• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

amd 7xxx series updates

Glad to hear the update. I've been worried the past couple of months having not heard much about the new cards.

Really looking to upgrade to a 7870 or 7970 variant.
 
when was the last time ati did a brand new architecture?
Every few years. :p

This is a departure from the R600 architecture introduced in 2007. That's mentioned in the first link near the beginning of the article.
 
You should read this.

Based on that and that I trust anandtech more on this type of things I don't think that southern islands with have the new architecture.
 
Looking forward to it. Now that I've recently moved up from 1680x1050 to 1920x1080 and dropped from the 40's to the 30's in Crysis and Metro, I'm looking for an upgrade but the only cards that are more than a few fps faster are the $500 and up lines. That's a little outta my reach.
 
Radeon 2900 I think, but their shader architecture has been within the same family since the Radeon 9000 series I think.
 
You should read this.

Based on that and that I trust anandtech more on this type of things I don't think that southern islands with have the new architecture.
That pretty much is the same info as other sites are reporting. What are you seeing there that makes you believe that is not describing the "HD 7000" series, which is AMD's next GPU architecture?
 
Woot, can't wait! I'm hoping to see at least 100% increase in performance with two 7xxx in crossfire versus 6xxx in crossfire. Do you guys think this is possible?
 
Woot, can't wait! I'm hoping to see at least 100% increase in performance with two 7xxx in crossfire versus 6xxx in crossfire. Do you guys think this is possible?

If it really is the new architecture... possibly.
 
This 4970x2 just isn't cutting it on my 30" monitor. I've been DYING to pick up a pair of 6970's so this news is very welcome! Hopefully we're not talking about December hehe
 
You should read this.

Based on that and that I trust anandtech more on this type of things I don't think that southern islands with have the new architecture.


parts of it will be. my guess is they will do somthing similar to the 6k series where mainstream and lower use the die shrink architecture of the 6900 VLIW4 and the 7900's will run a mixture of the newer architecture with the older. but all we can do is wait and see what happens.


http://forum.beyond3d.com/showpost.php?p=1561107&postcount=300

So yeah, 28nm AMD GPUs using FSA before the end of this year. Hopefully this means the AMD & [H] event in July is going to showcase the new cards. Pretty exciting.

very doubtful. i'm predicting they will have a few systems there running the bulldozer 8xxx cpu's and a few desktop versions of the Llano to show off their performance.
 
Bulldozer and Llano have been announced.
http://hardforum.com/showthread.php?t=1612784


yeah i know. but how a company perceives that and reality are different things, lol. technically 7k series and trinity have been announced as well, so theres not much AMD has that hasn't been announced. see how this works? we consider bulldozer as being announced, yet outside of leaked documents AMD has never announced the actual processors. so if they say had the FX 8150P there, then thats not an announced product because its only in leaked documents. same goes with any of the desktop versions of the Llano. everything we know about them is from leaked documents and not officially announced documents. now they may have ES models to show what the card may look like but i don't think they actually have a card that's playable yet. we probably won't see an ES working model until late Q3.
 
Last edited:
yeah i know. but how a company perceives that and reality are different things, lol. technically 7k series and trinity have been announced as well, so theres not much AMD has that hasn't been announced. see how this works? we consider bulldozer as being announced, yet outside of leaked documents AMD has never announced the actual processors. so if they say had the FX 8150P there, then thats not an announced product because its only in leaked documents. same goes with any of the desktop versions of the Llano. everything we know about them is from leaked documents and not officially announced documents. now they may have ES models to show what the card may look like but i don't think they actually have a card that's playable yet. we probably won't see an ES working model until late Q3.
Bulldozer and Llano have been announced. The rest of your reply is simply grasping at straws.

But that link states that southern islands will still be using VLIW4?
Read it again...
 
But that link states that southern islands will still be using VLIW4?
Old (or stubborn) speculation. The AMD slides are here: http://developer.amd.com/documentation/presentations/assets/6-Demers-FINAL.pdf ("next generation GPU/a new architecture" starts on slide 13)

Some context:
http://www.nordichardware.com/news/71-graphics/43460-amd-confirms-southern-islands-for-2011.html

2 or 3 days before the AMD Fusion 11 Development Summit Eric Demers keynote said:
"In a couple of days you are going to hear about our exciting new graphics architecture that will be coming out later this year and will be utilized by our future APUs," Rick Bergman, AMD.

I do see the conFusion :p here, and it's strange that the message is getting so muddled. There was no other new graphics architecture announcement made.
 
Woot, can't wait! I'm hoping to see at least 100% increase in performance with two 7xxx in crossfire versus 6xxx in crossfire. Do you guys think this is possible?

