AMD reveals new roadmap, Scorpius platform in 2011
- Thread starter MODEL3
- Start date
VirtualLarry
No Lifer
- Aug 25, 2001
- 56,587
- 10,225
- 126
Are AMD bypassing 32nm completely? I wondered about that, when they announced the 28nm (or was it 22?) for the ARM CPUs.
32nm SOI process technology is for AMD's own CPUs.
28nm high-K metal gate process technology is for ARM's CPUs.
http://www.xbitlabs.com/news/c...ays_for_Bulldozer.html
http://www.xbitlabs.com/news/o...M_Microprocessors.html
VR-ZONE is saying that there will not be a 32nm shrink for the Athlon II & Phenom II CPU line.
28nm high-K metal gate process technology is for ARM's CPUs.
http://www.xbitlabs.com/news/c...ays_for_Bulldozer.html
http://www.xbitlabs.com/news/o...M_Microprocessors.html
VR-ZONE is saying that there will not be a 32nm shrink for the Athlon II & Phenom II CPU line.
The Athlon II is extremely competitive, and Thuban will probable be a better value at the high end than quad core nehalems. If they are selling for the same price then I think Thuban will be able to outperform nehalem at most tasks. It does not appear like there will be 32nm quad cores in 2010, so people wanting high end CPUs for less than a grand(gulftown) are stuck on 45nm until Sandy Bridge and Bulldozer come out.
I don't really mind the lack of upgrade path for myself TBH. It is less tempting this way, and so I can spend money on something more important, like an SSD.
Anyway, what is the point of Fusion? Is it just going to be an SB850+ glued to a CPU? That's useful for laptops and other low end computers, but for the enthusiast it just seems like a waste of die space. Or is this going to be geared towards hardware accelerated OpenCL physics?
I don't really mind the lack of upgrade path for myself TBH. It is less tempting this way, and so I can spend money on something more important, like an SSD.
Anyway, what is the point of Fusion? Is it just going to be an SB850+ glued to a CPU? That's useful for laptops and other low end computers, but for the enthusiast it just seems like a waste of die space. Or is this going to be geared towards hardware accelerated OpenCL physics?
interesting, but i wonder if this will be at the beginning of 2010 or the end. would help in the decision making process (if to buy now or wait until the new stuff)
I am thinking that Bulldozer will work on AM3 motherboards, at the very least hte 800 series ones. It would make sense, given how weak AMD will be for the next year compared to Intel. If they can promise people that there is an upgrade path from K10 to Bulldozer, then people will purchase AM3 setups and wait for Bulldozer to get a drop in replacement, potentially instead of going with what would otherwise be a superior offering from Intel.
It can go the other way around too, with AM3r2 boards being backwards compatible with Thuban CPUs. If this is the case, it will increase the chance that I purchase a Thuban significantly, as it will allow me to reuse my Thuban from an AM2+ board in an AM3r2 board while waiting for Bulldozer to become cheaper and more mature.
It can go the other way around too, with AM3r2 boards being backwards compatible with Thuban CPUs. If this is the case, it will increase the chance that I purchase a Thuban significantly, as it will allow me to reuse my Thuban from an AM2+ board in an AM3r2 board while waiting for Bulldozer to become cheaper and more mature.
lopri
Elite Member
- Jul 27, 2002
- 13,314
- 690
- 126
If I'm not mistaken SB850 will support USB 3.0 and SATA III natively. It is also going to be the first AMD/ATI chipset with built-in gigabit networking.
We may see a very strange situation here because Intel's upcoming 57 series chipsets (P57/H57/Q57) won't have these (USB 3.0 and SATA III), according to the leaked roadmaps I've seen.
We may see a very strange situation here because Intel's upcoming 57 series chipsets (P57/H57/Q57) won't have these (USB 3.0 and SATA III), according to the leaked roadmaps I've seen.
manimal
Lifer
- Mar 30, 2007
- 13,559
- 8
- 0
http://www.farscapefantasy.com/creative/gallery/371.jpg
If this is the graphics on the and it ran like crap I would still buy it!!
heres hoping sb850 unlocks a single core into a eleventy core!!
Seriously wonder if this will matter. I guess I wont be too sad to not see a phenom II part duex. Wonder if the limit many of us hit around 4ghz is a wall the engineers are having trouble with.
