January 9 Launch Date for AMD Radeon HD 7900

[DannibusX]

House or video cards?

 

[bostonbuddy]

but will they be able to handle consul ports?

 

[John Doe]

My internal ATI contact now working for AMD has already confirmed to me that there’s a strong possibility of a HD 7900 series launch for mid to late January 2012. In other words, they’ve been baking in the oven for several months now and AMD has enough to meet any possible demand. I’ve gotten this info on September 22, 2011 but was told not to say anything until after Christmas. AMD does not want to affect HD 6000 series Christmas sales by announcing the HD 7000 too soon. So much for the Top Secret

Yeah, and Boy George told me he's in love with you the other day. I've had enough of your AMD nonsense. How many more posts, threads you have to post to praise AMD? Enough is enough.

 

[7mm]

Just about time I've being told to upgrade from Radeon HD 4350 and I made up my mind with Radeon HD 6670 DDR5, now shouldn't I wait for a month for better upgrade!

 

[newtekie1]

This is wrong. The highest OC's on 6970 cards have been seen on reference solutions. Yes, Lightnings OC well though with reference you're guaranteed to get quality. Non-ref, even a good solution can be a crapshoot and not OC due lack of QC. I.E. Gigabyte's older Ultra Durable cards.

Really? Care to prove that with some examples of reference cards overclocked on the stock air cooling?

 

[John Doe]

Really? Care to prove that with some examples of reference cards overclocked on the stock air cooling?

http://www.techpowerup.com/reviews/AMD/HD_6950_1_GB/25.html

http://www.techpowerup.com/reviews/Powercolor/HD_6950_Vortex_II/30.html

If you need to ask, I wouldn't expect you to understand. With reference, you most the time get a good sample. Reference boards are all made by Sapphire (PCPartner), while non-ref's are unknown. That's why they tend to have better QC and mess up less.

 

[pantherx12]

http://www.techpowerup.com/reviews/AMD/HD_6950_1_GB/25.html

http://www.techpowerup.com/reviews/Powercolor/HD_6950_Vortex_II/30.html

If you need to ask, I wouldn't expect you to understand. With reference, you most the time get a good sample. Reference boards are all made by Sapphire (PCPartner), while non-ref's are unknown. That's why they tend to have better QC and mess up less.

That's a 2.15% difference which is probably more down to individual chip variation rather than anything else.

For example my first 6870 could over-clock to 1100mhz and do 1000mhz with no additional voltage.

My second 6870 can't go over 960mhz without jibbing out.

My cards were identical in every other way ( reference design by the way with the fancy voltage control and such )

 

[John Doe]

That's a 2.15% difference which is probably more down to individual chip variation rather than anything else.

For example my first 6870 could over-clock to 1100mhz and do 1000mhz with no additional voltage.

My second 6870 can't go over 960mhz without jibbing out.

My cards were identical in every other way ( reference design by the way with the fancy voltage control and such )

That's not the point. With reference, you have a higher chance of getting a good sample.

VisionTek 900352 Radeon HD 6950 2GB 256-bit GDDR5 ...

 

[pantherx12]

That's not the point. With reference, you have a higher chance of getting a good sample.

VisionTek 900352 Radeon HD 6950 2GB 256-bit GDDR5 ...

I don't think there is a higher chance man, maybe you've just got lucky.

I see the same variation regardless of whether it's reference or not.

In fact some of the highest over clocks I've seen are on non reference cards.

How ever I know it's all down to luck, any over-clocking always is.


I don't just speak from my experience either, I speak from this forums entirely collective user-base XD

 

[John Doe]

I don't think there is a higher chance man, maybe you've just got lucky.

I see the same variation regardless of whether it's reference or not.

In fact some of the highest over clocks I've seen are on non reference cards.

How ever I know it's all down to luck, any over-clocking always is.

That's not what I meant. Companies go cost cutting when they go non-ref. So if you just buy some random non-ref board without research (like that one), chances are you'll get a cheaply made board with QC all over the place. With reference, you always get that solid, tested and proven design.

 

[pantherx12]

That's not what I meant. Companies go cost cutting when they go non-ref. So if you just buy some random non-ref board without research (like that one), chances are you'll get a cheaply made board with QC all over the place. With reference, you always get that solid, tested and proven design.

