January 9 Launch Date for AMD Radeon HD 7900

[n-ster]

Not that many people OC GPUs compared to CPUs at least. I OC my GPU only slightly

 

[John Doe]

You, my dear, are an idiot.

If you took few minutes to check out their rumor posts from the past you would surely notice that they have been correct on pretty much everything. The only exception to this are the inferences people make from those announcements. People like you.

No, I'm not. However you are if you trust that site.

Their most recent rumor was Bulldozer. They claimed to have tested the chip, and put it in pace with a 980x. When in reality, most people knew it was going to blow. Then it did.

I've used Volterra PWM-powered reference AMD boards, and cheapo Sapphire/HIS boards that use the same GPU but cheaper UPI/CHIL/OnSemi + DPAK circuits. I could never tell any quality difference. I guess the PWM-powered boards only have an edge with extreme cooling...that is if AMD doesn't skew you over with clock speed limits.

There's no accurate software monitoring or voltage adjustment on cheapo cards. And either way, I'd prefer to have solid Volterra's powering my card. For example, you can power your system with both a $100 or a $150 PSU. Both would do it well as long as they're under ATX specs. Though, I'd take the $150 one.

 

[btarunr]

There's no accurate software monitoring or voltage adjustment on cheapo cards.

And you wouldn't need it either, unless you have extreme cooling at your disposal. That was my point. Most cheap UPI/OnSemi/CHIL controllers still give you 'some' voltage tuning headroom.

 

[John Doe]

And you wouldn't need it either, unless you have extreme cooling at your disposal. That was my point. Most cheap UPI/OnSemi/CHIL controllers still give you 'some' voltage tuning headroom.

Well, I personally want to have it all in one. You don't get accurate software voltage sensors with those. Then again I'd like to make sure my card is powered cleanly and well. Not with some cheap regulators that can't suppress ripple as good.

 

[btarunr]

Well, I personally want to have it all in one. You don't get accurate software voltage sensors with those. Then again I'd like to make sure my card is powered cleanly and well. Not with some cheap regulators that can't suppress ripple as good.

Besides, Volterra+CPL boards were, historically, always produced in limited quantities, so those cost-effective board designs that allow greater price flexibility step in just a couple of months later. Those high-grade boards were more of PR vehicles. Only the early adopters got the pie.

Looks like a 'posh' reference board with Volterra+CPL PWM at first glance, doesn't it?

Look closely:

^That's a common cost-effective VRM.

The Volterra+CPL design should look like this from the back of the PCB:

Again, my point is, the whole discussion about missing digital PWM on those Tahiti boards is moot. Very few actually end up getting those cards, AMD itself gives out cheapo PCBs once the launched products "settle down", as is the case with the example above. You'll find the same cheap PCB in use with many other AIBs.

 

[John Doe]

Yeah. Ripple is a part killer. People that ran their machines off old units that're on the higher side of it need good VRM. Good VRM = better ripple surpression. You either upgrade your PSU or VRM to reduce it, or else it might be an another point of failure. Both PSU and cards die from ripple.

 

[erocker]

Looks like the card pictured has more phases than the GTX 570 which I believe used a 4 phase design. Really though, I don't care as long as the performance is there out of the box. Has anyone measured the hole spacing for the cooler to see if it matches up with any previous cards?

 

[Damn_Smooth]

I am looking forward to seeing how these perform. They are part of my current upgrade plan, but they can be taken off if we get anything resembling the other half of AMD's recent product launches. If these launch dates are true, this will give them almost a full year before Nvidia has anything to answer to it. Hopefully they can still pull off a decent GPU.

 

[v12dock]

W1zzard is currently viewing this thread and lol'ing at our speculations

 

[John Doe]

W1zzard is currently viewing this thread and lol'ing at our speculations

lol I doubt it. Has the card even went to full production yet?

 

[DarkOCean]

Not sure where the Ti or the 6870 came in discussion. lol. I was on about the high end (7900/6900).

Ti=texas instruments

 

[John Doe]

Ti=texas instruments

Ok, you said 4 phase. The amount of phases don't equal to the amount of power they deliver. Put five 50W phases against ten 25W phases. See? The thing here is, this design isn't as quality as the high current inductors MSI uses, let alone a Volterra. See the pics I compared to on the bottom of the 2nd page.

 

[dj-electric]

lol I doubt it. Has the card even went to full production yet?

