AMD CPU speculation... and expert conjecture

[jdwii]

I love how TR tries to twist the actual numbers around. First off, they used skyrim, amd's worst game in their benches.

Intel pre-encoding 3570k, 104FPS
AMD 8350 pre-encoding, 77 fps

with endocing, Intel, 82fps, amd 62

both dropped by ~20%, so what does it prove? nothing really, pretty much every one of the cpus dropped close to 20% until you look at the dual core cpus, pentium g2120 and I5 655 hit the hardest, go from 78 fps to 37. Conclusion : dual core are not for multitasking at all.

other than that, take a small sample for a program that already heavily favors intel, it still favors intel if you try other things.

I agree i wish they would make a site that would review 20+ games on a CPU kinda like one of my favorite sites techpowerup(which they did do what i wanted with bulldozer)

http://www.techpowerup.com/reviews/AMD/HD_7970_CPU_Scaling/

Intel shot themselves shifting Haswell to a new socket, it made end users irrate when IB a side grade to SB came out with a new chipset with neglible performance gains despite the faster IMC and goodies thrown onto the new chipsets. I feel they are going to take the flack for releasing a new socket for a same process part which ironically is another side grade made more expensive by the platform change. Then there is the iGPU, now many sane reviewers told people to hold their horses and not expect a golden lolipop, well the hype over GT is going to hit the ground hard when it barely manages to compete with Llano. The ultimate realisation here is that Intel has its fails as well.

Wow that is funny and i find it quite sad that they are going to be forcing users to upgrade their boards for this and i also find it even more funny how much they hyped of their IGPU. They had me believing it as well since well its Intel they really have no reason at all to hype anything. If you cant answer this then fine but do you think Trinity to richland is a bigger upgrade then Ivy to haswell?

 

[gamerk316]

^^ Ivy -> Haswell is probably a bigger upgrade, due to how weak Intels iGPU is in comparison to Trinity. Not saying Haswell will be better then Richland, but in terms of improvement?

Still, if Intel can get its iGPU even at Trinity performance levels, that would be worrying for AMD. Thats where they are strongest, and thus have most to lose.

 

[de5_Roy]

I agree i wish they would make a site that would review 20+ games on a CPU kinda like one of my favorite sites techpowerup(which they did do what i wanted with bulldozer)

http://www.techpowerup.com/reviews/AMD/HD_7970_CPU_Scaling/

Wow that is funny and i find it quite sad that they are going to be forcing users to upgrade their boards for this and i also find it even more funny how much they hyped of their IGPU. They had me believing it as well since well its Intel they really have no reason at all to hype anything. If you cant answer this then fine but do you think Trinity to richland is a bigger upgrade then Ivy to haswell?

wow, i got a feeling that the tpu article might have incited more vitriol from amd fanboys than tom's 2011 sub $200 gaming cpu roundup. in some cases 20% faster at stock speeds, wth. okay.. the tpu article only tests 3 cpus. Phew! good thing the pentium or the core i3 wasn't there at that time. 20% was plenty of space for core i3 to fit into.
actual problem with haswell might come with mainstream broadwell which is rumored to be soldered on the motherboard (bga packaging). that means, users with haswell cpus might be in a bit of a pickle if they want to upgrade from a haswell core i3/pentium to broadwell core i5/i7. it's in the future though. i wanna see how intel handles that issue..... well... how big a jerk they turn into...
the bigger problem is intel's casual attitude about haswell's rumored usb 3.0 issues that intel won't fix at launch. i wonder if thunderbolt has the same issue or intel is crippling usb 3.0 to push thunderbolt. the last one is conspiracy-bait for amd fanboys to go crazy over.

intel's policy seems to be supporting two lineups on each socket since lga 1156.

intel would hype a pice of rock if they thought it could make them some money. they claimed triple digit performance increase with new haswell igpu. forgot if it was 120% or 250%. iirc, turned out that they were comparing ivb hd2500 (6 shaders) to haswell hd4600 (20 shaders). hd4000 has 16 shaders, btw.

imo, richland is not a worthy upgrade over trinity. as a new buy, it might be okay for laptops.

don't belive intel/amd/nvidia pr hype.

 

[sarinaide]

Yep, Richland is only a worthwhile upgrade if;

a) new buyer
b) have money to spend on the 10% gain you get, which is sizeable but not enough compared to Kaviri, just seems to be a mid step waste for existing owners.
c) Ties into b) you are a avid AMD collector like me

 

[-Fran-]

Uhm... It's not an upgrade if you're a new buyer, Sarinaide, hahaha. But yea, I agree with those points.

