Caporegime
It wants to be a bit higher than that or its, Fail!
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Launches Three Kaveri APU SKUs in February 2014 – Feature Set For A10 and A8 APUs Detailed
- Thread starter Boomstick777
- Start date
It wants to be a bit higher than that or its, Fail!
Caporegime
I somewhat disagree, while I'd like it to be higher, I can't see it.
An 8 core high clocking CPU Lynnfield IPC will still be a very good CPU.
If it was any higher (Just slightly slower than Sandy) but 8 cores, you'd be pretty much guaranteed that'd it be a ridiculous price, as it's pretty much going toe to toe with a 3930K.
We'll have to wait and see anyways.
Caporegime
Needing to run insane clocks out of the box to get "some" single threaded performance out of it is a big problem factor with temps and power consumption.
Lynnfield is still a country mile behind Hswell, it will have to run at silly clocks again to keep up.
Caporegime
Lynnfield's are the original i5's, IPC wise it's not bad at all, bearing in mind with instruction sets it's somewhat old now too.
An 8 core CPU with IPC like that overclocking to 4.8GHZ would be insane tbh.
It'd be over double the performance of a first gen i5 maxed out (Probably like 120% of the performance going 100%, and then you're not factoring in gains from instruction sets)
IPC's moved on maybe ~25% from Lynnfield (But then with FPU it's a fair bit more, Cinebench shows quite some gains, 4.8GHZ core > 5.2GHZ Ivy core)
An 8 core CPU with IPC like that overclocking to 4.8GHZ would be insane tbh.
It'd be over double the performance of a first gen i5 maxed out (Probably like 120% of the performance going 100%, and then you're not factoring in gains from instruction sets)
IPC's moved on maybe ~25% from Lynnfield (But then with FPU it's a fair bit more, Cinebench shows quite some gains, 4.8GHZ core > 5.2GHZ Ivy core)
Last edited:
Soldato
Just for you!
Lynnfield IPC + decent iGPU + several potentially great technologies rumoured and I'm a happy bunny
Tbh, IPC is much more important for their desktop/FX cpu's than it is for their APU's and the two don't really target the same market.
Though, I can see a lot of people picking up a Kaveri APU if the performance is good for a media pc etc, or as a cheap and rather good way of entry into the market of gaming pc's.
Though, I can see a lot of people picking up a Kaveri APU if the performance is good for a media pc etc, or as a cheap and rather good way of entry into the market of gaming pc's.
Soldato
It seems Kaveri will be made at TSMC:
http://translate.google.com/translate?hl=en&sl=auto&tl=en&u=http://www.expreview.com/29187.html
http://translate.google.com/translate?hl=en&sl=auto&tl=en&u=http://www.expreview.com/29187.html
Caporegime
Good! Most of AMD's CPU delay problems over the past 10 years can be traced back to Global Foundries issues and delays, i have been saying for years they need to dump them.
Caporegime
Caporegime
But Kaveri is still delayed if it comes out in 2014
Paper release before this year is out, AMD didn't say it would be available this year
Caporegime
Are we allowed to make jokes about it being the same as AM3+ now 
?
As you can get a few select FM2+ boards
?As you can get a few select FM2+ boards
Caporegime
Are we allowed to make jokes about it being the same as AM3+ now
As you can get a few select FM2+ boards
Yes, alight....
Caporegime
I've got some decent hopes for FM2+ to be honest, it all depends just how much they can get out of that IGP.
Caporegime
I'm more interested in CPU IPC, having said that if AMD can find a way to make hybrid CF work and work well for any GCN GPU the iGPU could offer a nice boost to discrete GPU's.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Firstly Global Foundries was only formed in 2009, so blaming them for products late before then is laughable daft.
Likewise Intel has had frequent delays, as have TSMC, as have Samsung, IBM and each and every foundry around because you're talking about research that now can stretch to 100's of billions amongst all the fabs and various unused paths of research for each process node increase.
Intel has 5 fabs, running 4 at max capacity while switching one over and ramping up volume while working through problems is the single easiest and best way to do it. AMD had one fab, and no capital(largely because Intel was out their paying everyone to not stock AMD products) to expand it's production.
Try upgrading an in use fab when you're at max capacity, can't afford downtime and need as high yields as possible as fast as possible to make it worthwhile shutting parts of the fab down.
