AMD Piledriver rumours ... and expert conjecture

Page 33 - 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 ...
 
It just goes back to Bulldozer being a server-oriented concept that AMD tried to wedge into an enthusiast product.

It most likely would have worked if they had been able to keep single-thread performance consistent with Phenom II's. Unfortunately it appears that they dropped the ball on cache performance and/or prediction and IPC took a sizable hit.

If they can get IPC back to Phenom II levels with PD then they may have a decently competitive CPU again.

If they get it back to Phenom II levels they would be back at square one. Intel has IB coming out in April. It will have better performance than SB that much we know. Then in 9 months to a year is Haswell.

PD needs to do two things: Drop power draw a lot because man BD is a hog and get performance above Phenom II in every aspect. Then it may be competitive in the DT market.

Then again AMD isn't focusing on Intel anymore as they have moved past that.

Still if PD performs better than IB it will cost you plenty to have that as we are seeing from the HD7970 thats priced way to high for my liking. And I do want a HD7970, just don't want to spend that much for it. Of course I will probably wait until Kepler is out as that will probably help drive the price of it back down to normal levels.
 
Watching with interest. Still loving my 1100T / 3.8ghz-4T on stock cooler :) Hope PD does end up AM3+ as I would like an upgrade path at some point. Regardless, VERY pleased with my current processor.
 
when i was playing with vm i found this strange thing that can help to get more performance from more core for single threaded or less threaded tasks.

as you can see in vm (only with 1 core)
i am running prime95 in this vm (running 1 thread only)
but in base os (task manager) both cores are getting utilized (only 50% of total cpu)
this means that 2 threads are running (i.e, 1 thread of prime95 in vm is converted into 2 thread to share workload on cores)

and thus we (software or microcode developer) can develop a software/microcode that will help to share workloads of 1 thread on different cores and will utilize all cores of cpu (that means more core will give more performance in lightly/single threaded task), i.e, bd and pd can be good for games (that use only 1-4 threads) and thus amd can make a comeback with bd and pd

what you think
 
How much better is PD supposed to be over BD? AMD said 10-15% on some of their marketing slides, any other bumps i'm missing?




Well the rule of the thumb is to take what Amd says and divide it by 3. So i would say about a real increase of 5% over BD per/clock. And i'm not kidding. The real increase will come from higher clocks. I lost faith in Amd in making high-end CPU's after Some Amd engineers said bulldozer wasn't supposed to be high-end even though their asking a high-end cost for them(anything over 200$). I never knew Amd would redo a whole design and spend 3+ years on a project to get ok performance i was pretty sure Amd already had that. But getting off track Amd needs to improve their IPC by 33% or more.

I also hate it when Amd says they made BD for high-end res when the I5 still beats it at that. And how they made it for multithreading when a 6 core phenom pretty much tie's it in most benchmarks with 2 less cores and 300mhz less clock rate and pretty much No turbo. At a much cheaper price.
 
with these new hotfixes on my bd rig I got a 12% increase over the standard win 7 os so if this is any sign of how much a improvement we could expect with the pd i'll be waiting to get on as long as it stays am3+
 
so now the FX-8120 can finally equal Thubans in overall performance.? :heink:

in gaming I'll still take a higher clocked Deneb instead.
I agree that it's definitely a start.
with a new stepping (Bulldozer FX-4170, FX-8190) and then release of Piledriver.
we'll see how it goes 1 year from now...
my rig is nothing special just a single gpu rig but i hope someone else gets the same on a multi gpu rig, also it is a passmark test not sure how reliable it still is but it as a 12% improvement for me
 
about 1629 before and now 1886 but i had other stuff running at the same time and that is without overclocking the cpu as well i'm expecting maybe 2000+ if overclocked that was for full system i'll get the bd its self here in a minute
 
my rig is nothing special just a single gpu rig but i hope someone else gets the same on a multi gpu rig, also it is a passmark test not sure how reliable it still is but it as a 12% improvement for me

Passmark is synthetic so overallit wont give a realistic increase, only a more generic increase. Games, apps like encoding, zipping etc will be true judges. You should run the apps pre hotfix and then after then calculate the difference.

As for the WEI, I don't trust it at all. I have built many systems and I have seen the CPU score for a 2500K be 5.0 and other 7.5 and in between. I have also watched the GPU score change by 1-2 points depending on the drivers themselves. Its too erratic to trust.

Right now my Q6600 is at 7.3 but my HD5870 is at 6.0 when it was at 7.6 when I first got it. Just to give an example, I just redid our security DVRs setup at work. Its running a Core i3 2100 and using the HD2000 IGP. It gets, as of latest drivers, a 5.1 WEI. .9 points away from my HD5870. Of course I am running the 12.1 betas but still it makes no sense as to why the WEI is soo erratic.
 
^^ Thats why I was always worried about Trinty; llano was best in low-power devices like laptops, which is very power concious. BD simply does NOT seem that well suited to that task.
remove 8mb l3 cache and 4mb (2 per core)l2 cache, you reduced power consumption by ?? In the die shots, the cache looks to be ~75% to the size of the module. Just a swag, Id say ~1.2ghz at a voltage of 0.7ish. only one module and a tiny graphics section
 
Interesting scores, my 1100T overclocked at 4ghz (for testing only, normally run at 3.8ghz on stock cooler...
- WEI 7.6 Once hit 7.7 but have never seen that again so must have been a fluke
- Passmark at 4ghz 7636.6

Bigtex2011, were those scores at the stock clock? Interesting if it is, with the normal 4.6ghz overclock I see on BD, she should be well in the high 8K range

Malmetal, agreed. I have also seen very consistent numbers chip per chip save the oddball but it was not off by much on WEI. I was recently comparing A4, A6 and A8 equipped laptops and they were practically dead on consistency wise depending on what chip they had. Memory values and other were all over the place. So gives a general idea of a change.

Edit: New score, 7636.6 on the stock cooler :)
 
Nice! I am confident I can wring a 7,800 out of my 1100T with real cooling (as opposed to my fake cooling haha) Which I am happy with.

Plans to overclock at all? With a good CPU-NB overclock + getting her to 4.5ish ghz, number wise she would be pretty impressive. I do not see 9K being out of the question if one was just looking at passmark and really pushed the chip.
 
I ran my X6 at 4Ghz (1090T) but it seemed pretty pointless to wind the rubber band up that much ... ok it was a def improvement for games with the 2X5850's in CF but the additional stress on the system was considerable.

Putting it on water might have been a bit better I guess.

Bit like overclocking the old Q6600 ... best done with asbestos gloves.

The Penryn cpu's didn't thermally ramp as quickly ... as are the SB / SBE's.
 
I ran my X6 at 4Ghz (1090T) but it seemed pretty pointless to wind the rubber band up that much ... ok it was a def improvement for games with the 2X5850's in CF but the additional stress on the system was considerable.

Putting it on water might have been a bit better I guess.

Bit like overclocking the old Q6600 ... best done with asbestos gloves.

The Penryn cpu's didn't thermally ramp as quickly ... as are the SB / SBE's.

try this
disable or downclock 2 cores of your cpu and then try to overclock, i think then you can achieve 4.5ghz without any issue of temp
(i am curious to know if it works or not [:truegenius:7] )
 
Status
Not open for further replies.