AMD Piledriver rumours ... and expert conjecture

Page 100 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
We have had several requests for a sticky on AMD's yet to be released Piledriver architecture ... so here it is.

I want to make a few things clear though.

Post a question relevant to the topic, or information about the topic, or it will be deleted.

Post any negative personal comments about another user ... and they will be deleted.

Post flame baiting comments about the blue, red and green team and they will be deleted.

Enjoy ...
 
and what's even more disappointing is that we were promised a stepping revision for the FX-Bulldozer.
I thought it was to be the FX-4170, I was wrong.
now word of three or four more FX chips coming out and still no stepping revision.
so basically it's the same chip @ different speeds.
same crap.

When is Piledriver coming again.? :??:

That means there wasn't a small enough fix to make to justify the cost of a new step. Since AMD gave up their fabs they can't eat the cost of new mask as easily as they used to.

We could see one but I wouldn't bet on it. AMD has a lot of silicon in the queue as it is.
 
and what's even more disappointing is that we were promised a stepping revision for the FX-Bulldozer.
I thought it was to be the FX-4170, I was wrong.
now word of three or four more FX chips coming out and still no stepping revision.
so basically it's the same chip @ different speeds.
same crap.

When is Piledriver coming again.? :??:
This is the same reason BD got a crappy welcoming mat, people taking scoured documents as AMD said this and that.

AnandTech has found ‘an AMD BIOS and kernel document’ mentioning a B3 revision of the AMD FX-8150 CPU, which is based on the Bulldozer architecture.

AMD promised ... where? AMD promised B3 but screwed the pooch and didn't deliver ... this was never the case and bad memory is what causes bad reputation. Remember how BD was supposed to blow the doors off of Sandy Bridge? I don't remember any claims by AMD stating that, just a little twist on comparing it to a $1000 CPU on a price/performance slide that got interpreted as such. Its things like this that makes the welcoming mat into a trap door to hades.

Chances are they are re-working BD to include clock mesh and did away with B3 and going straight to PD.

If AMD switched thier goal to making sure PD is on schedule rather than tweaking B2, good for them. Keep in mind when trying to figure out whats going on that AMD is on a tight budget and doesn't have Intel's 1.5B/quarter profit to play with.
 
trinity benchmarked (hopefully not a repost):
http://www.fudzilla.com/home/item/26445-amd-trinity-a10-5800k-benches-are-out
http://www.legitreviews.com/news/12666/
http://wccftech.com/amd-trinity/
some nitpicking -
legitreviews says that trinity will have 4mb l3 cache. yay...wait.. the cpu-z screeny shows no sign of l3 lol.
the package section shows socket fm2 (904) - so trinity will indeed be incompatible with llano. desktop llano users are :sol:
looks like full instruction set support like earlier rumors - this will gain trinity a lot of value at entry level.
fudzilla, wccftech and legitreviews all said the exact thing while trinity's lack of l3 cache was known already. i wonder if they mixed up trinity with low end pd specs.
 
fudZila conjecture;

AMD has nothing to counter Haswell, Ivy Bridge..."
http://www.fudzilla.com/home/item/26452-intel-to-dominate-high-end-processor-market-throughout-2013

I think the sad thing is that those are not the high end market for Intel. For Intel, high end is SB-E and will be IB-E. Haswell and IB are mainstream which is shy the CPUs only go to $300ish at the top end.
 
Limited Trinity news and now Kaveri is ahead of schedule? LoL

http://www.fudzilla.com/home/item/26451-kaveri-28nm-apu-scheduled-for-january-2013
Well, GCN is already out and ready anytime they want to fuse that to a die, so its not going to be a wait on that. I worry about the time they are taking to improve Steamroller though.
3 things:
1.Trinity is probably not going to use GCN, as mentioned before, because GCN is 28nm, so it only makes sense for them to wait and use it with much less hassle with Kaveri.
2.Trinity is probably going to be Q2 release if Kavari is coming Q1 2013.
3.Piledriver will be out not long after, because releasing 2.5 before 2 is something even AMD would not do.
 
