Sign in with
Sign up | Sign in
Your question

Chiphell.com has released RV770's every single detail!!!!

Last response: in Graphics & Displays
Share
March 3, 2008 7:32:34 AM

BREAKING NEWS! RV770 IS ON ITS WAY!!


1. The core speed of RV770 is higher than RV670(3870), however, it has not reached 1000MHz yet.
2. The memory that this baby gonna use will be GDDR5
3. Die size of RV770 is 250 mm^2
4. Both US Sharders and TMU units will be as twice as RV670's. Perherps even more.
5. It will lanuch in May.


Sounds really exicting, huh? :D 
March 3, 2008 7:51:28 AM

I am certainly looking forward to it. I just read bunch about the new innovations being used with it, and it does sound pretty exciting!
March 3, 2008 7:53:12 AM

The red team is fighting back.

I wonder what will be if they introduce a 4870x2?
Related resources
Can't find your answer ? Ask !
March 3, 2008 8:43:42 AM

VR-Zone never seems too trustworthy .. but I hope these specs are true.... as it'll push nVidia to bring out what the real power users desire :) 
March 3, 2008 4:33:08 PM

Twice the shaders and TMUs?

Blimey. Id settle for 50% more shaders and 100% more TMUs though. Thats where the r600 & rv670 are falling short ihmo.
March 3, 2008 4:44:22 PM

Great news if its true. If this comes in May then its realy great news for AMD. This would put a hudge crunch on Nvidia and create some super great deals for consumers for sure.
March 3, 2008 4:49:47 PM

Why are they conflicting?

The RV670 would compete where it currently competes, the mid range, except it would probably move lower, then the RV770 competes for the higher mid range or low hanging high end GTS-512 fruit, and then a potential R700 would compete with whatever other upper high-end is there.

I don't give creadance to rumours other than for discussion purposes, but the possibility of these products doesn't cause conflict anymore than the G92 eliminated the GF8800Ultra etc. There's room for all of them especially if the RV770 is anywhere near 50+% more powerful.

Personally I'll wait for the real deal, but it would be nice if this were true especially with regards to the TMUs (although I'd like to see improved address ratios too).

Either way, a long way off before anything solid, until then it's interesting to consider the implications of the hardware changes.
March 4, 2008 4:18:05 PM

GDDR5? I thought V4 was deemed as a moot difference from 3?
March 4, 2008 4:45:33 PM

Thats where GDDR5 makes the difference. Thats what its supposed to do, what the GDDR4 didnt
March 4, 2008 4:53:40 PM

xkm1948 said:
BREAKING NEWS! RV770 IS ON ITS WAY!!


1. The core speed of RV770 is higher than RV670(3870), however, it has not reached 1000MHz yet.
2. The memory that this baby gonna use will be GDDR5
3. Die size of RV770 is 250 mm^2
4. Both US Sharders and TMU units will be as twice as RV670's. Perherps even more.
5. It will lanuch in May.


Sounds really exicting, huh? :D 

Color me skeptical, but GDDR5? :lol: 
March 4, 2008 10:02:37 PM

Well GDDR4's main issue was the cost of getting the high end 1.8V modules, not the slower 1.5V modules.

With GDDR5 you get the benefit of much higher speeds, lower power consumption, and lower heat. They've been sampling it since before December with 1.8(3.6)Ghz modules running at 1.5V. Faster modules means less need for wire traces and transistors required for higher bitwidth support that would be required to equal that overall bandwidth, this reduces the cost of the chip and the cost of the PCB if you can support GDDR5.

Also they are increasing the transfer size and they are moving to larger single chips with 1Mbit at launch and 2Mbit already testing. That would mean less chips required for the same memory size.

Currently GDDR4 needs to be power hungry and fast for the high end, and the extension of GDDR3 meant the minor benefits GDDR4 were outweighed by it's shortage at the higher speeds. That's why it was cheap and plentiful at 1.5V on the lower end cards, but expensive and rare for the higher end cards wanting the 1.8V modules.

The other nice thing about GDDR5 is there isn't much additional tweaking to the hardware required to add GDDR5 support to something that already supports GDDR4 they designed it that way to ease adoption.

