just want to lash out in frustration and anger that it is taking this long for the goodies to come out. Never thought I would say this but I wish we were back on the 6 month new gen cycle... This 1.5 year thing is terrible for everything except my wallet. Even then my wallet still suffers because I get so bored that I just buy new cards to try something different.

Gotta feed the addiction!

Seriously, good on ya' for having fun. You haven't been too bored though. If I'm not mistaken you've owned every high end card the previous gen and the refresh so far. Hawaii will just be another. I hope it meets with your approval. :thumbsup:

 

[3DVagabond]

Did a lot of you really miss the "AMD Sez" part of Kyle's post?

No, I didn't. I took that as justification for Kyle to post it as I'm sure he's under NDA about any knowledge he has due to his tester status.

 

[blackened23]

Why would you make all that crap up? And then admit it and post it here? :thumbsdown:

He could be lying and trolling this forum about trolling R. Who says he's telling the truth? And who cares? Anyone can make any claim around here. Anyway, there is no reason to take anyone around here at face value, besides which, who the hell cares. Off topic btw, don't propagate this nonsense. Take it to the Steam Box thread in PC gaming.

 

[3DVagabond]

Mhm. Leave it to nVidia's marketing team to always have a card up their sleeve...

If the rumors are true about Hawaii's performance, I don't expect nVidia to take this lying down. I would expect more than price cuts too. Personally the Titan ultra with unlocked voltage and a cooler that can keep it cool under high voltages should do it. I can only see a couple of possible problems. 1, backlash from their previous high prices. Although the "blame it on AMD' marketing strategy should be able to buffer that a bit. 2, Maybe they don't have enough fully functioning chips to supply the consumer market with ~$600 cards without shortages. The fact they sell Titan for a grand, the law of supply and demand insinuates they don't have heaps of even 2nd level binned chips. There just might not be enough "Ultra" gpu's to compete at lower price points.

 

[Jaydip]

If the rumors are true about Hawaii's performance, I don't expect nVidia to take this lying down. I would expect more than price cuts too. Personally the Titan ultra with unlocked voltage and a cooler that can keep it cool under high voltages should do it. I can only see a couple of possible problems. 1, backlash from their previous high prices. Although the "blame it on AMD' marketing strategy should be able to buffer that a bit. 2, Maybe they don't have enough fully functioning chips to supply the consumer market with ~$600 cards without shortages. The fact they sell Titan for a grand, the law of supply and demand insinuates they don't have heaps of even 2nd level binned chips. There just might not be enough "Ultra" gpu's to compete at lower price points.

I believe NV took a stupid decision when they didn't allow AIBs to have custom Titans/690s.It's not like that AIBs can't create a better design than NV.Anybody here remembers ASUS MARK II ? that was one hell of a card except for the price.

 

[24601]

I believe NV took a stupid decision when they didn't allow AIBs to have custom Titans/690s.It's not like that AIBs can't create a better design than NV.Anybody here remembers ASUS MARK II ? that was one hell of a card except for the price.

For the Titan they probably did it for validation reasons, since I doubt the AIBs will do all the validation required for error free compute performance.

AMD sure as hell didn't with their AIB 7970s.

 

[Jaydip]

For the Titan they probably did it for validation reasons, since I doubt the AIBs will do all the validation required for error free compute performance.

AMD sure as hell didn't with their AIB 7970s.

Each and every design from AIBs gets validated by NV/AMD.There is just no way around it.

 

[24601]

Each and every design from AIBs gets validated by NV/AMD.There is just no way around it.

Sure explains why late model XFX Double Dissipation Black Edition 1000 mhz 7970s invariably fail at compute then, huh?

Same thing with late model Gigabyte Windforce 1000 mhz 7970s.

 

[Jaydip]

Sure explains why late model XFX Double Dissipation Black Edition 1000 mhz 7970s invariably fail at compute then, huh?

Same thing with late model Gigabyte Windforce 1000 mhz models.

Fails at compute? I am not sure I understand.Some overheating issue?