1.Trinity is probably not going to use GCN, as mentioned before, because GCN is 28nm, so it only makes sense for them to wait and use it with much less hassle with Kaveri.

Yeah I think the confusion came from mixing GCN+VLIW in the Radeon 7000 naming scheme. They need to be able to Crossfire with a similar low end 7K GPU.

Trinity will be VLIW but at 32nm. A slight boost over the 40nm implementation.
 
Yeah I think the confusion came from mixing GCN+VLIW in the Radeon 7000 naming scheme. They need to be able to Crossfire with a similar low end 7K GPU.

Trinity will be VLIW but at 32nm. A slight boost over the 40nm implementation.

VLIW4 isn't bad and is a nice improvement over VLIW5 that is in Llano but it made Trinity's die even larger than Llano's. The gpu takes up just a little over half the total die but a little surprised that got such good clocks with less TDP than Llano on the gpu side. We already know how the cpu is going to turn out but still it will do great gaming for those on the mobile. I am interested in the next generation that is to come out next year. Next year's apu will be GCN but likely still Piledriver based with spit shine improvements.
 
VLIW4 isn't bad and is a nice improvement over VLIW5 that is in Llano but it made Trinity's die even larger than Llano's. The gpu takes up just a little over half the total die but a little surprised that got such good clocks with less TDP than Llano on the gpu side. We already know how the cpu is going to turn out but still it will do great gaming for those on the mobile. I am interested in the next generation that is to come out next year. Next year's apu will be GCN but likely still Piledriver based with spit shine improvements.
Krishna (apu generation after Trinity) will be using Steamroller cores, thus the reason it will be 28nm.
 
All those sites are posting the exact same set of bench's, meaning they all got the same screenshots. And while I doubt the validity of all prerelease bench's those are very indicative of how BD works in general. Amazing increase in GPU performance, moderate increase in CPU performance.

To all those stamping their feet and flaring their nostrals, SuperPI is an absolute horrible benchmark these days. It use's old 8087 FPU instructions instead of the decade old SIMD instructions even though the math it's attempting to do is perfect for conversion to SIMD. AMD doesn't do 80-bit 8087 directly in hardware, instead they translate it into a 128-bit SIMD instruction and execute it.

So what your seeing is the 8087 execution performance of one core on a BD CPU. Really just test's the cache performance more then anything. So what we've got is a nice 12% boost to cache performance at the same or less TDP and with a big boost to GPU performance.
 
if they can hit 20% for cpu it would be pretty amazing.

I would love a trinity ultra thin.

I was going to say build your own but is that even possible? I know I have seen notebook cases for sale in the past. Would be fun though to be able to select your own components and upgrade.
 
if they can hit 20% for cpu it would be pretty amazing.

I would love a trinity ultra thin.

I don't care about ultra thins much but want to see Trinity being backward compatible with socket FS1 but maybe just day dreaming at this point as I am yet to find any data on the difference between socket FS1 and FS1r2. Who knows it might only just be a chipset revision and a bios update.
 
I don't care about ultra thins much but want to see Trinity being backward compatible with socket FS1 but maybe just day dreaming at this point as I am yet to find any data on the difference between socket FS1 and FS1r2. Who knows it might only just be a chipset revision and a bios update.
I highly doubt this since I think FM2 has more pins
 
Globalfoundries: 32nm Yield Problems Are Behind, Shipments to AMD Now Unconstrained..."

http://www.xbitlabs.com/news/other/display/20120321200633_Globalfoundries_Celebrates_Shipment_of_250_Thousand_32nm_HKMG_Wafers_to_AMD.html

Thanks for the link, gallardo, but I shall have to report you for having a frightful avatar if you don't change it.
 
Status
Not open for further replies.