If this is the graphics on the and it ran like crap I would still buy it!!
heres hoping sb850 unlocks a single core into a eleventy core!!
Seriously wonder if this will matter. I guess I wont be too sad to not see a phenom II part duex. Wonder if the limit many of us hit around 4ghz is a wall the engineers are having trouble with.
Consider it a redundant throw-in instead of a waste of space, and don't lose sleep over it. For enthusiasts, it means you've always got a backup GPU in case you fry your discrete card somehow, eevn if you don't have an IGP. The high-end motherboards don't want to waste space on IGPs, so at least this way you're not offline while you order a new video card.
Why do CPUs tend to be more unstable in 64 bit OSs, and is there a way to fix that problem with a change in architecture?
As for a 4ghz wall, that seems pretty likely to me, seeing as how neither company has ever shipped a 4ghz CPU in any form. That said, I was hoping for 32nm Thuban. At 45nm it might end up being relatively expensive.
64-bit OSes make use of on-chip 64-bit registers which are invisible to 32-bit OSes. these registers are more numerous and much larger than x86 registers. if you have a certain overclock and that is borderline stable/unstable, and you are only testing it in a 32-bit environment, then you are only using x86 registers and you cannot conclude whether or not your x86-64 registers are functioning properly because they are not used. it has nothing to do with the quality of the architecture; you just aren't using the "entire" chip, and your stress testing is therefore inconclusive. if the same overclock is unstable in a 64-bit environment, then something related to the execution of x86-64 instructions needs more volts or less frequency to run.
registers are not always the limiting factor in overclocking. on some chips it could be cache, or I/O, or pretty much anything. for a failure to occur under extreme frequencies and voltages, the whole architecture is up for grabs. however, to say that 64-bit systems are less stable than 32-bit systems, in general, is incorrect. 64-bit registers simply consist of many more transistors than the x86 registers and can be said to be slightly more error-prone as the chip approaches its frequency limit for a given voltage.
definitely do not be worried about the size of a fusion IGP. it'll be smaller than a core.
registers are not always the limiting factor in overclocking. on some chips it could be cache, or I/O, or pretty much anything. for a failure to occur under extreme frequencies and voltages, the whole architecture is up for grabs. however, to say that 64-bit systems are less stable than 32-bit systems, in general, is incorrect. 64-bit registers simply consist of many more transistors than the x86 registers and can be said to be slightly more error-prone as the chip approaches its frequency limit for a given voltage.
definitely do not be worried about the size of a fusion IGP. it'll be smaller than a core.
I don't think I believe the vr-zone article...
It's based on nothing but that inpai image, and I'm fairly certain that the image is wrong or faked.
For example RD890 is coming out in Q4 2009, not in 2010...and the DX11 discrete graphics have already come out (so not to be 2010).
It's based on nothing but that inpai image, and I'm fairly certain that the image is wrong or faked.
For example RD890 is coming out in Q4 2009, not in 2010...and the DX11 discrete graphics have already come out (so not to be 2010).
I don't know if vr-zone is correct or not,
but before vr-zone (and inpai) (1,5 month earlier...) there were many sites like techpowerup and many others that said the same things.
For example, lets take XbitLabs article (09/01/2009)
(http://www.xbitlabs.com/news/c...Desktop_Processor.html
In this article, XbitLabs was saying Thuban for Q3 2010 but other sites was saying Q2 2010 (but all of the sites was saying RD890 in 2010...)
Possibly AMD delayed the RD890 (it was originally to be launch in Q4 2009) and accelarated the Thuban launch in order the launch to be simultanious. (just a guess)
About DX level the logic thing is for RD890 to be based on RV710 (4350).
The DX11 succesor to 4350 (cedar) will launch in Q1 2010.
Yes i read your post:
I was commenting on something else, i was commenting on the DX level of 890GX. (the roadmap doesn't say which DX version)
---------------------------------------------------
Now regarding your comment: (so not to be 2010)
The roadmap shows 2010 for the "Leo" platform, which consists (according to the roadmap) from:
Q2 2010 "Thuban" CPUs
Q2 2010 RD890+SB850 chipset
Q3-Q4 2009 ATI Radeon DX11 discrete graphics
just like the "Dragon" platform (8 Jan 2009) consists from:
Q1 2009 "Phenom II" CPUs
Q3 2008 790GX/FX chipset
Q2 2008 ATI Radeon HD4000 (4870) discrete graphics
http://game.amd.com/us-en/landings/Dragon.aspx
EDIT*
RD890 will be a must-have chipset with USB3 and SATA3.