This is why I read reviews before buying things

See now I get what your saying, but the way you were saying it was well not right


Should of gone with that statement straight away, save yourself all the grief you got

 

[newtekie1]

http://www.techpowerup.com/reviews/AMD/HD_6950_1_GB/25.html

http://www.techpowerup.com/reviews/Powercolor/HD_6950_Vortex_II/30.html

If you need to ask, I wouldn't expect you to understand. With reference, you most the time get a good sample. Reference boards are all made by Sapphire (PCPartner), while non-ref's are unknown. That's why they tend to have better QC and mess up less.

That's a 2.15% difference which is probably more down to individual chip variation rather than anything else.

For example my first 6870 could over-clock to 1100mhz and do 1000mhz with no additional voltage.

My second 6870 can't go over 960mhz without jibbing out.

My cards were identical in every other way ( reference design by the way with the fancy voltage control and such )

Actually, the difference is probably due to the fact that the reference board is pushing 1.19v through the core by default and the PowerColor card is only using 1.15v by default. But he probably was hoping we didn't notice that.

And when the PowerColor card was ran at ~1.2v(or 1.19v maybe, the graph is a little hard to read exact voltages) it hits 965MHz, so 35MHz more than the reference with the Volterra when using the same voltages.

Not a great example... Keep trying there John Doe.

 

[pantherx12]

Actually, the difference is probably due to the fact that the reference board is pushing 1.19v through the core by default and the PowerColor card is only using 1.15v by default. But he probably was hoping we didn't notice that.

And when the PowerColor card was ran at ~1.2v(or 1.19v maybe, the graph is a little hard to read exact voltages) it hits 965MHz, so 35MHz more than the reference with the Volterra when using the same voltages.

Not a great example... Keep trying there John Doe.

Completely missed that, but then I've always ignored that graph

 

[John Doe]

Not a great example... Keep trying there John Doe.

Great, you know actually it's like discussing with a brick wall when arguing with you. You do realize reference is the best with the Cayman GPU, right? It has the highest QC and the best power delivery. As long as temps are kept in check, reference with Volterra is a beast. Most non-ref solutions of Cayman are cheaper to make, inferior. Especially after the revision change, 2nd revision boards haven't been doing as good as the first 6950/70's did.

 

[newtekie1]

Completely missed that, but then I've always ignored that graph

Oh, one more thing. Did you happen to wonder why he picked the 1GB review instead of the reference 2GB? My guess is because the 2GB reference sample only managed 860Mhz...

Yep, those reference cards will definitely give you a great overclock. Guaranteed!</sarcasm>

Great, you know actually it's like discussing with a brick wall when arguing with you. You do realize reference is the best with the Cayman GPU, right? It has the highest QC and the best power delivery. As long as temps are kept in check, reference with Volterra is a beast. Most non-ref solutions of Cayman are cheaper to make, inferior. Especially after the revision change, 2nd revision boards haven't been doing as good as the first 6950/70's did.

You keep saying that, but have yet to show any proof. And the proof you tried to show just made you look worse. As I said, keep trying.

Yeah, I can understand the brick wall thing, you just keep posting things and I just keep shooting them down for the BS they are... I can understand that becoming frustrating.

 

[John Doe]

Oh, one more thing. Did you happen to wonder why he picked the 1GB review instead of the reference 2GB? My guess is because the 2GB reference sample only managed 860Mhz...

Yep, those reference cards will definitely give you a great overclock. Guaranteed!</sarcasm>

Yes, they will. I didn't give attention to those details at all. Good making up.

You keep saying that, but have yet to show any proof. And the proof you tried to show just made you look worse. As I said, keep trying.

The burden of proof is on you. It's your ignorance that doesn't know about how the revision change affected the boards. Do your research, ask Lord Jummy. He has reference boards. I don't own the cards in question so this is all I can say. What proof do you want me to show you? I had a bunch of non-ref boards with that red PCB made by Flextronics (same company), they didn't OC crap. Not even a few Mhz. Such thing would pretty much never happen with a reference board, because it's higher quality make with higher QC.