So for lower power, lower cost, lower heat, but faster speeds sounds good to me. If nVidia can get over their wire noise issues they'll be happy to skip past GDDR4 and go straight to GDDR5.
a b } Memory
March 5, 2008 12:38:21 AM

Woot! ATI (not AMD, I know....I know...I like ATI but hate AMD right now cause of the Phenoms, ATI on the other hand 3870X2s are good) is back in the game with this and the 780G.
March 5, 2008 1:00:02 AM

What's Nvidia got?
March 5, 2008 1:27:36 AM

After a year and a half, still waiting.....
March 5, 2008 2:14:00 AM

I think that the R700 is going to have a release date much later than May. I am thinking maybe July or August. And I just say that because of the news of AMD doing a revision of the HD38XX series. The current RV670 series is known as A11. The new revamped faster RV670 series will be known as A12 and will be released in April or May.

If the release date of the R700 was in May then why would AMD even bother to do a remake of the 38XX series with in that same time period?

http://www.fudzilla.com/index.php?option=com_content&ta...
Quote:
April / May for R12 revision

The new RV670 revision A12 will make enough space for a new version of 3870. The new Radeon 3870 with revision A12 chips will work at faster clocks and should score better than existing cards.

The other thing that we can report is that it is expected in April to May time frame, which clearly indicates that the R700 family will launch a bit later than that.

The story behind the RV670 chip is that the current A11 revision that was launch back in November was not supposed to be the launched, but it was the first A11 silicon that ATI managed to ship. Usually by ATI, A12 silicon is the launched one, but for a change something went well in DAAMIT.
March 5, 2008 3:01:55 AM

TheGreatGrapeApe said:
So for lower power, lower cost, lower heat, but faster speeds sounds good to me. If nVidia can get over their wire noise issues they'll be happy to skip past GDDR4 and go straight to GDDR5.

So that's it, wire noise issues. What exactly does that mean?
March 5, 2008 3:47:35 AM

rwayne said:

If the release date of the R700 was in May then why would AMD even bother to do a remake of the 38XX series with in that same time period?


I think this was covered a few posts back;
http://www.tomshardware.com/forum/249091-33-chiphell-re...

Remember they wouldn't fight for the same segment at least not at first until you had significant amounts to start crippling them. It also averages out the costs further for the HD3K series.

Launching a faster RV670 would still give people in the $125-200 segment something worth buying in both faster HD3850 and 3870 replacements, while letting the RV700 compete for the $200-400 segment and the R700XT compete for the $400+ segment, which at first would likely gobble up all production.

No reason to keep avoiding the high-end if you think you can bring something worthwhile to market.

Also the RV770 should have similar performance but lower cost compared to an HD3870X2. So it makes alot of sense to go that route.
March 5, 2008 4:10:27 AM

homerdog said:
So that's it, wire noise issues. What exactly does that mean?


With the way that nVidia uses a traditional memory bus interface it focuses all the wire traces into a very tight space where they contact the chip on the pcb, and this proximity of ever smaller wires going to ever smaller memory interfaces increases wire cross-talk. Due to the GDDR4's requirement for low signal noise (to allow for those higher speeds) it means your tolerances are very tight. ATi's ringbus helps combat this by spreading out the interface around the chip and at multiple points.

So far the major reason I've heard for no GDDR4 in the nV cards has been the signalling issues, and it does make sense. Especially since the GDRR4 production/cost drawbacks wouldn't have been apparent during the design development stage. It makes it a wise hindisght decision, but wouldnt have made it a good decision to limit options, unless there was a reason.
March 5, 2008 8:29:41 AM

rwayne said:
I think that the R700 is going to have a release date much later than May. I am thinking maybe July or August. And I just say that because of the news of AMD doing a revision of the HD38XX series. The current RV670 series is known as A11. The new revamped faster RV670 series will be known as A12 and will be released in April or May.

If the release date of the R700 was in May then why would AMD even bother to do a remake of the 38XX series with in that same time period?

http://www.fudzilla.com/index.php?option=com_content&ta...
Quote:
April / May for R12 revision

The new RV670 revision A12 will make enough space for a new version of 3870. The new Radeon 3870 with revision A12 chips will work at faster clocks and should score better than existing cards.

The other thing that we can report is that it is expected in April to May time frame, which clearly indicates that the R700 family will launch a bit later than that.

The story behind the RV670 chip is that the current A11 revision that was launch back in November was not supposed to be the launched, but it was the first A11 silicon that ATI managed to ship. Usually by ATI, A12 silicon is the launched one, but for a change something went well in DAAMIT.


