Worth going back to AMD?

[hokies83]

Get a Used 2600k / Z68 board for about the same money and enjoy a cpu 25% faster then the Amd fail cpu they just released.. just new and alot of hype.. it is just on Avg 10% faster then BD....

The 2600k Should last you until Sky Lake if you do not like to upgrade all the time...

And will prolly still be 10% faster then Steamroller..

 

[BLaber]

The energy efficient part is the key. IB is something like 35% better than Sandy in the same metric.

Dont let marketing fool you(again).

GLFO 32nm sucks, clock mesh tech is good.

http://www.eteknix.com/reviews/processors/amd-fx-8350-am3-piledriver-processor-review/9/

 

[Idontcare]

AMD has never claimed steamroller to be 30% faster than piledriver or bulldozer on an ipc level. They did claim steamroller to be 15 to 30% more energy efficient compared bulldozer. During that slide they showed a redesign bulldozer core with a high design library (instead of being drawn by hand) that took 30% less size for the fpu unit (remember you still have the interger unit, interface, and l2/l3 cache which take up a huge part of the die.)

And even then amd uses the word up to 30% (15 to 30%) and they are comparing cpus from generation 1 (bulldozer) and generation 3 (steamroller) skipping the gains from generation 2 (piledriver) to make the numbers look better.

Image is here

Anand talks about this and other slides here
http://www.anandtech.com/show/6201/amd-details-its-3rd-gen-steamroller-architecture/2

I think THE message is getting lost because AMD conflated this information (the embedded slide above and its related technical details) with that of the release of information regarding steamdriver capabilities.

If we read the anandtech link provided above in which Anand discusses this information you will note he states the following at the end of that segment:

We won’t see these new libraries and automated designs in Steamroller, but rather its successor in 2014: Excavator.

In other words, even in the best-case-scenerio regarding the optimistic implementation of high-density libraries and so forth it won't be included in steamroller, it was all targeting excavator.

And on the topic of Excavator, AMD still kinda mentions it, but in a bad way as Excavator (and all of the bulldozer microarchitecture line) fall on the left-hand side of the following AMD "refocusing" slide - excavator does not have a place at AMD when it comes to AMD's focus on the right-hand side priorities:

^ you don't transition from the "From" side to the "To" side by continuing to invest money into products that fall under the umbrella of the "From" side...that money must be redirected into the product development of those products which fall on the "To" side of the slide.

And with layoffs and cost reductions happening in parallel, that makes even less resources available for AMD to enable their priorities on the "To" side. There simply is no room or resources at AMD to justify completing Excavator. I saw this happen at SUN when they cancelled Rock (I worked on Rock on the foundry side of that equation, they took it through to silicon in hand before cancelling it).

It would be insanity for Rory to externally argue the case that AMD is downsizing and refocusing their even more limited resources towards jaguar-based product derivatives only to then internally continue shoveling money at products on a now deprecated roadmap that includes excavator.

 

[Idontcare]

GLFO 32nm sucks, clock mesh tech is good.

http://www.eteknix.com/reviews/processors/amd-fx-8350-am3-piledriver-processor-review/9/

I don't see where that assertion is proven.

piesquared will tell you Glofo's 32nm process technology is superior to Intel's 22nm process technology, and its 28nm process tech will be superior to Intel's 14nm. Just ask him.

I believe the current leading theory is that GloFo's process tech is awesome, untouchable, and AMD's bulldozer microarchitecture is awesome, unbeatable innovation, but the world of hardware reviewers and software programmers have all banded together and conspired against AMD by cherry-picking applications to benchmark that favor Intel's microarchitecture while compiling all the applications with compilers that intentionally slow down the code if a non-Intel CPU is detected. :| In other news, the straw industry is projecting record revenue and growth owing to a new fad they dub to be the "grasping at straws" craze.

As for the clock mesh technology, AMD wrote a white-paper in which they found at best it reduced power-consumption by 5% provided you were using it at its optimal clockspeed.

Too low (<2.9GHz) or too high (>4.6GHz) and the resonant mesh actual caused clock power to exceed that of the traditional clock gate...which is why they have both resonant and classic clock gating technologies implemented on trinity/piledriver so the resonant can be disabled when the clockspeed is not in the optimal regime.