 

[newtekie1]

Yes, they will. I didn't give attention to those details at all. Good making up.

And your reason for the 2GB sample clock like crap then? If the reference boards are so much better, why only 860Mhz. Note: I know the answer, I'm just trying to figure out if mister know-it-all can figure it out.

The burden of proof is on you. It's your ignorance that doesn't know about how the revision change affected the boards. Do your research, ask Lord Jummy. He has reference boards. I don't own the cards in question so this is all I can say. What proof do you want me to show you? I had a bunch of non-ref boards with that red PCB made by Flextronics (same company), they didn't OC crap. Not even a few Mhz. Such thing would pretty much never happen with a reference board, because it's higher quality make with higher QC.

Not it isn't. You made the initial claims, you have to prove them. You have to prove that the Volterra PWM is important. All you've done so far is make claims, and the small little amount of proof you tried to provide actually proved the opposite of your claim, that when a non-reference card without a Volterra is set to the same voltage it will clock higher than the reference with the Volterra.

 

[John Doe]

Not it isn't. You made the initial claims, you have to prove them. You have to prove that the Volterra PWM is important. All you've done so far is make claims, and the small little amount of proof you tried to provide actually proved the opposite of your claim, that when a non-reference card without a Volterra is set to the same voltage it will clock higher than the reference with the Volterra.

I didn't say it's important. They took off what made the GPU fancy. It's now just an inferior GPU with inferior drivers. It has nothing left to make itself really distinguish from nVidia. Understand? And no, a non-ref card will NOT clock higher than a reference with these cards. It depends on your sample; the GPU you got.

 

[DannibusX]

Oh, hey look. 7900's are releasing on Jan. 9!

C-c-c-c-combo Breaker!

 

[newtekie1]

I didn't say it's important. They took off what made the GPU fancy. It's now just an inferior GPU with inferior drivers. It has nothing left to make itself really distinguish from nVidia. Understand? And no, a non-ref card will NOT clock higher than a reference with these cards. It depends on your sample; the GPU you got.

What you posted seems to contradict that statement, doesn't it. Seems the Volterra isn't important at all, you basically only got one thing right. It depends on the sample, the Volterra doesn't improve your changes at all.

 

[pantherx12]

What proof do you want me to show you? I had a bunch of non-ref boards with that red PCB made by Flextronics (same company), they didn't OC crap. Not even a few Mhz. Such thing would pretty much never happen with a reference board, because it's higher quality make with higher QC.

Cmon man you buy a cheap board what do you expect?

It's like buying a £35 motherboard and expecting it to be as good as a £200 one.

If you get the non reference cards that are the same price or more expensive than a reference one you'll typically get decent over-clocking headroom unless your unlucky.

Anyway, think that's me done in this thread.

Way off topic


Shall we make a new thread to discuss reference vs 3rd party designs?

 

[John Doe]

What you posted seems to contradict that statement, doesn't it. Seems the Volterra isn't important at all, you basically only got one thing right. It depends on the sample, the Volterra doesn't improve your changes at all.

Yes, it does. It provides accurate software voltage monitoring. Right at point and very stable regulation. It shows it's benefits under more extreme cases, so as to what degree it helps depends on your case. There haven't been extensive testing on it to show how much it adds to the reliability/stability of GPU. Though, more advanced VRM designs get lower ripple, which equates to cleaner power.

 

[cadaveca]

Yes, it does. It provides accurate software voltage monitoring.

guess which one is reference:

 

[JustaTinkerer]

Well I was wanting to show you guys later but by a freak miracle mix up at the delivery centre I managed to get my hands on a 7900 series card.

First overclocking result are promising but I wish W1zz would fix GPU-z

http://www.techpowerup.com/forums/attachment.php?attachmentid=44657&stc=1&d=1323750651


Wait until I get it under water though

 

[John Doe]

guess which one is reference:
http://www.techpowerup.com/forums/attachment.php?attachmentid=44656&stc=1&d=1323750640

You know you have to keep track of voltage for an amount of time, look at how much regulation changed percantage-wise, right? I can't "guess" which one is reference by a single screenshot. I've to extensively test and keep track of both.

Can you guess how long an obstacle is by looking at it? You have to measure it. This is ironic.