It could be around august I reckon, but ATI will want to get into the market with their new products before Nvidia does, and generally having played second fiddle for just over a year they will be trying to get to release as fast as possible imho. I dont think we'll see the launch date blow right out to september/october time.
March 5, 2008 8:30:42 AM

TheGreatGrapeApe said:
With the way that nVidia uses a traditional memory bus interface it focuses all the wire traces into a very tight space where they contact the chip on the pcb, and this proximity of ever smaller wires going to ever smaller memory interfaces increases wire cross-talk. Due to the GDDR4's requirement for low signal noise (to allow for those higher speeds) it means your tolerances are very tight. ATi's ringbus helps combat this by spreading out the interface around the chip and at multiple points.

So far the major reason I've heard for no GDDR4 in the nV cards has been the signalling issues, and it does make sense. Especially since the GDRR4 production/cost drawbacks wouldn't have been apparent during the design development stage. It makes it a wise hindisght decision, but wouldnt have made it a good decision to limit options, unless there was a reason.


cheers for that, found that interesting.
March 5, 2008 8:31:39 AM

no... that's too late... I think it's more AMD who is behind schedule... ATI seem to be doing their thing now, like back in the good old days :) .

any word on if these cards can Xfire with the current (and A12) gpus? i think not...
March 5, 2008 1:40:05 PM

TheGreatGrapeApe said:
With the way that nVidia uses a traditional memory bus interface it focuses all the wire traces into a very tight space where they contact the chip on the pcb, and this proximity of ever smaller wires going to ever smaller memory interfaces increases wire cross-talk. Due to the GDDR4's requirement for low signal noise (to allow for those higher speeds) it means your tolerances are very tight. ATi's ringbus helps combat this by spreading out the interface around the chip and at multiple points.

So far the major reason I've heard for no GDDR4 in the nV cards has been the signalling issues, and it does make sense. Especially since the GDRR4 production/cost drawbacks wouldn't have been apparent during the design development stage. It makes it a wise hindisght decision, but wouldnt have made it a good decision to limit options, unless there was a reason.

Cool. Thanks.
March 5, 2008 2:41:40 PM

From everything I've read, the R700 will have a core 50% faster than the 3870 core. With multi-die, this board will be faster than any future nvidia 9000 series.
So the point is, while the 2900XT had great specs (lots of shaders, ringbus, etc), it had some serious bottlenecks and even the 3870 had texture unit problems.
This time though, I think all bottlenecks will be un-corked and problems will be ironed out since the 4870 will be the 3rd-gen 10.1 card. I think Ati was wise to focus on the next gen cards and Ati will finally SMOKE!
The good-old ATI is back.
Using this logic: The phenom while ahead of its time, will also finally get ironed out and will finally get at least faster than the kentfield.
2008 may be a very rosy year for AMD!
March 5, 2008 2:44:44 PM

enewmen that doctor who avatar is rather unsettling i find lol.
March 5, 2008 3:42:56 PM

The BEST doctor of them all!!!
March 5, 2008 9:32:57 PM

TheGreatGrapeApe said:
Launching a faster RV670 would still give people in the $125-200 segment something worth buying in both faster HD3850 and 3870 replacements, while letting the RV700 compete for the $200-400 segment and the R700XT compete for the $400+ segment, which at first would likely gobble up all production.

No reason to keep avoiding the high-end if you think you can bring something worthwhile to market.

Also the RV770 should have similar performance but lower cost compared to an HD3870X2. So it makes alot of sense to go that route.
So do you really think the R700 is still on schedule for May or June? (I Hope So)

That would be amazing if some of the HD38XX 512MB cards did come in at $125 around May.

What do you think about the A12 revision? Do you think those cards will still be labeled as HD3870s and 3850s? Or do you think ATI will name them something else like HD4670s and 4650s? and reserve the HD4870 title for the R700?

Renaming the ATI A12 cards to the HD46XX series would make sense and would follow in line with Nvidia making the performance jump of the 8600GT to the 9600GT and making it their NEW standard for a midrange card.

I bet that is what is going to happen. That would explain how A12 cards and the R700 cards would be released all with in the same time frame of May. It would be the hard launch of the ATI Radeon 4XXX series.
March 5, 2008 9:49:27 PM

JAYDEEJOHN said:
The BEST doctor of them all!!!

Thanks :D  And I agree.

But what do you think of my post? Right, wrong, too early to know anything, or I have no idea what I'm talking about??



