AMD talks up Griffin
Forum CPU & Components : CPUs - AMD talks up Griffin
Today AMD has released info on it's new Mobile rocessor called Grifin. Griffin, contrary to what many people believed IS NOT based on K10 but on K8. It has several advances that should mae idle power nearly negligible with additional sleep states and the ability to throttle each core independently.
This announcement also makes fools of all those wo doubted the addition of the 4 cycles of latency they SAID was to allow for larger caches at 65nm.
As we all know Brisbane only has 512KB L2 while Griffin has 1MB. I guess they are going to increase the cache after all. Yet another time when Intel fangirls should have listened.
Links everywhere, The Register, Anand, THG, etc.
They also reportedly tweaked the IMC for independent operation and better prefetching. It looks like it is following AMDs usual power before perf mantra but the enancements talked about may provide 10-20% improvement over Turion.
If this is a "system on a chip" and even if this is not a "star" core, I won't mind having a Athlon X2 in a UMPC!
Baron, you forgot the most interesting part Mid-2008.
| Quote : If this is a "system on a chip" and even if this is not a "star" core, I won't mind having a Athlon X2 in a UMPC! |
It may be my next laptop. I have a 1.6GHz Turion now, but I can't upgrade the CPU so I need to buy a new one. This gives me a reason to skip the vanilla 65nm Turion.
| Quote : Baron, you forgot the most interesting part Mid-2008. |
That is a ways out but I think it's better since they have 65nm Turions shipping now and need to concentrate most of their efforts on K10 and its chipsets.
Baron, I doubt Intel fangirls will be be impressed with the Griffin and the Puma. You forgot that at same time Intel will release Nehaleem, which will have be more integrated platform and its performance numbers would be at least +50% as same clocked Turion.
Once again, your pro AMD news failed to make AMD look better than Intel.
| Quote : This announcement also makes fools of all those wo doubted the addition of the 4 cycles of latency they SAID was to allow for larger caches at 65nm.
|
This sounds most like flame bait. I don't see it making fools of anyone the way it is written.
90nm Toledo and Windsor are capable of handling 1MB L2. A direct shrink into 65nm should likewise handle the same amount of cache with no additional latency, at the same process quality.
But from your description, Griffin sounds like K8 computational cores hooked up to K10 IMC and power circuitry. That power circuitry could be the reason for increasing L2 latency - it's a by-product of being able to clock the separate caches asynchronously.
| Quote : Baron, I doubt Intel fangirls will be be impressed with the Griffin and the Puma. You forgot that at same time Intel will release Nehaleem, which will have be more integrated platform and its performance numbers would be at least +50% as same clocked Turion.
|
Agreed.
| Quote : Baron, I doubt Intel fangirls will be be impressed with the Griffin and the Puma. You forgot that at same time Intel will release Nehaleem, which will have be more integrated platform and its performance numbers would be at least +50% as same clocked Turion.
|
Why would I be trying to make one company look better than the other? Im not you. Besides, if these chips can make a battery last for 7 hours (they are claiming 5 for the vanilla 65nm chips), that will get more business than if it can do SuperPi in 10 sec.
| Quote : This announcement also makes fools of all those wo doubted the addition of the 4 cycles of latency they SAID was to allow for larger caches at 65nm.
|
This sounds most like flame bait. I don't see it making fools of anyone the way it is written.
90nm Toledo and Windsor are capable of handling 1MB L2. A direct shrink into 65nm should likewise handle the same amount of cache with no additional latency, at the same process quality.
But from your description, Griffin sounds like K8 computational cores hooked up to K10 IMC and power circuitry. That power circuitry could be the reason for increasing L2 latency - it's a by-product of being able to clock the separate caches asynchronously.
The Brisbane was originally designed for 512KB, but they wanted the option of increasing the cache. Obviously for this very purpose. It's not my fault that everyone here negatively viewed the latency change without considering the reasons.
They even said why and still foolishness prevailed. Now they have announced the increase in cache and it's flame bait?
Fine. It is.
Cache is not the reason for flaming. It is your tone and the message delivered, curiously deviod of any balance. There is no news out of AMD that means anything right now, because they over promise and under deliver. Until there is a product releasedand and it benches worth a darn, this entire topic is moot.