We deep-dived into the resonant clock-mesh stuff in quite a few threads here already, for pointers to the take home message sifted from the white-paper linked above I recommend reading this post and this post plus the other posts on the same topic on that thread.

 

[Abwx]

I think THE message is getting lost because AMD conflated this information (the embedded slide above and its related technical details) with that of the release of information regarding steamdriver capabilities.

If we read the anandtech link provided above in which Anand discusses this information you will note he states the following at the end of that segment:

^ you don't transition from the "From" side to the "To" side by continuing to invest money into products that fall under the umbrella of the "From" side...that money must be redirected into the product development of

Dont be fooled , for some reason Anand removed the relevant slide
from his article , nothing is lost , it s just that important information
is downplayed.

I put again the explicit mention by AMD of the expected perfs :

http://techreport.com/review/23485/amd-cto-reveals-first-steamroller-details

 

[Phynaz]

It say 30% more ops/cycle , i suppose that it s not just to
fill more cache memory and wait for the exe units to be saturated..

30% more compared to what? (hint- they don't say - it could be corn flakes)

Are these the same simulation tools that told them IPC of Bulldozer increased over Phenom?

 

[tonyfreak215]

I have a Core 2 Quad Q9550 clocked at 3.8ghz (Sig system) and it has easily been the best processor I have ever owned in terms of how long its remained viable and what a low voltage pain free OC it has given me.

However my first love is AMD, I'd been building AMD system up to the Core 2 Line and I've let Nehalem and Sandy Bridge pass by waiting for AMD to come up with something worth upgrading to.

Have they? While my processor is great, the P45 chipset it has to be coupled with is really showing its age and I'm starting to come around to thinking about a bottom up system rebuild. Would I see a reasonable performance boost going from my 3.8ghz Q9550 to an FX8350 (which I would undoubtedly overclock) and are AMD's Chipsets any good? If not, then I'll just have to accept that the AMD ship has sailed and hold out for Haswell.

Thoughts?

We are in the same boat.

I've decided to wait for Haswell.

Here is why:
Less power usage than Ivy
10+% faster
PCI-E 3
All ports will be SATA 6Gbps (rumored)
Trim in Raid 0
More USB 3 Ports

Plus with Windows 8 coming out; I've always found it to be a good idea to wait. There might be some new feature that involves new hardware to take advantage of.

Plus, since I've waited for so long, why not wait a little more? What's 4-5 more months?

 

[hokies83]

We are in the same boat.

I've decided to wait for Haswell.

Here is why:
Less power usage than Ivy
10+% faster
PCI-E 3
All ports will be SATA 6Gbps (rumored)
Trim in Raid 0
More USB 3 Ports

Plus with Windows 8 coming out; I've always found it to be a good idea to wait. There might be some new feature that involves new hardware to take advantage of.

Plus, since I've waited for so long, why not wait a little more? What's 4-5 more months?

I have a De lidded lapped 3770k and i will sell it and get Haswell to..

Simple fact is my Cpu trade off price wise will not be to much 50 - 75$ i can sell my MB for 200$ and almost get 1150 Socket from my guy for free.

 

[pauldun170]

I have a Core 2 Quad Q9550 clocked at 3.8ghz (Sig system) and it has easily been the best processor I have ever owned in terms of how long its remained viable and what a low voltage pain free OC it has given me.

However my first love is AMD, I'd been building AMD system up to the Core 2 Line and I've let Nehalem and Sandy Bridge pass by waiting for AMD to come up with something worth upgrading to.

Have they? While my processor is great, the P45 chipset it has to be coupled with is really showing its age and I'm starting to come around to thinking about a bottom up system rebuild. Would I see a reasonable performance boost going from my 3.8ghz Q9550 to an FX8350 (which I would undoubtedly overclock) and are AMD's Chipsets any good? If not, then I'll just have to accept that the AMD ship has sailed and hold out for Haswell.

Thoughts?

I have a Q9550+ ep45-udp3.
I moved up to 2700K z77 setup.

Looking at the stats on how the FX compares to Ivy Bridge and knowing how Ivy Bridge compares to my 2700K AND knowing how my 2700K compares to my Q9550 set-up....

Wait for haswell. Save your money next summer.