It's inherently difficult to upgrade a fab in use. These days TSMC generally starts a new fab or a new expansion on launch of a new process which then gets used to ramp up the next process. IE when 45nm was actually the 28nm fab was well into production and as that was finished they had a relatively easy way to begin to test and ramp production.
AMD and then GloFo simply haven't had that luxury, it's a logistical nightmare and yet they've been catching up pretty quick, but catching up in foundry terms is something you see over a couple of processes, not a few months.
AMD's biggest and smartest move was getting planning permission for the New York state fabs, in 2009 when GloFo was founded, production on that fab hadn't started, 3-4 years building a world class fab is standard, not slow, and the New York fabs are getting their, but GloFo were new to the game in 2009, they had to start work immediately, and work finished at a not particularly useful time basically. In between 32 and 28nm, if GloFo was in business in 2005, they would have started the fab building in time to finish for the beginning of 32nm and started the expansion in time for 20nm. That is the advantage Intel/TSMC have had, they started their fabs at roughly speaking the right time for their processes, not years late. Those issues will iron out over time for GloFo, and no AMD shouldn't have dumped them.
TSMC has no process capable of making a 4Ghz Bulldozer, let alone a 5Ghz one, they had no process capable of making a 4Ghz Phenom 2. IBM could likely make it but isn't an options, Samsung don't offer what is required, Intel do... pretty sure AMD couldn't switch to them.
In terms of GloFo, they helped hugely in reducing AMD's debt, they've removed what is billions and billions each process upgrade on research and equipment and allowed them to expand production. The new york state fab cost probably 10billion by now to get where they have, AMD were 4billion in debt or so when GloFo was split off, GloFo(or lets call it AMD's manufacturing arm) used to be 1 real fab in Germany, it is now a better fab in germany, a huge new plant in New York with permission for expansion to three times the size there and several billion in fabs in Singapore.
GloFo is many many many times stronger than they were when AMD sold the german fabs basically. But in this industry it was always going to take GloFo the full length of building the new fab, and at least one more process change to get up to speed with other fabs and be able to offer multiple different processes.
TSMC and Glofo are both doing risk production of 20nm parts which are shipping, and both are set to offer real quantities of proper 20nm production roughly mid 2014. That means GloFo has caught up to TSMC in real terms which is actually a pretty outstanding achievement considering they have done, in the same time frame a 32nm process that TSMC dropped, and have an entire extra range of higher performance processes that TSMC don't offer at all.
Cliffs ;- the ONLY other company able to produce Phenom 2/bulldozer would have been Intel, a complete non starter.
Glofo are doing pretty well, if AMD had the cash that plant would have been started 18-24months earlier, finished in time for 32nm to ramp there and in time for a seamless and timely transition to 28nm, 20nm wouldn't have been delayed so long.
people think GloFo don't have 28nm working and in full scale production because AMD aren't making 28nm cpu's yet(high end cpu's), GloFo have been shipping 28nm like TSMC for a long time now, mostly arm chips like TSMC.
Last edited:
Caporegime
With the lack of L3 cache, I don't expect Kaveri to blow me away, so if it's at lynnfield IPC without the L3 cache colour me impressed.
Just don't go a bit over zealous like you did with the initial Trinity preview stuff
Caporegime
With the lack of L3 cache, I don't expect Kaveri to blow me away, so if it's at lynnfield IPC without the L3 cache colour me impressed.
Just don't go a bit over zealous like you did with the initial Trinity preview stuff
I'm not interested in any CPU that cannot match my i7 Bloomfied per core per clock for £150 at most.
Its serving me pretty well, i see no reason to swap it out for another unless i can do it brand new for that price.
Caporegime
Lynnfield and Bloomfield are pretty much the same IPC, Lynnfield it's like Sandy and SB-E. (Your i7 920 being a Sandy 3820 and the Lynnfield being Sandy on 1155 ; 2600k for the i7's or 2500k's for the i5's)
I don't see the point in that, as you are buying an additional card to crossfire anyway so you might aswell have had a decent igpu or just go for a discreet card in the first place.
A decent gcard say 7850, would be held back by the igp and the ddr3 system memory anyway.
Soldato
Thanks, that was a bloody good interesting read, and not the first one I've read from you.