The fact that this product is not due until mid 2008 errrr.......... 2009, errrr.......... 2010 or whenever they get around to it.
gOJDO, why don't we wait until there is more information about the chips before we try to make performance speculations? To do otherwise is pretty much just trolling. Intel hasn't said much about 45 nm laptop chips nor let anybody do a direct clock-for-clock comparison between Wolfdale/Yorkfield and Conroe/Kentsfield and AMD hasn't said any performance numbers about Griffin yet, so it's all just conjecture at this point.
| Quote : Baron, I doubt Intel fangirls will be be impressed with the Griffin and the Puma. You forgot that at same time Intel will release Nehaleem, which will have be more integrated platform and its performance numbers would be at least +50% as same clocked Turion.
|
Why would I be trying to make one company look better than the other? Im not you. Besides, if these chips can make a battery last for 7 hours (they are claiming 5 for the vanilla 65nm chips), that will get more business than if it can do SuperPi in 10 sec.
Baron
This does come off as a bit 9-inch-ish.
Dont get riled, unlike him you dont post and run, and you do stay around to defend your perpective. But neglecting to include the fact that release is set for mid 2008 is a glaring omision, and the "many people" and "Intel fangirls" comments lend a decidedly skewed flavor.
The 2008 bit is huge though. They are still working to introduce new K8 products 2 years after C2D. Not particularly heartening news. I hope they plan on supplying them at low costs.
| Quote : Baron, you forgot the most interesting part Mid-2008. |
Why are they going to launch a 65nm K8 derivative when supposedly K10 is fine and 45nm should launch around then?
It is a ways off, which brings up another question to me. In these recent mobile announcements, there is no mention of "hawk". This was supposed to be Q2, but then slipped to Q3/Q4 like everything else of AMDs. Had support for Hybrid HDD and had the on-the-fly switch from discrete to integrated video, i believe. I assume DX10 video also, but not sure. Maybe not as impressive as Puma, but as you say, Puma's a ways off.
Is the Puma fanfare to make us forget about a hopelessly delayed Hawk, or is Hawk just not a big enough deal to hype?
Of concern to those like myself who will get a new laptop in the fall...
| Quote : As we all know Brisbane only has 512KB L2 while Griffin has 1MB. I guess they are going to increase the cache after all. Yet another time when Intel fangirls should have listened. |
There were K8 before with less latency and 1MB L2 per core, so I don't know where the Intel fangirls are wrong.
Well, cmon now Vern, ya have to have a little patience. It takes a lot longer to achieve relability when your busy remaining virginally pure of heart and all that other jazz.
Besides, didnt you see the interview....AMD sells "freedom", (
) not perfromance. At least not when Intel holds the performance crown that is.
| Quote : Cache is not the reason for flaming. It is your tone and the message delivered, curiously deviod of any balance. There is no news out of AMD that means anything right now, because they over promise and under deliver. Until there is a product releasedand and it benches worth a darn, this entire topic is moot.
|
News is news. Flame away. I have flame retardant skin. Look at how long Intel has been promising CSI and dual core Itanium. Is that different? How?
| Quote : As we all know Brisbane only has 512KB L2 while Griffin has 1MB. I guess they are going to increase the cache after all. Yet another time when Intel fangirls should have listened. |
There were K8 before with less latency and 1MB L2 per core, so I don't know where the Intel fangirls are wrong.
Maybe it's because the forest is in the way.
| Quote : It is a ways off, which brings up another question to me. In these recent mobile announcements, there is no mention of "hawk". This was supposed to be Q2, but then slipped to Q3/Q4 like everything else of AMDs. Had support for Hybrid HDD and had the on-the-fly switch from discrete to integrated video, i believe. I assume DX10 video also, but not sure. Maybe not as impressive as Puma, but as you say, Puma's a ways off.
|
Hawk is just a 65nm shrink for Turion and they say that it's shippng to OEMs now. with system availability by the end of the month.
| Quote : If you take what henri said this time into the context of providing an escape from Intels megalomaniacal world domination plans, this could be seen as an adequate statement.
|
I can only wonder why you still have that avatar and location.
| Quote : Yet another time when Intel fangirls should have listened.
|
It may be my next laptop. I have a 1.6GHz Turion now, but I can't upgrade the CPU so I need to buy a new one. This gives me a reason to skip the vanilla 65nm Turion.
