ATI Radeon HD 4000 Series discussion

[AliG]

@ Hornet:That means after gt200, but that still says absolutely nothing about when the specific launch date is, and besides, December of 2011 is after August, so once again, we still know just about nothing

[BenchZowner]

It is possiblefor the g92 gts to run there stock (I haven't seen any non modded gt though, don't know where you've seen that)

I'm not talking about stock Vgpu.
I believe you also know that 8800GT has lower stock Vgpu than 8800GTS512 and 9800GTX.
Nothing also dictates that nVIDIA will be using the same stock voltage they're using now with the 8800GT ( which is rather low ).

but the thing is, g92's yields were bad enough on 65nm, I don't see how they'll be that much better on 55nm that they can run stock 800mhz (I predict probably around 700mhz for the 9900gt and perhaps 750, top of 775 for the 9900gtx). Also why would g92b get gddr5? You said it yourself, the reason why nvidia didn't bother with gddr5 for gt200 (aside from more ROPs, but they could have changed their design to work around that) is because ATI would get just about all the GDDR5 in the channel.

ATi has a GDDR4 channel domination ( and not only channel ).
I said that nVIDIA will never use GDDR4, never said they won't be using GDDR5.

The yields weren't that bad, and there was also another reason to "stop" the G92 flow when they introduced the 8800GT 512MB

800MHz Core is very possible, and unless they keep the voltages real low to maybe use single slot cooling, it'll be it.
And there'll be more headroom as well, and they'd never push the chips close to their limits, not even the pre-overclocked vendors editions.

The 1GB RAM already upped the cost, using GDDR5 would get it even higher, so they decided to implement a wider BUS width and GDDR3. ( "GT200" )

When's the launch date? Is it before or after gt200?

Like I said, after "GT200" and more than likely in August.

[DilTech]

http://www.hardware.info/en-US/news/...launch_HD4800/

The final card in the HD4800 line-up will be the Radeon HD4870 X2, which will have 512MB GDDR5 memory per core, busting earlier myths on shared memory.

Looks like no shared memory controller, it's still going to be 256bit x2.

[Shintai]

http://www.hardware.info/en-US/news/...launch_HD4800/



Looks like no shared memory controller, it's still going to be 256bit x2.

Surprise surprise. Just as predictable as Amen in christian church.

[Hornet331]

good, finally this rumors is dead (for sure now).

[Shintai]

good, finally this rumors is dead (for sure now).

Unless the fanboys starts about talking about R800 series. Then we get it all back again and then some. Only to end up like this again.

[Origin_Unknown]

if the x2 has 512mb per core then it looks like ill be going for a 4870 1gb * 2

[Milos-stancene]

Why don t you just stop with it and wait until the REAL bencmarks comes???

[STaRGaZeR]

NV and ATI can burn in hell with their GX2 and X2 crap.

[Hornet331]

NV and ATI can burn in hell with their GX2 and X2 crap.

indeed, or at least as long as till they fix the microstuttering.

[STaRGaZeR]

indeed, or at least as long as till they fix the microstuttering.

Micro stuttering is only one of their problems unfortunately

[Tim]

Micro stuttering is only one of their problems unfortunately

GX2 is a great card pal, nothing like the 7950, got good drivers, and excellent performance.

[STaRGaZeR]

GX2 is a great card pal, nothing like the 7950, got good drivers, and excellent performance.

Of course it is a great card and it has excellent perfomance... when it works properly. It's the same with CF or SLI. The drivers you call good are good only for certain games, if your usual games are in that list, GX2 or X2 will kill any other card. But if they're not... and I don't know why the games I usually play have close to 0 (if not 0) support in your "good drivers", so I call them crap drivers and crap card. I feel kinda insulted when my $600 card does weird things or doesn't work at all in those games I want to play, and you?

[biohead]

indeed, or at least as long as till they fix the microstuttering.

doesn't that only occur with unmatched pci-e bandwidths? (e.g. p35's 16/4)

[Hornet331]

doesn't that only occur with unmatched pci-e bandwidths? (e.g. p35's 16/4)

it occures with every multi-gpu solution (NV/ati/s3) on every board, right now its a limitation of the system itself, but it might be fixable in the future.

[Leeghoofd]

it occures with every multi-gpu solution (NV/ati/s3) on every board, right now its a limitation of the system itself, but it might be fixable in the future.

Think for sure SLI, which has been running for such a long time already, a real anti-micro stuttering solution would have been found already...

Let's advance time and whished it was mid june already....

[LordEC911]

http://www.hardware.info/en-US/news/...launch_HD4800/



Looks like no shared memory controller, it's still going to be 256bit x2.

Doesn't anyone find it a little strange that their "sources" are another news site which is not reporting the same news?
There is also some other interesting BS in there...

[Macadamia]

no way the 4850 is up against the gt260 with the current rumored specs. Imho the direct competition of the 4580 will be the G92b

2 4850s do.

[zerazax]

Doesn't anyone find it a little strange that their "sources" are another news site which is not reporting the same news?
There is also some other interesting BS in there...

Yep looks like BS to me. They cited NH but didn't see them quote that there or here!

[freeloader]

Until you hear it from the horse's mouth, it's all horse crap! A few more weeks and the world will know.

[STaRGaZeR]

Yep looks like BS to me. They cited NH but didn't see them quote that there or here!

Looking at the leaked pictures of the 4870X2 cooler the quick conclussion is that it's not BS...

[zerazax]

Looking at the leaked pictures of the 4870X2 cooler the quick conclussion is that it's not BS...

I know the cooler pic, I mean the article source and conclusion. At least if they cited the cooler picture, it would be better journalism than claiming someone else claimed something, when they didn't

[LordEC911]

Looking at the leaked pictures of the 4870X2 cooler the quick conclussion is that it's not BS...

Except how do you know that it is a bridge chip?
Also they are using a slightly tweaked version of the R680 cooler.
Plus the fact that AMD/ATi themselves said that R700 wasn't simply two RV770's on the same PCB like R680...

[zerazax]

Well I've always said there had to be some bridge for the two to talk to each other since putting the two die on one chip just wasn't going to happen, no this soon at least.

Also, the memory config doesn't mean they are independent for each core or if they are shared.

Anyways, the fact that ATI has for R600 and for RV670 kept the internal ring bus, and I highly doubt ATI is leaving it around for the sake of wasting resources (in which case if they are, people need to be fired) - i expect it to be put to use one way or another.

Anyways, with all the hush hush about R700 (that "leaked" AIB notice, whether fake or not, didn't even mention R700 specs), concluding one or another way decisively on what little evidence we have is pointless. We'll know soon enough.

[adamsleath]

Of course it is a great card and it has excellent perfomance... when it works properly. It's the same with CF or SLI. The drivers you call good are good only for certain games, if your usual games are in that list, GX2 or X2 will kill any other card. But if they're not... and I don't know why the games I usually play have close to 0 (if not 0) support in your "good drivers", so I call them crap drivers and crap card. I feel kinda insulted when my $600 card does weird things or doesn't work at all in those games I want to play, and you?

you mean advertise the problems? never