Regarding Haswell and IB. Intel decided to beef up the iGPU instead of beef up the CPU thanks (in part, I think) to AMD's big BD blunder. Since PD showed a very good improvement, I believe they'll be more pressed with SR and will improve CPU (maybe a lot) with Broadwell.

That's my take in the future. And yes, I do hope SR turns out to be better than expected.

Also, why TSMC's process wouldn't go beyond 3.1Ghz? ARM designs are very different from what AMD and Intel do to reach certain speed targets. I'd love to know about that.

Cheers!

 

[mayankleoboy1]

Is it clear from the rumours whether the USB3.0 bug is in the chipset or the CPU ?
EDIT : What all is integrated on the CPU die ?

 

[gamerk316]

actual problem with haswell might come with mainstream broadwell which is rumored to be soldered on the motherboard (bga packaging). that means, users with haswell cpus might be in a bit of a pickle if they want to upgrade from a haswell core i3/pentium to broadwell core i5/i7. it's in the future though. i wanna see how intel handles that issue..... well... how big a jerk they turn into...

Remember that ~90% of all Intel CPUs sold are to Dells, HP's, and so on. These are the people who NEVER upgrade, and buy a new PC every decade or so. Point being, by the time they DO start looking for upgrades, they are typically on a socket thats no longer supported anyway, necessitating a motherboard upgrade. So I view this as a non-issue. Gamers will go to the non-soldered versions on the chip.

the bigger problem is intel's casual attitude about haswell's rumored usb 3.0 issues that intel won't fix at launch. i wonder if thunderbolt has the same issue or intel is crippling usb 3.0 to push thunderbolt. the last one is conspiracy-bait for amd fanboys to go crazy over.

Remember the time it takes to re-spin. Intel probably doesn't want a three month delay, so they are forcing the chipset makers to "agree" to the issue now, while they re-spin and fix the problem [Similar to the B2->B3 SB revision].

 

[de5_Roy]

Remember that ~90% of all Intel CPUs sold are to Dells, HP's, and so on. These are the people who NEVER upgrade, and buy a new PC every decade or so. Point being, by the time they DO start looking for upgrades, they are typically on a socket thats no longer supported anyway, necessitating a motherboard upgrade. So I view this as a non-issue. Gamers will go to the non-soldered versions on the chip.

yeah...well... that's true for intel-based pre-built pcs. it's more like they cannot ever upgrade. by the time a haswell prebuilt pc will show signs of slowness, a new platform will be available.
the second part about gamers, or people who build their own pcs - those might have issues. especially the section of the diy builders who start with low cost builds. for example, if intel decides to solder sub $200 cpus, then core i3 and lower will never be able to upgrade. incidentally, core i3 and lower cpus are also prime candidates for pre-built pcs. otoh, if intel decides to solder core i5 and lower, then even mainstream gamers will have problems upgrading cpus, in case they want to switch to a core i7, or a if partially locked core i5 owner wants to switch to an unlocked cpu. lga 2011 being fully customizable won't help at all, because it'll let intel [strike]extort[/strike] charge even more.

Remember the time it takes to re-spin. Intel probably doesn't want a three month delay, so they are forcing the chipset makers to "agree" to the issue now, while they re-spin and fix the problem [Similar to the B2->B3 SB revision].

ugh. i sorta remember. i frequently use usb, so it coulda make an issue for me. y'know, this effectively makes ivy bridge the most problem-free platform despite the poor t.i.m. and overclockability. cougar point bug was the second biggest reason i missed out on sandy bridge. the first being bulldozer hype.

at least amd fanboys will sleep in peace mumbling,"22nm and intel still has a hardware bug! amd ftw!!"


listen up c.a.l.f., here is a weapon to use against the m.i.l.f. - haswell first rev. has a hardware bug. intel sucks. this bug is So big that it totally balances phenom i's bug. intel has had cougar point bug and this one now, and another two with virtualization - vt bug on sb-e and the one where guest can infect the host pc or something like that (every 64-bit cpu prior to haswell). be Happy!

 

[mayankleoboy1]

at least amd fanboys will sleep in peace mumbling,"22nm and intel still has a hardware bug! amd ftw!!"


listen up c.a.l.f., here is a weapon to use against the m.i.l.f. - haswell first rev. has a hardware bug. intel sucks. this bug is So big that it totally balances phenom i's bug. intel has had cougar point bug and this one now, and another two with virtualization - vt bug on sb-e and the one where guest can infect the host pc or something like that (every 64-bit cpu prior to haswell). be Happy!