If you aren't choking on anything now hold on to the Q9550.
Even though the chipset is old, it still rocks with current SSD's hooked up on the Sata 3gb/s ports. (In my case a Samsung 830 256gb)

The FX8350 seems like a nice proc but unless you are looking for some specific performance targets that your Q9550 isn't hitting I'd stick with the OC Q9550. Still useful life in it

 

[Abwx]

30% more compared to what? (hint- they don't say - it could be corn flakes)

It was released at Hotchips if you did take the time to read TR article
rather than being invariably caught by your thead trashing habits....

So we can safely conclude that it s in comparison of BD 1....

Are these the same simulation tools that told them IPC of Bulldozer increased over Phenom?

You are never tired of spreading your usual and worthless junk.??..

 

[Phynaz]

You are never tired of spreading your usual and worthless junk.??..

You didn't answer the question.

Why do you hold AMD marketing slides as Gods' truth?

 

[T_Yamamoto]

You didn't answer the question.

Why do you hold AMD marketing slides as Gods' truth?

Because he's an AMD fanboy thinking that steamroller will somehow beat Intel

 

[Abwx]

You didn't answer the question.

Why do you hold AMD marketing slides as Gods' truth?

You asked what the comparison was done with and i answered that it was
in respect to BD1....

Voluntarly deaf/blind but not mute ?...

 

[Abwx]

Because he's an AMD fanboy thinking that steamroller will somehow beat Intel

An exemple of straw man.....
No valuable argument so thrashing a thread all the way is all what
is left for half brains...

 

[Phynaz]

You asked what the comparison was done with and i answered that it was
in respect to BD1....

Voluntarly deaf/blind but not mute ?...

I also asked if they used the same simulation tool as their BD estimates, but instead you chose to denigrate my post.

And now you failed to answer why you believe in AMD's marketing slides, and again resorted to a personal attack rather than answering the question.

So you state we should believe what AMD says about Steamroller, but refuse to state why. Others would say that calls the credibility of your posts into question.

See my sig why you don't believe marketing...Ever. From anyone.

 

[AtenRa]

I have a Core 2 Quad Q9550 clocked at 3.8ghz (Sig system) and it has easily been the best processor I have ever owned in terms of how long its remained viable and what a low voltage pain free OC it has given me.

However my first love is AMD, I'd been building AMD system up to the Core 2 Line and I've let Nehalem and Sandy Bridge pass by waiting for AMD to come up with something worth upgrading to.

Have they? While my processor is great, the P45 chipset it has to be coupled with is really showing its age and I'm starting to come around to thinking about a bottom up system rebuild. Would I see a reasonable performance boost going from my 3.8ghz Q9550 to an FX8350 (which I would undoubtedly overclock) and are AMD's Chipsets any good? If not, then I'll just have to accept that the AMD ship has sailed and hold out for Haswell.

Thoughts?

What you do with your PC ???

FX8320 is way faster in multithreaded apps than Q9550 and it has more performance even in single thread and more new games starting using more than two threads.

I would get the 8320 and OC to 4.5GHz or more.

 

[Abwx]

I also asked if they used the same simulation tool as their BD estimates, but instead you chose to denigrate my post.

It s not denigrating but it should be obvious to you that
one cant honnestly answer valuably such a question unless he works
at AMD and is willing to disseminate inside knowledge....

And now you failed to answer why you believe in AMD's marketing slides, and again resorted to a personal attack rather than answering the question.

So you state we should believe what AMD says about Steamroller, but refuse to state why. Others would say that calls the credibility of your posts into question.

See my sig why you don't believe marketing...Ever. From anyone.

I didnt say that i believe whatever you re pretending , if you did read the post to wich you did answer you ll notice that i used this slide as an answer to another poster that claimed that AMD said nothing about the expected IPC improvements of Steamroller and went quoting a twisted ANAND review from Hotchips that , curiously , didnt mention this picture...

Now , as to know if the 30% figure will be accurate , that s another
story but considering their current financial situation they better
hit the target if they want to survive , as the low power SoC strategy
cant be implemented in one or two years and the traditionnal CPU/APU
will keep on being their first money provider for quite a long time.

 

[mrmt]

Dont be fooled , for some reason Anand removed the relevant slide
from his article , nothing is lost , it s just that important information
is downplayed.

I think that it is not important information downplaying, but filtering AMD marketing bullshit. It is the same company that hyped bulldozer to the death for years, the same company that was promising that Trinity would solve all its problems, etc.

I would take every performance prediction by AMD with a mountain of salt, but some people never learn.