Baron, is this imagined futuristic vision of a laptop going to compliment your power horse brute of a PC... you know the one we have envied since inception from your vivid and detailed tales of lore told about your uber efficient Geode 4x4... fabricated by AMD [/inconjunction with SuperSwamper] ... it's a mate made in heaven, Like Little Big-Man... or the Engine that Could...
| Quote :
|
After reading through your posts today, I think I will keep my avatar as is for just a bit longer as well.
Wow
Baron, you have been getting no love what so ever accross several threads the past few days. Feels like your a target.
| Quote : As we all know Brisbane only has 512KB L2 while Griffin has 1MB. I guess they are going to increase the cache after all. Yet another time when Intel fangirls should have listened. |
There were K8 before with less latency and 1MB L2 per core, so I don't know where the Intel fangirls are wrong.
Maybe it's because the forest is in the way.
No, it is because you understand nothing, nothing at all ..... and swallow any excuse AMD provides you as the gospel truth.... the latency is process driven and any excuse otherwise is pure FUD.
BTW, I hear that the dual socket enthusaist part won't support > 4 gigs...
I don't CARE. I guess that would be a good excuse as the 65nm process is redone to be mobile specific. Or vice versa.
FUD is not propping up your products it's devaluing other's products.
| Quote : FUD is not propping up your products it's devaluing other's products. |
That sounds a lot like AMD's CTO interview on Tom's.
| Quote : Detailing a product that won't launch for at least a year, talking up the future without having seen any of Intel's products that will be out when it launches .... hmmmmmmm, where I have heard this before:
|
-- Henri Richard DigiTimes Inteview March 2006 before the Conroe Benchmarks
It would appear that AMD is engaging in "traditional market tatics"
Nice find. Exquisitely apropos
| Quote : Having read through most of the thread now, I think it is time for lockage... |
Flexing the old gigolo muscle are we?
This just means that AMD can´t get their new Architecture to the mobile segment in time. Instead they use some patched together K8. Maybe there really is something to the rumors about a delayed K10.
I vote a title change to "Baron Talks Up AMD....again...sigh"
| Quote : Cache is not the reason for flaming. It is your tone and the message delivered, curiously deviod of any balance. There is no news out of AMD that means anything right now, because they over promise and under deliver. Until there is a product releasedand and it benches worth a darn, this entire topic is moot.
|
News is news. Flame away. I have flame retardant skin. Look at how long Intel has been promising CSI and dual core Itanium. Is that different? How?I'm absolutely ecstatic that my E6600 doesn't have an IMC. They used the die space for 4MB Cache... and the large L2 combined with off-die MC...i can run DDR-400, and my machine still positively rocks. I don't lose much performance at all...using slower RAM. My RAM isn't even running @ 200MHz(~185MHz@2-2-2-5-1T). Try that on an AM2 K8.
So 'griffin' is the only transitional mobile chip before fusion....
griffin link
i really wanted a k10 mobile core before fusion(see signature below).... How can I go from a core2duo T5600 to a 'griffin'/'puma'? This may end up being a downgrade.... And considering 'griffin' is like 6/12 months away, it REALLY will need a k10 core to be a viable purchase....
purhaps I should email AMD, and express my concerns....
| Quote : Flexing the old gigolo muscle are we? |
The Tsunami thread was getting outta hand and Baron was only going to dig himself deeper.
I hope Wombat2 picks up on it though, there are some good Baron quotables in that one....
Posts like this show that you are just a trouble maker.
| Quote : Cache is not the reason for flaming. It is your tone and the message delivered, curiously deviod of any balance. There is no news out of AMD that means anything right now, because they over promise and under deliver. Until there is a product releasedand and it benches worth a darn, this entire topic is moot.
|
News is news. Flame away. I have flame retardant skin. Look at how long Intel has been promising CSI and dual core Itanium. Is that different? How?I'm absolutely ecstatic that my E6600 doesn't have an IMC. They used the die space for 4MB Cache... and the large L2 combined with off-die MC...i can run DDR-400, and my machine still positively rocks. I don't lose much performance at all...using slower RAM. My RAM isn't even running @ 200MHz(~185MHz@2-2-2-5-1T). Try that on an AM2 K8.
But it's OK for you to make excuses for Intel's lateness? See, you're a dink. I think that Griffin will make a nice splash.
WHat mobo has DDR for C2D?
Some Asus dual-VSTA motherboard combines C2D with AGP and DDR as well as PCI-e and DDR2... it's somewhat popular among the budget (non-overclocking) upgraders. That's an advantage of off-die memory controllers - you can design a new one for custom needs and combine it with an existing mass-production CPU.
Could we not argue about lateness, especially with server concerns beyond the practical reach of this forum (Itanium, interconnect bandwidth). One company has reason to be late if the other company doesn't provide competition.
| Quote : Some Asus dual-VSTA motherboard combines C2D with AGP and DDR as well as PCI-e and DDR2... it's somewhat popular among the budget (non-overclocking) upgraders. That's an advantage of off-die memory controllers - you can design a new one for custom needs and combine it with an existing mass-production CPU.
|
OK. Perhaps that's why AM3 will supposedly support both. It would have been great had AMD put a dual type controller in AM2, but again, Intel is 10X bigger so AMD may not have had the resources at the time. There is the big joke that Intel has more profit than AMD has revenue every year.
I guess like VernDewd I wish the two companies were closer in revenues, but Intel has been making sure that that doesn't happen even though with the recent numbers for Vista people make note that there are 2X as many PCs being sold now as when XP came out.
So a change in market share balance wouldn't have affected Intel's growth.
Anyway, I'm ranting. I do believe though that the changes for Griffin will be a vast improvement over Turion and with the 65nm chips trickling out at up to 2.4GHz, they will have a good stopgap.
I have a 1.6GHz Turion and it flies even with a 5400RPM HDD, so I can see myself with a TL66 before I upgrade my desktop. Unfortunately, the pinouts will be different so I can't upgrade but could always have two laptops next year.
ALL HAIL THE DUOPOLY!!!
| Quote : Some Asus dual-VSTA motherboard combines C2D with AGP and DDR as well as PCI-e and DDR2... it's somewhat popular among the budget (non-overclocking) upgraders. That's an advantage of off-die memory controllers - you can design a new one for custom needs and combine it with an existing mass-production CPU.
|
OK. Perhaps that's why AM3 will supposedly support both. It would have been great had AMD put a dual type controller in AM2, but again, Intel is 10X bigger so AMD may not have had the resources at the time. There is the big joke that Intel has more profit than AMD has revenue every year.
I guess like VernDewd I wish the two companies were closer in revenues, but Intel has been making sure that that doesn't happen even though with the recent numbers for Vista people make note that there are 2X as many PCs being sold now as when XP came out.
So a change in market share balance wouldn't have affected Intel's growth.
Anyway, I'm ranting. I do believe though that the changes for Griffin will be a vast improvement over Turion and with the 65nm chips trickling out at up to 2.4GHz, they will have a good stopgap.
I have a 1.6GHz Turion and it flies even with a 5400RPM HDD, so I can see myself with a TL66 before I upgrade my desktop. Unfortunately, the pinouts will be different so I can't upgrade but could always have two laptops next year.
ALL HAIL THE DUOPOLY!!!Yes...i have the ASRock 775Dual-VSTA. I'm not making excuses for their being slow to adapt CSI, etc. i'm just saying that for me, and many others....the lack of IMC has turned into a benefit, not a hinderance. I would still be using my P4 if i had to pair DDR2 with C2D, as money's very tight for me.
| Quote : Today AMD has released info on it's new Mobile rocessor called Grifin. Griffin, contrary to what many people believed IS NOT based on K10 but on K8. It has several advances that should mae idle power nearly negligible with additional sleep states and the ability to throttle each core independently.
|
why shuld anyone bother about it? by that time, with k10 n penryn on the market it will make this "new processor" looks old. and i doubt anyone will buy notebooks or even upmc with this slow k8 core.
| Quote : Today AMD has released info on it's new Mobile rocessor called Grifin. Griffin, contrary to what many people believed IS NOT based on K10 but on K8. It has several advances that should mae idle power nearly negligible with additional sleep states and the ability to throttle each core independently.
|
why shuld anyone bother about it? by that time, with k10 n penryn on the market it will make this "new processor" looks old. and i doubt anyone will buy notebooks or even upmc with this slow k8 core.
K8 is not a slow core. If the improvements can decrease power by 30% and perf by even 20%, it will make a kick ass notebook for even gaming.
The first SLI notebook I saw ADVERTISED had a Turion and 7800Go. Refitting K8 should save on masks and enable them to get the core up and running quicker than a totally new design.
| Quote :
|
I would not expect a K8-based architecture can achieve 30% power reduction and 20% performance increase without heavily revamping the core.
| Quote : WHat mobo has DDR for C2D? |
I have one sitting right here that I ran my E6400 on for a good while and it works fine - ECS board...Asrock has one and there are a few others...best solution in the world? No. Does it work and work well, both perf wise and stability wise? You bet...
| Quote :
|
I would not expect a K8-based architecture can achieve 30% power reduction and 20% performance increase without heavily revamping the core.
If you read the Toms' article, you will see that they redesigned the North bridge and cores for split plane voltage. Since a majority of the time the CPU won't need to run both cores, they could turn off one core and lower the power of the other core which with the idle efficiency of the current K8 (again see the tom's article).
By allowing the CPU to lower the voltage of RAM and turning down the HT3 links, that will enable excellent office perf with the lowest amount of power.
Redesigning everything for power means they can up the frequency from the current 2.3GHz (TL66). My TL-52 1.6GHz is very snappy for even Visual Studio, so if they get Griffin to 2.7GHz it will be more than competitive.
Griffin
Power
| Quote : Detailing a product that won't launch for at least a year, talking up the future without having seen any of Intel's products that will be out when it launches .... hmmmmmmm, where I have heard this before:
|
-- Henri Richard DigiTimes Inteview March 2006 before the Conroe Benchmarks
It would appear that AMD is engaging in "traditional marketing tatics"
There is a better quote from the same article:
| Quote : But frankly, they've made so many claims in the past – you know, the Netburst architecture was supposed to scale to 10GHz, and look at where we are today. Then their new-generation micro-architecture (NGMA), is, quite frankly, a quick fix on the front-side bus. I don't think that's the future of the Intel architecture. I think it's another quick fix until 2008 or later, when they're going to come out with a genuinely new architecture. So again, from a pure technology perspective, my assessment is that it's a lot of marketing – it's clever marketing, but it's not revolutionary. And calling it a new-generation microarchitecture is a little bit out of balance. |
Henri Richard is full of BS! He only makes me to trust AMD less.
And don't forget the forum (I've forgotten
) where he got owned by a newb (not noob) with 5 post....It's ok to be a fanboy IMO, but with a brain to match...
My first thought when I read this was how stupid can AMD get..... sure it would be great if they could release this now, but they seriously want to pit a refitted K8 core against Nehalem when its already at a substantial disadvantage to Conroe? When I've thought things through though, it could be there is a method to AMD's madness....
What if we're going to go back to the days when we had 2 classes of mobile chips from AMD - remember when there was Turion CPUs alongside Mobile Athlon 64's? I wonder if this is what they're planning again. Think about the way the market is going. The average joe is probably more likely to buy a laptop now than a desktop, and they also are probably looking for value as opposed to absolute performance. In addition, they are probably more concerned about having a tangible feature such as 20% more battery life than 20% more points in PCMark.
That being the case, perhaps the point of Griffin is that by using the old K8 core which dosn't have all the core enhancements of K10 (and therefore uses fewer transistors), the power consumption will be lower and also because the dies will be smaller they will be cheaper to produce. The idea would be that Griffin would provide acceptable performance with optimum value and battery life. This approach would strengthen AMD's product branding - Turion would become known to your average user as giving good value for money, whilst Phenom would also propogate into (power hungry) mobile parts and establish itself as AMDs flagship chip for performance on both desktops and laptops.
What are your thoughts on this Jack?
There are 9 identified and unidentified users. To see the list of identified users, Click here.
Please mind
You are about to answer a thread that has been inactive for more than 6 months.
If you still wish to proceed, please ensure that your posting is original and does not duplicate or overlap any prior responses to this thread.