Dont.Bait.The.Fanbois :lol:

 

[de5_Roy]

AMD Kaveri Unveiled: PC Architecture Gets GDDR5
http://www.brightsideofnews.com/news/2013/3/5/amd-kaveri-unveiled-pc-architecture-gets-gddr5.aspx
calling this interesting is an understatement. i guess when amd scrapped sr last year they really went back to the drawing board.
i thought that gcn enabled unified addressing, with kaveri it looks like amd developed some kind of a unified/hybrid memory controller.
amd developing gddr standards help a bit more. rambus would bleed people to skeletons before doing something like this.
this'll be great for excavator-based apus - ddr4, gddr6, full integration and massive bandwidth that the igpu rightfully deserves. let the dreaming begin.

 

[-Fran-]

Ever heard of "it takes a stupid to call on a stupid"? Come on guys, you're not 10 year olds... Right? hahaha.

The problems with hardware after the design and build are common issues across the board for every maker. You just have some bad ones and terrible ones, but every maker has them. If you wanna go back to "terrible" ones, the first gen Pentiums had a very nasty math issue in them; just as bad as the TLC bug in PhI.

Moving things into the die space for Intel is a natural evolution of things, given history. We need to adapt. The markets and buyers have to adapt. I still remember the Cache on-board for Pentiums and older CPUs. Yeah, you lose something, but the gains *should* overweight the problems, right?

Cheers!

EDIT: GDDR5 soldered to the MoBo? That could work wonder in embedded systems, but not in regular DIY PCs... Interesting non the less!

 

[Cazalan]

calling this interesting is an understatement. i guess when amd scrapped sr last year they really went back to the drawing board.
i thought that gcn enabled unified addressing, with kaveri it looks like amd developed some kind of a unified/hybrid memory controller.

Jaguar also has a GDDR5/DDR3 interface so not a big stretch for AMD to add to all their APUs. These new APU platforms could end up looking more like a big graphics card.

The downside is that 4GB cap with current chips. It doesn't leave much for Windows.

 

[gamerk316]

ugh. i sorta remember. i frequently use usb, so it coulda make an issue for me. y'know, this effectively makes ivy bridge the most problem-free platform despite the poor t.i.m. and overclockability. cougar point bug was the second biggest reason i missed out on sandy bridge. the first being bulldozer hype.

I still have a B2 revision board. Lost the DVD-ROM drive, likely due to the SATA II bug, but the two physical HDDs are on SATA III, and I have an external DVD Drive, so I don't care.

AMD Kaveri Unveiled: PC Architecture Gets GDDR5
http://www.brightsideofnews.com/news/2013/3/5/amd-kaveri-unveiled-pc-architecture-gets-gddr5.aspx
calling this interesting is an understatement. i guess when amd scrapped sr last year they really went back to the drawing board.
i thought that gcn enabled unified addressing, with kaveri it looks like amd developed some kind of a unified/hybrid memory controller.
amd developing gddr standards help a bit more. rambus would bleed people to skeletons before doing something like this.
this'll be great for excavator-based apus - ddr4, gddr6, full integration and massive bandwidth that the igpu rightfully deserves. let the dreaming begin.

Usage of DDR3 and GDDR5 is mutually exclusive.

I dislike this, very much. Good for the GPU, horrid for the CPU. Non-memory bottlenecked applications that are not sensitive to CPU cache are going to suffer due to GDDRs higher latencies.

 

[mayankleoboy1]

^ but who cares for the CPU ? GPU FTW!!!1!

 

[$hawn]

Usage of DDR3 and GDDR5 is mutually exclusive.

I dislike this, very much. Good for the GPU, horrid for the CPU. Non-memory bottlenecked applications that are not sensitive to CPU cache are going to suffer due to GDDRs higher latencies.

I think its more probable that the mobo will have like 1-2GB's of GDDR5 soldered close to the APU socket for exclusive use by the GPU, and normal DDR3 RAM slots for the CPU.

The next big question is, if so, then how do they manage that unified memory architecture for the CPU and GPU? Confusing!!

 

[jdwii]

Mid 2014 we might have DDR4 ram and GDDR6 video cards just found out about GDDR6 today.

 

[Cazalan]

I think its more probable that the mobo will have like 1-2GB's of GDDR5 soldered close to the APU socket for exclusive use by the GPU, and normal DDR3 RAM slots for the CPU.

The next big question is, if so, then how do they manage that unified memory architecture for the CPU and GPU? Confusing!!

When they say mutually exclusive that means the IMC can run in either GDDR5 mode or DDR3 mode. Not both at the same time. They would have to add a second memory controller to do both.