 

[Phynaz]

Now , as to know if the 30% figure will be accurate , that s another
story but considering their current financial situation they better
hit the target if they want to survive , as the low power SoC strategy
cant be implemented in one or two years and the traditionnal CPU/APU
will keep on being their first money provider for quite a long time.

Ok, so you admit the 30% figure should not be taken as a statement of fact. Wasn't that easy?

 

[tweakboy]

I have a Core 2 Quad Q9550 clocked at 3.8ghz (Sig system) and it has easily been the best processor I have ever owned in terms of how long its remained viable and what a low voltage pain free OC it has given me.

However my first love is AMD, I'd been building AMD system up to the Core 2 Line and I've let Nehalem and Sandy Bridge pass by waiting for AMD to come up with something worth upgrading to.

Have they? While my processor is great, the P45 chipset it has to be coupled with is really showing its age and I'm starting to come around to thinking about a bottom up system rebuild. Would I see a reasonable performance boost going from my 3.8ghz Q9550 to an FX8350 (which I would undoubtedly overclock) and are AMD's Chipsets any good? If not, then I'll just have to accept that the AMD ship has sailed and hold out for Haswell.

Thoughts?

FX 8300 the top model of Vishera is 8 cores 199 dollars. But 8 slower much slower cores then Sandy or Ivy or Haswell. You don't need 8 core unless you do video editing or autocad, or studiomax or maya or run a DAW

Sandy OCed still slaps the sh*T out of this 199 dollar 8 core processor.

D:

 

[Abwx]

Ok, so you admit the 30% figure should not be taken as a statement of fact. Wasn't that easy?

It can be false but it can also be true , while you re insisting
on the first possibility.

The thing is to discuss of the possibilities without thoses prejudices
that you are always prone to exibit , and this require a little more
than the average basher INTELligence....

 

[Abwx]

I think that it is not important information downplaying, but filtering AMD marketing bullshit. It is the same company that hyped bulldozer to the death for years, the same company that was promising that Trinity would solve all its problems, etc.

I would take every performance prediction by AMD with a mountain of salt, but some people never learn.

Curious that other site didnt "filter" this slide in their articles
on the same subject...

One never "filter" without purpose but still , you put the charge
on AMD rather than see the obvious , that ANAND is not as professionnal
as it pretend to be , or rather , yes , they are....:biggrin:

 

[T_Yamamoto]

An exemple of straw man.....
No valuable argument so thrashing a thread all the way is all what
is left for half brains...

No valid argument? Faildozer ews a bust. Thank fully pile driver is somewhat better but in all honestly it still sucks compared to sandy and ivy

 

[Phynaz]

It can be false but it can also be true , while you re insisting
on the first possibility.

The thing is to discuss of the possibilities without thoses prejudices
that you are always prone to exibit , and this require a little more
than the average basher INTELligence....

Oh, look you're now trying to make silly puns.

I don't get why you are still failing to read my posts. I pay no attention to marketing slides from any company. If you were were to post Intel slides I would also say they are mostly calculated BS.

 

[GodisanAtheist]

I have a Q9550+ ep45-udp3.
I moved up to 2700K z77 setup.

Looking at the stats on how the FX compares to Ivy Bridge and knowing how Ivy Bridge compares to my 2700K AND knowing how my 2700K compares to my Q9550 set-up....

Wait for haswell. Save your money next summer.

If you aren't choking on anything now hold on to the Q9550.
Even though the chipset is old, it still rocks with current SSD's hooked up on the Sata 3gb/s ports. (In my case a Samsung 830 256gb)

The FX8350 seems like a nice proc but unless you are looking for some specific performance targets that your Q9550 isn't hitting I'd stick with the OC Q9550. Still useful life in it

Well looks like that settles it.

Honestly nowadays I like TALKING about computers more than building them. Get married and have a kid and suddenly you realize how incredibly solitary computers are. Will probably begin my slow slide back into console gaming but I still want to have a solid gaming/office/media rig *JUST IN CASE* I somehow find the time and the alone-ness. I'm starting to get that 3 year upgrade itch REAL BAD and I have a lump of spare severance money going from my last job to my new one...

I've waited this long, might as well treat myself and get a nice 30" monitor and HD7950 and see how long this core set-up lasts me.

Thanks for the responses... and the crazy ranting that happened in between!