1 - yes, probably. It doesnt take 1 night to produce thousands of graphics cards from nada
2 - Since when reviewers need a shelf prodect, they get an ES
3 - Yes, W1zz probably look at this thread with a trollface smile

 

[H82LUZ73]

Besides, Volterra+CPL boards were, historically, always produced in limited quantities, so those cost-effective board designs that allow greater price flexibility step in just a couple of months later. Those high-grade boards were more of PR vehicles. Only the early adopters got the pie.

Looks like a 'posh' reference board with Volterra+CPL PWM at first glance, doesn't it?
http://img.techpowerup.org/111212/bta8746kjhdc.jpg

Look closely:
http://img.techpowerup.org/111212/bta8764jhd.jpg
^That's a common cost-effective VRM.

The Volterra+CPL design should look like this from the back of the PCB:
http://img.techpowerup.org/111212/bta8764hd.jpg

Again, my point is, the whole discussion about missing digital PWM on those Tahiti boards is moot. Very few actually end up getting those cards, AMD itself gives out cheapo PCBs once the launched products "settle down", as is the case with the example above. You'll find the same cheap PCB in use with many other AIBs.

Is Volterra+CPL PWM made in Taiwan? If it is it could have been flooded like the Wd factory or they come from the northern part of Japan were the Nuclear plant has been leaking water out....either way... Like Erocker and Bta have said... If the performance is there
yep you bet I will be getting one,When was the last card from ATI/AMD with memory bust over 256bit....2900xt 512bit with the robust fire bit right.348 bit is new for them And i hope they take both crowns from the Green camp.

 

[John Doe]

1 - yes, probably. It doesnt take 1 night to produce thousands of graphics cards from nada
2 - Since when reviewers need a shelf prodect, they get an ES
3 - Yes, W1zz probably look at this thread with a trollface smile

And can you tell me how you know about that? There's zero info out there stating such thing. The card may as well not even be designed properly yet. It's not that easy to build a GPU in months time.

 

[TheoneandonlyMrK]

his point is that it IS that easy to slap a gpu on a board and ES gpu are most deff tapped out

 

[John Doe]

his point is that it IS that easy to slap a gpu on a board and ES gpu are most deff tapped out

No, it's not. You're making it sound like as easy as cooking beef. You don't just "slap a GPU on a board". ES samples are fully functional GPU's, just like how ATi/AMD has been doing it before. I have an ES HD 2900 dated at March of 07. It's the same as any other 2900 technically. You have to design an entire GPU architecture, which we still aren't sure if they've done yet.

 

[erocker]

You have to design an entire GPU, which we still aren't sure if they've done it yet.

What are those pictured cards then? .. if that's what you are meaning.

 

[John Doe]

What are those pictured cards then? .. if that's what you are meaning.

It's a half-cut picture. The core of the GPU looks some sort of toy. I'm not even sure the card in that single photo is finished yet.

 

[cadaveca]

It's a half-cut picture. The core of the GPU looks some sort of toy. I'm not even sure that single shot is finished yet.

No offense, dude, but just because you are skeptical doesn't mean everyone else has to be as well.

 

[Wile E]

Analog vs digital pwm is 100% meaningless. All that matters is what the regulatory circuits are capable of, not how they get there.

 

[erocker]

Meh, nothing means a thing until the company that produces these cards actually provide some information. Arguing about speculation, rumors, etc. has to be one of the biggest wastes of time I can think of. I'm just going to say these cards exist and I've seen enough proof to think this way.

 

[newtekie1]

Sooo... Unveil means they will just remove the NDA or at least some of it, but doesn't mean the cards will actually be available on the market, correct?

Looks like the card pictured has more phases than the GTX 570 which I believe used a 4 phase design. Really though, I don't care as long as the performance is there out of the box. Has anyone measured the hole spacing for the cooler to see if it matches up with any previous cards?

Well accoring to him the VRM design on the GTX570 is rubbish, can't handle any overvolting(pure BS), and the same 4 phase design on the GTX470 was better and the GTX470 overclocked better too...

There is no use arguing with him, he acts like he knows what he is talking about, but in the end has no clue.

 

[TheoneandonlyMrK]

Meh, nothing means a thing until the company that produces these cards actually provide some information. Arguing about speculation, rumors, etc. has to be one of the biggest wastes of time I can think of. I'm just going to say these cards exist and I've seen enough proof to think this way.

and to get giddy with anticipation no less

vregs dont overly matter to 99% of their users so long as they work

 

[John Doe]

Analog vs digital pwm is 100% meaningless. All that matters is what the regulatory circuits are capable of, not how they get there.

Digital solutions (Volterra) are most the time superior to Analog, but this specific Analog design (used on non-ref Cayman GPU's) is nowhere as quality as a Volterra Digital PWM.