 

[de5_Roy]

Jaguar also has a GDDR5/DDR3 interface so not a big stretch for AMD to add to all their APUs. These new APU platforms could end up looking more like a big graphics card.
The downside is that 4GB cap with current chips. It doesn't leave much for Windows.

that's for console version of jaguar. iirc i didn't see any mention of gddr5 in desktop/laptop/tablet version of jaguar-related rumors/leaks until ps4. it'd be good if amd kept the gddr5 controller when they said they'd sell a stripped down version of the ps4 apu. spinning jaguar/kabini hype as ps4 apus is a good move from amd. before it was 'amd fusion' and crap with gcn and hsa mixed in. but consoles bring in much larger exposure, so if amd hypes it as a 'version of the ps4 apu' instead of amd fusion/vision/convolution they stand to gain more. since the apu is well known to be low power, low cost apus, amd has very little to lose. this could end up being the polar opposite of bulldozer hype (thankfully, jag doesn't seem to share architecture with bd apart from modularity ).
an ultrathin based off kabini with a gig of gddr5 soldered on to the mobo would be good for the form factor. slightly different take: put the gddr5, more storage, fatter batteries on the keyboard/charger dock so when kabini tablet is combined with the dock, the apu gains desktop-class gpu power with ram bandwidth to match.

^ but who cares for the CPU ? GPU FTW!!!1!

when it comes to amd apus, hell yeah! sidenote - amd's ddr3 controller is a bottleneck anyway, cpus are [strike]meh[/strike] adequate. i guess directly using gddr5 was the best way for amd to alleviate bw demands from the igpu - if teh rumorz are true.

....
sorry if my speculations sound more like delusions. i heart mobile apus.

 

[gamerk316]

The next big question is, if so, then how do they manage that unified memory architecture for the CPU and GPU? Confusing!!

Not really; its trivial to do. Its one address space. No more "the application exists within a 4GB Address Space, but the GPU has its own ~2GB data cache specifically for its tasks"; its all the same pool now. Simplifies the memory subsystem quite a bit actually.

 

[griptwister]

Mid 2014 we might have DDR4 ram and GDDR6 video cards just found out about GDDR6 today.

Link Pwease?

 

[noob2222]

AMD Kaveri Unveiled: PC Architecture Gets GDDR5
http://www.brightsideofnews.com/news/2013/3/5/amd-kaveri-unveiled-pc-architecture-gets-gddr5.aspx
calling this interesting is an understatement. i guess when amd scrapped sr last year they really went back to the drawing board.
i thought that gcn enabled unified addressing, with kaveri it looks like amd developed some kind of a unified/hybrid memory controller.
amd developing gddr standards help a bit more. rambus would bleed people to skeletons before doing something like this.
this'll be great for excavator-based apus - ddr4, gddr6, full integration and massive bandwidth that the igpu rightfully deserves. let the dreaming begin.

I wonder if thier NDA source actually named the cpu when they talked about it or if BSN just added in "Kaveri". This is NDA information, likely with a lot of "fill in the _____"

http://www.tomshardware.com/news/A-Series-APU-PlayStation-4-John-Taylor-Radeon,21290.html

looks more plausable that this article is talking about the PS4 chip without sony's IP.

But as calazan stated, its possible that they put the memory controller on the chip specifically for oem vendor machines, laptops in particular. User built pcs will likely only be ddr3.

As for steamroller, there is no need to put gddr5 support on it, there is no igpu. Most likely part of the delay for steamroller was to get engineering samples for Sony's ps4 and MS's xbox apus.

 

[jdwii]

Link Pwease?

http://vr-zone.com/articles/gddr6-memory-coming-in-2014/16359.html

Over 8 months old buy hey

http://www.anandtech.com/show/6619/crucial-demonstrates-ddr42133-modules

DDR4

They say late 2014 Amd probably wont have it until excavator my guess since Steamroller is going to work on AM3+ and its to late to put a DDR4 controller in steamroller.

 

[-Fran-]

Not really; its trivial to do. Its one address space. No more "the application exists within a 4GB Address Space, but the GPU has its own ~2GB data cache specifically for its tasks"; its all the same pool now. Simplifies the memory subsystem quite a bit actually.

Heeeeh... Now that you mention it, using 64bit addressing they should have *plenty* of space to reserve some addresses, hahaha.

Thing is, 64bits

Cheers!

 

[griptwister]

My question is... Will DDR4 effect gaming performance like DDR3 compared to DDR2? I'm very interested in this GDDR6 though...

About these APUs... I like them. They're a fantastic idea for laptop users and HTPCs. I have a i7 3630QM laptop, I have to run BF3 in windowed mode at 480p to get playable FPS... my friend has a APU Laptop (I've got no idea which APU tbh) and he runs it on full screen at 720P at about 35FPS

 

[-Fran-]

AFAIK, DDR4 is quite a major overhaul of how DDR operates. So it won't only have increased density and higher speeds, it will also sport better latency and GANG BANG... No, wait, better gouging... Yes, that... I think, hahaha.

Cheers!