 

[24601]

Fails at compute? I am not sure I understand.Some overheating issue?

No, there's something about their bios design that detects compute workloads and automatically limits real work done somehow.

They probably didn't design the power delivery systems on them properly and implemented the bios hack to try to stop as many RMAs from people using it for compute.

It results in them having 1/2 to 1/3 compute performance.

If you do something to try to get around it, the power delivery system indeed results in a hard system lock without blue screen or anything.

This is at stock speeds with very good case ventilation.

 

[Makaveli]

No, there's something about their bios design that detects compute workloads and automatically limits real work done somehow.

They probably didn't design the power delivery systems on them properly and implemented the bios hack to try to stop as many RMAs from people using it for compute.

It results in them having 1/2 to 1/3 compute performance.

If you do something to try to get around it, the power delivery system indeed results in a hard system lock without blue screen or anything.

This is at stock speeds with very good case ventilation.

i've not heard of this before got anymore links or info?

 

[24601]

i've not heard of this before got anymore links or info?

Ok, what I think is happening is that the power delivery systems on these two cards (at least) is absolutely subpar, causing too much vdroop. The ROPs/IMC on Tahiti require higher voltage than the rest of the core, making this deficiency only show up on ROP intensive workloads.

The combination of a high actual core utilization of the chip plus the high utilization of the ROPs make the vdroop caused by the inferior power delivery system to cause a hardlock in the operating system.

This is exacerbated in the design of the XFX Double Dissipation Black Edition 1000 mhz especially since the VRMs are placed in such a way to have low actual airflow, leading to extremely high temperatures.

These high temperatures were tolerable in the reference board design, but the subsequent switch to XFX's custom board design lead to obvious deficiencies in many parts of the board.

This was obvious to these two board manufacturers, so they put something in their bios's that probably limits core speed drastically when they detect compute heavy workloads.

This explains the severe under-performing in Litecoin workloads at stock speeds.

When you force the cards to stay at the supposed "stock" speeds with MSI afterburner, they will work for a while until they hardlock the system.

This even carries onto lower clock speeds.

This is evident since the Litecoin workload actually does NOT bump into the powertune limits, because it does not fully utilize the compute section of the card. The extreme stress to the ROP/IMC section of the card is what causes the hardlock when combined with the vdroop from both inferior power delivery as well as the inferior cooling provided to the inferior power delivery.

The crappiness of these two cards for Litecoin workloads is common knowledge throughout the litecoin community.

 

[Makaveli]

Ok, what I think is happening is that the power delivery systems on these two cards (at least) is absolutely subpar, causing too much vdroop. The ROPs/IMC on Tahiti require higher voltage than the rest of the core, making this deficiency only show up on ROP intensive workloads.

The combination of a high actual core utilization of the chip plus the high utilization of the ROPs make the vdroop caused by the inferior power delivery system to cause a hardlock in the operating system.

This is exacerbated in the design of the XFX Double Dissipation Black Edition 1000 mhz especially since the VRMs are placed in such a way to have low actual airflow, leading to extremely high temperatures.

These high temperatures were tolerable in the reference board design, but the subsequent switch to XFX's custom board design lead to obvious deficiencies in many parts of the board.

This was obvious to these two board manufacturers, so they put something in their bios's that probably limits core speed drastically when they detect compute heavy workloads.

This explains the severe under-performing in Litecoin workloads at stock speeds.

When you force the cards to stay at the supposed "stock" speeds with MSI afterburner, they will work for a while until they hardlock the system.

This even carries onto lower clock speeds.

This is evident since the Litecoin workload actually does NOT bump into the powertune limits, because it does not fully utilize the compute section of the card. The extreme stress to the ROP/IMC section of the card is what causes the hardlock when combined with the vdroop from both inferior power delivery as well as the inferior cooling provided to the inferior power delivery.

The crappiness of these two cards for Litecoin workloads is common knowledge throughout the litecoin community.

So is using thie litecoin software the most common way to check for this deficiency?

 

[24601]

So is using thie litecoin software the most common way to check for this deficiency?