March 6, 2008 1:24:30 AM

Its what Ive read, and expect as well, tho the PCI-2 will add an interesting twist in things because of the bus speeds
March 6, 2008 2:03:22 AM

JAYDEEJOHN said:
Its what Ive read, and expect as well, tho the PCI-2 will add an interesting twist in things because of the bus speeds

:hello:  Ha, I just replied to you in another thread. You do know that PCIe 2.0 makes virtually no difference for the today's video cards? And yes I include the 9800GTX in the "today's video cards" category.
a b } Memory
March 6, 2008 8:38:48 PM

enewmen said:
From everything I've read, the R700 will have a core 50% faster than the 3870 core. With multi-die, this board will be faster than any future nvidia 9000 series.
So the point is, while the 2900XT had great specs (lots of shaders, ringbus, etc), it had some serious bottlenecks and even the 3870 had texture unit problems.
This time though, I think all bottlenecks will be un-corked and problems will be ironed out since the 4870 will be the 3rd-gen 10.1 card. I think Ati was wise to focus on the next gen cards and Ati will finally SMOKE!
The good-old ATI is back.
Using this logic: The phenom while ahead of its time, will also finally get ironed out and will finally get at least faster than the kentfield.
2008 may be a very rosy year for AMD!


I agree with everything you have said about ATI and I am glad that it seems they are just pushing AMD to the side and doing their own thing like they should always have done even from the begining.

As for Phenom being faster than Kentsfeild.... More than likely not. Right now on a clock per clock overall Phenom has to be at 2.7GHz to be on par with a Q6600. No B3 might help a little but I doubt much. Maybe their 45nm will help but we cannot tell. I don't think Phenom will really be able to beat Kentsfeild maybe catch up. And even if it does catch up it now has to deal with Yorkfeild. And at the end of this year Nehalem. So its got its work cut out for it.
March 6, 2008 10:04:22 PM

rwayne said:
So do you really think the R700 is still on schedule for May or June? (I Hope So)


Unfrtunately I don't know. It could be and the rumblings are positive, but like so many things last minute issues like plagued the GF9600 and GF9800 series could crop up (the R520's memory issue didn't crop up until late into production and delayed them 6-9 months to the point where they almost were able to skip to the R580 altogether).

What I'm saying is that the launch of a refreshed RV670 doesn't preclude them also launching the RV770, it just means that the end target would be different than it currently is.

Quote:
What do you think about the A12 revision? Do you think those cards will still be labeled as HD3870s and 3850s? Or do you think ATI will name them something else like HD4670s and 4650s? and reserve the HD4870 title for the R700?


I would prefere they label them the HD3880-3890 and the HD3860 instead of doing something like naming them the HD4xxx. If they did that it would kind of negate all their talk about what this new numbering scheme was supposed to mean. Where 3 is supposed to denote the generation. If it's "launch generation" instead of chip/feature generation, then that would require a new one every frickin' refresh then.
Of course don't put it past them to forget that point to and do the PR-thing as usual. IMO, the only thing to justify a new number would be new features according to their own strategy. However marketing is a very powerful motivator.

Quote:
Renaming the ATI A12 cards to the HD46XX series would make sense and would follow in line with Nvidia making the performance jump of the 8600GT to the 9600GT and making it their NEW standard for a midrange card.


From that perspective it makes sense, but from their own internal perspective it doesn't.
However I agree, most people will look at the cards the same way they look at the GF9600 and think that since the AMD number didn't change it must obviously be less 'modern' because it still uses the same numbers. But those would be like the tools who bought FX5200s to replace their GF4ti cards, and those who bought X1300s to replace their X800s.

Quote:
I bet that is what is going to happen. That would explain how A12 cards and the R700 cards would be released all with in the same time frame of May. It would be the hard launch of the ATI Radeon 4XXX series.

It does make sense from a marketing perspective and I agree it's likely to happen, it just negates any silly notion of this whole numbers versus suffixes thing they went through last year.
[/quotemsg]
March 9, 2008 11:15:03 AM

There are some details of the new HD 4000 Series.

(HD 4870 X2, HD 4870, HD 4850, HD 4670, HD 4650, HD 4470, HD 4450)

The HD 4870 X2 will be four times faster than the HD 3870 watching the FLOPS.

http://mygpu.info/newshd4000.htm


xkm1948 said:
BREAKING NEWS! RV770 IS ON ITS WAY!!


1. The core speed of RV770 is higher than RV670(3870), however, it has not reached 1000MHz yet.
2. The memory that this baby gonna use will be GDDR5
3. Die size of RV770 is 250 mm^2
4. Both US Sharders and TMU units will be as twice as RV670's. Perherps even more.
5. It will lanuch in May.


Sounds really exicting, huh? :D 

March 9, 2008 9:44:52 PM

amd_fanboi said:


any word on if these cards can Xfire with the current (and A12) gpus? i think not...



Yeah that is a filthy rumor that is snowballing throughout the forums. So far, xfirex has been only announced for RV670 and RV680 chips. Given the trend, it looks like it is between the same GPUs and wont translate into RV670 / RV770 xfire-ing.

However, that is just my assumption, but... I think it is better to assume no xfirex than something that hasn't been announced yet.

March 9, 2008 10:03:27 PM

What exactly is the A12 revision going to change? I was under the impression that GPU revisions were similar to CPU steppings, and don't really bring anything new to the table in terms of added functionality. Just lower heat/power consumption and higher clocks right?
March 10, 2008 2:47:20 PM

higher clocks = better performance, which means, for HD3870, better performance against the 8800gt and 9600GT
March 10, 2008 3:56:59 PM

amd_fanboi said:
higher clocks = better performance

Thanks for clearing that up :ange: 
a b } Memory
March 11, 2008 3:42:45 PM

waassy said:
There are some details of the new HD 4000 Series.

(HD 4870 X2, HD 4870, HD 4850, HD 4670, HD 4650, HD 4470, HD 4450)

The HD 4870 X2 will be four times faster than the HD 3870 watching the FLOPS.

http://mygpu.info/newshd4000.htm


1050MHz GPU???? Nice. And 2200MHz memory for the HD4870 is nice. I can get 850MHz GPU and 2200MHz memory with my HD2900Pro 1GB GDDR4. Looks nice. If the price is right I might get one :D 
March 11, 2008 5:23:23 PM

Ive heard using GDDR5 itll be more like 3600MHz
March 11, 2008 6:24:22 PM

Nobody get your hopes up. Think 15% increase in performance and July release, that way you won't be too disappointed.
March 11, 2008 6:37:28 PM

Hell, might as well say, itll be worse than the 2900 then anything will look good
March 11, 2008 8:24:50 PM

higher clocks does NOT = better performance. It's more calculations per cycle. There's only so far you can push higher clockspeeds. This is why a 2ghz core2 beats the snot out a 3ghz p4.
March 11, 2008 8:25:08 PM

jaydeejohn- HUH? Are you joking?

computertech82- That's pretty much the truth, if you look at DDR3 over DDR2, or a 3Ghz Nutburst CPU compared to a 1.83Ghz C2D. Clocks don't mean as much as the tecnnology that's put into them.
March 11, 2008 8:31:28 PM

I think as any new unreleased tech comes to be, we like to speculate. Who knows with a completely different arch? I like to speculate too, its just sooo little to go on... now if it were the 9800GTX? heheh
March 11, 2008 9:03:14 PM

T8RR8R said:

computertech82- That's pretty much the truth, if you look at DDR3 over DDR2, or a 3Ghz Nutburst CPU compared to a 1.83Ghz C2D.


Well the first one makes little sense as the move to significantly faster speeds does matter despite any additional latency concerns short term, that's like comparing the end of DDR to the beginning of DDR2, now would you still want ultra low timing DDR 400 compared to your DDR2 PC2-10600 for that C2D?

And comptech's example using faster older technology as an example makes little sense, because this is progressive architecture of similar characteristics, improved if anything. So faster and improved, hm let's see if it could possibly have better performance. :sarcastic: 
Guess what's faster a 2.4 ghz E4600 or 3.0ghz QX9650?
I'll put my money on the one with a faster core, faster FSB, more cores, more cache.

If the architecture is the same with more units it's more than reasonable to think faster speed = faster performance.
March 11, 2008 9:04:58 PM

computertech82 said:
higher clocks does NOT = better performance. It's more calculations per cycle. There's only so far you can push higher clockspeeds. This is why a 2ghz core2 beats the snot out a 3ghz p4.

If the architecture is unchanged then higher clocks do equal better performance. Of course other potential bottlenecks come into play once the clocks get too high, but we shouldn't have to worry about that for a while.

What we could be seeing with the RV770 is an improved (but not entirely ground-up new) architecture coupled with higher clocks. So ATI will finally have a GPU (not card, they arguably already have that) faster than G80 :lol: 
!