CF scaling is more of a bandwidth and latency issue, not something that will likely get significantly effected with a different GPU arch. They'd have to a new and faster bus among other things to really make a big difference. Make the CF link a full blown 8x PCIe 2.0 or better yet 3.0 bus would be one way to possibly do it I guess. Maybe a 16 bit HT 3.1 bus would better to replace the CF link, I don't really know for sure and I don't know why they haven't bothered, probably cost. Supposedly if you get the bandwidth high enough and the latency low enough you virtually eliminate all the software issues associated with CF as well as nearly linear scaling.
 
But will it have more cores than the 6XXX or the 5XXX series?
Probably not. Poor utilization in VLIW5/4 is why AMD is moving away from that architecture.

Better question: will it have higher performance at similar die size and power consumption? Clearly, those are tradeoffs Nvidia deals with, and lagged behind AMD GPUs in those two metrics for a while.

Best question: will it have higher performance? AMD seems to believe so, at least in GPGPU performance. ;)
 
Would be smart of AMD to get these new VideoCards out in October, just in time for the two biggest PC releases in years, BattleField 3 and SW;TOR, both games due out October.
 
Would be smart of AMD to get these new VideoCards out in October, just in time for the two biggest PC releases in years, BattleField 3 and SW;TOR, both games due out October.
Pretty Sure SW:TOR isn't coming out until 2012 now :rolleyes:
B3 will be good fun however :)
 
CF scaling is more of a bandwidth and latency issue, not something that will likely get significantly effected with a different GPU arch. They'd have to a new and faster bus among other things to really make a big difference. Make the CF link a full blown 8x PCIe 2.0 or better yet 3.0 bus would be one way to possibly do it I guess. Maybe a 16 bit HT 3.1 bus would better to replace the CF link, I don't really know for sure and I don't know why they haven't bothered, probably cost. Supposedly if you get the bandwidth high enough and the latency low enough you virtually eliminate all the software issues associated with CF as well as nearly linear scaling.

you know, if they use a HT link instead of their current CF Interconnect link, they could get closer to pxc's speculated Fusion MCM Opteron, lol :p
 
Pretty Sure SW:TOR isn't coming out until 2012 now :rolleyes:
B3 will be good fun however :)

I read all the SW;ToR forums, and MMO sites, and all the beta testers, and buzz surrounding the game points to a Fall release. And some big time beta tester, who is famous for leaks, actually gave a date today of October 4th, for whatever that's worth :D

http://betacake.blogspot.com/2011/06/star-wars-old-republic-exclusive-tester_20.html
Q. Based on what you've played so far, what is your prediction for the release date of the game?
A. October 4th.

Q. Please elaborate on why it is the "best game you've ever played."
A. Uh, a billion reasons really. I suppose I could write an essay, but the basic reasoning here is that I have a lot of games I can play and things I could be doing, but the only thing I want to do is play this game.
 
Last edited:
Probably not. Poor utilization in VLIW5/4 is why AMD is moving away from that architecture.

VLIW4 utilization wasn't poor at all for graphics workloads. This change is only about making it better for general computing.
 
VLIW4 utilization wasn't poor at all for graphics workloads. This change is only about making it better for general computing.
Poor was a bad word choice. More underutilized compared to Nvidia SP utilization would have been better.
 
Poor was a bad word choice. More underutilized compared to Nvidia SP utilization would have been better.

Poor, yet its still significantly more efficient? I know VLIW5 underutilized the 5th IW, which is why they went to VLIW4, but I've not heard anything about VLIW4 being underutilized. Do you have a source for that?
 
I can't find it now, but I've read the average packing rate in games was 3.4 (not sure if that was all shaders or just pixels). That's pretty good. It's more efficient in terms of die space than Nvidia.
 
Last edited:
Poor, yet its still significantly more efficient? I know VLIW5 underutilized the 5th IW, which is why they went to VLIW4, but I've not heard anything about VLIW4 being underutilized. Do you have a source for that?

depends. In graphics, it's great. In HPC, well... search around for Intel's recent larrabee anouncement :p
 
Yea for HPC/GPGPU their VLIW4 still isn't good enough. It depends too much on the compiler and software developers to write for the hardware to get OK much less good performance which is a kiss of death even in that field. For graphics its fine which is probably why its going to stick around for the low end and on die APU stuff since its more die space efficient than GCN. GCN is supposedly going to have a huge die size, much like Fermi did. If the B3D speculation thread is to be believed it may very well be significantly faster for the same or somewhat less die space and power usage then the updated version of Fermi it'll be going up against. While having simpler hardware (from a scheduling perspective) to boot, which is really quite impressive.
 
Last edited:
Supposedly if you get the bandwidth high enough and the latency low enough you virtually eliminate all the software issues associated with CF as well as nearly linear scaling.

They must be close already then, because from what I've seen, going from one card to two is basically a 100% performance increase (or darn close).
 
Back
Top