It is not notorious in any other community that I know of, probably because there is little to no other common application that I know of that uses high amount of compute as well as high to maximum ROP/IMC use on 7970 cards.

The same notoriety is not shared even in the Bitcoin community, most likely because of their utter lack of use of ROPs and IMC. (which is why Bitcoin ASICs are so easily produced)

 

[pcslookout]

Do you all think people who spread misinformation and trolls should be sued ?

It just happen with fake reviewers.

 

[24601]

Do you all think people who spread misinformation and trolls should be sued ?

It just happen with fake reviewers.

Considering marketing in and of itself is allowed, no amount of falsehoods are regulated in a realistic sense (enough to stop the intended goal).

 

[Attic]

Titan vs AMD's Next is 100x more exciting to me then the UFC 1XX fight night.

Let's get it on!!!!!!

 

[pcslookout]

Titan vs AMD's Next is 100x more exciting to me then the UFC 1XX fight night.

Let's get it on!!!!!!

I agree it is exciting but have no idea what UFC is. What is it?

 

[blackened23]

I agree it is exciting but have no idea what UFC is. What is it?

Rory Reed and JHH brawling MMA style in the ring. Any bets on who would win?

 

[pcslookout]

Rory Reed and JHH brawling MMA style in the ring. Any bets on who would win?

I have no idea who either of those people are. Never heard of them.

 

[Arzachel]

Ok, what I think is happening is that the power delivery systems on these two cards (at least) is absolutely subpar, causing too much vdroop. The ROPs/IMC on Tahiti require higher voltage than the rest of the core, making this deficiency only show up on ROP intensive workloads.

The combination of a high actual core utilization of the chip plus the high utilization of the ROPs make the vdroop caused by the inferior power delivery system to cause a hardlock in the operating system.

This is exacerbated in the design of the XFX Double Dissipation Black Edition 1000 mhz especially since the VRMs are placed in such a way to have low actual airflow, leading to extremely high temperatures.

These high temperatures were tolerable in the reference board design, but the subsequent switch to XFX's custom board design lead to obvious deficiencies in many parts of the board.

This was obvious to these two board manufacturers, so they put something in their bios's that probably limits core speed drastically when they detect compute heavy workloads.

This explains the severe under-performing in Litecoin workloads at stock speeds.

When you force the cards to stay at the supposed "stock" speeds with MSI afterburner, they will work for a while until they hardlock the system.

This even carries onto lower clock speeds.

This is evident since the Litecoin workload actually does NOT bump into the powertune limits, because it does not fully utilize the compute section of the card. The extreme stress to the ROP/IMC section of the card is what causes the hardlock when combined with the vdroop from both inferior power delivery as well as the inferior cooling provided to the inferior power delivery.

The crappiness of these two cards for Litecoin workloads is common knowledge throughout the litecoin community.

I don't doubt that some reference cards are lemons(the XFX5770 comes to mind), but you'll have to explain what you mean with "compute" and "compute heavy workloads", and how running a bunch of hashes causes "extreme stress on the ROP/IMC section of the card", because it sounds like you have absolutely no idea what you're talking about.

 

[Gloomy]

Do you all think people who spread misinformation and trolls should be sued ?

It just happen with fake reviewers.

The day lying on the internet becomes illegal I will take my leave of this wretched world.

 

[24601]

Rory Reed and JHH brawling MMA style in the ring. Any bets on who would win?

10:1 JHH, he looks pretty buff.

 

[24601]

I don't doubt that some reference cards are lemons(the XFX5770 comes to mind), but you'll have to explain what you mean with "compute" and "compute heavy workloads", and how running a bunch of hashes causes "extreme stress on the ROP/IMC section of the card", because it sounds like you have absolutely no idea what you're talking about.

Arguing from ignorance is fun i take it.

It's not my job to teach you how things work.

Use the magic of Google, its the "good enough" teacher of the modern student.

 

[The Alias]

Rory Reed and JHH brawling MMA style in the ring. Any bets on who would win?

jhh easy that guy looks built