Possibly with SB850 it will have also better raid performance than the past.
I've seen rumahs of PCIe Gen3 in the sb8xx specs, too.
I'm a little annoyed because sb8xx was 'unofficially' to be released right now but I understand from an economic/tech standpoint it really doesn't make any sense.
Tweaking the arch for a smooth transition for their next round of CPUs seems logical.
Remember: Fusion's main function is to harness the parallel processing power of the GPU. The graphics aspect is certainly a function but the new XOP instruction set is meant as a leap in overall IPC.
Initially it was to debut on a mobile platform but who knows these days ....
looks like in 2010 not too much no stuff from amd, just release of 6cores and some refreshing of older cpus. this is going to be a tough year for them facing onslaught of 32nm new CPUs from intel. i see intel getting another winning year here. of course with x4 620 selling for so low, I think 2010 is the year for quads for new builds.
i wouldn't call it onslaught, but it's going to be tough. intel isn't going to introduce any 32nm quads, either. we're probably just going to see a price war built around the same stuff we have right now with new chipsets and clarkdale, and probably the pineview atom processor. the desktop performance segment isn't going to get any shakeups, unless 5 ghz clarkdales become commonplace in the overclocking community. i would jump on that in a heartbeat.
AMD needs something a little more compelling in the mobile segment. they should have realized years ago that procrastinating on HKMG would be a huge mistake. they cannot achieve anywhere near the low power of intel's processors. in fact, their entire lineup is inefficient across the board. their 45nm quads consume more energy than intel's 65nm generation without HKMG, and that sucks. without high-k they have no chance of a SoC device to compete with pineview. the netbook market exploded and AMD has nothing suitable for it. how many times do they have to miss the boat on emerging segments before someone buys them out?
they should've known that a lack of HKMG in their process would ruin their performance in the mobile segment, yet they aren't even doing anything, even though the mobile segment grows larger each year.
AMD needs something a little more compelling in the mobile segment. they should have realized years ago that procrastinating on HKMG would be a huge mistake. they cannot achieve anywhere near the low power of intel's processors. in fact, their entire lineup is inefficient across the board. their 45nm quads consume more energy than intel's 65nm generation without HKMG, and that sucks. without high-k they have no chance of a SoC device to compete with pineview. the netbook market exploded and AMD has nothing suitable for it. how many times do they have to miss the boat on emerging segments before someone buys them out?
they should've known that a lack of HKMG in their process would ruin their performance in the mobile segment, yet they aren't even doing anything, even though the mobile segment grows larger each year.
Even better, you know someone is going to release a driver for the IGP that allows it to be a physics engine or somesuch, you can bet that modders aren't going to allow AMD to put a useless graphics card onto their processors. It will be useful whether you use it for graphics or not.
Bateluer
Lifer
- Jun 23, 2001
- 27,730
- 8
- 0
This is a serious concern, the mobile market across all its segments is growing rapidly. And AMD has nothing for that market. Even on the midrange and highend, they have nothing. Its truly pathetic.
On the positive side though, their graphics card division is slapping the competition around. Adding to that that their competition could very well lose 1/3rd of their revenue by discontinuing their chipset business, and AMD's graphics and chipset business looks brighter. On the desktop side.
I've been considering upgrading my A64 X2 Brisbane test beds to Athlon II X4 620s.
cbn
Lifer
- Mar 27, 2009
- 12,968
- 221
- 106
This doesn't sound too optimistic. I must be the server market is more lucrative than the enthusiast market.
cbn
Lifer
- Mar 27, 2009
- 12,968
- 221
- 106
USB 3.0 and SATA III might be the best reason to look into AM3. When will we see these new boards?
EDIT: nvm...I found the answer-->http://www.techpowerup.com/91528/AMD_SB850_Southbridge_to_Pack_Gen._3_SATA_6_Gbps_Support.html (Notice no mention of USB 3.0)
Last edited:
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter