News - AMD FX-8350 Piledriver-based CPU to land mid-Q3

[crossy]

Hot Ivy Bridge cpu's that barely improve on Sandy, but sure help intel to save up even more $billions? "K" series chips, turbo, hyperthreading - all these things on the same bit of silicon but laser cut off so you can't use them unless you pay more $?

Strange to hear you accuse IB of being "hot" - I was looking at IB and SB parts at the weekend, and I got the impression that the newer IB ones, while being a s**tload more expensive (shame on you Intel) also had lower TDP's.

E.g. 2700K SB v's 3770K IB - both 3.5gig parts with 8MB caches, but the 2700K is 95W TDP whereas the 3770K is "only" 77W.

And to join you in dissing Intel - how the heck can they justify such a horribly complex product offering? Looking at Scan, eBuyer etc, you see we've got four different sockets (1155, 1156, 1366 and 2100; three different major families (i3, i5, and i7); and then you've got "plain", "K" and "S" versions of some of those processors. It's almost as if they've decided to go with "whatever your budget, we've got a processor to fit".

[Jimbo75]

They can't change the design, they are stuck with it for the next 3-4 years at least. Same as intel was stuck with the old Pentium architecture when Athlon was all-conquering. All they can do is improve it by more than what intel improves by, and with them setting such a low target with Ivy Bridge AMD will close that gap without breaking sweat.

The gap is closing, no it's not closing enough for gamers but for anything else the top Piledriver should be a good match for the 3770K.

[scaryjim]

Not questionable at all for the stuff i use my PC for ...

Unless every computer user has the same use profile as you, then it's questionable I've looked back at the review benchmarks: there are tests where i5 2500k is twice as fast as Phenom II, there are tests where it's barely any faster at all. There are conceivable workloads where the difference would be negligible, and there are conceivable workloads where it'd be night-and-day.

Of course, for the vast majority of computer users these processors are already vastly more powerful than they need, which is why AMD has spent a lot more time focussing on APUs, and it's pretty clear from the Trinity previews that it's paying off for them. In the discreet CPU area they're much more interested in the server and HPC market, and there they're pushing the combination of CPU and GPU for heterogenous systems. Basically AMD has become a platform company, *not* a CPU company. Somewhat sucky for CPU enthusiasts, sure, but given the lead Intel managed to grab with Core 2, the rather disappointing Phenom launch (monolithic quad-core really failing to work for AMD), and the poorly optimised BUlldozer CPUs they rushed out to have something new in the consumer space last year, it's hardly surprising that they've started focusing efforts where they have a genuine competitive advantage over Intel.

[Hicks12]

@ crossy, i believe the hot part is aimed at intel using thermal paste instead of the tried and proven solder, its why that even though they may use slightly less power they are actually like 10 - 15c hotter than SB as it cant transfer the heat efficiently, thermal paste is a horrible effciency and its only used because its better than air but solder is like over 10x more effective in exchanging heat so your heatsinks will work better on SB than IB hence the hot term.

I dont see how bulldozer was a flop, it wasnt competitive to intels offerings but its a strong foundation. AMD has designed a chip that is the future but its a little early, hence why its not that great yet as developers arent taking advantage of it. A good example is Win7 vs Win8 performance, apparently its like 10% better on Win8 due to how it handles cores, it works much more effectively and makes use of AMDs design.

Intel is strong but its also expensive, AMD have good value and the motherboards are much cheaper! If you just want a "cheap" system like a HTPC or a low end pc then AMD should have the money as their trinity chips are excellent. We are in a time where the GPUs are slowly becoming the norm in terms of doing work, AMD know that floating point isnt an issue for the cpu, the graphics cards are the new FP power houses and most work will be done on them in the future, it works very well for them as their graphics cards are complete computing beasts so it takes the slack.

more cores is better for cpu dependant calculations as long as the software is coded properly and this is what i mean when AMD designed a chip to early, we are only a few years into GPGPU and its working extremely well for the mobile computing sector (laptops/mobiles etc) but we arent at a point where the desktop/high end computing can make proper use of these designs.

Maybe in a few years when steam roller is out we will be, but for now i think AMD is on the right track... the improvements over BD is a big one especially if you compare it to IB vs SB and that pile driver doesnt have the advantage of a smaller process, if that happened we would be seeing even stronger performance or simply a TDP reduction as well as the extra 10 - 15%.

The futures bright for AMD, i feel as if it will be a repeat of the core2duo but reversed (AMD being the winners). Hopefully, intel need to be brought back to competitive pricing and AMD needs the money!.


Edit: Missed jims post while i was typing , basically what he said! And i meant APU .

[KeyboardDemon]

I'm really looking forward to seeing the reviews and benchmark results as soon as they are available, I don't mind admitting that I felt let down by the Bulldozer performance results and stuck with my 965be at the time. I hope that the FX8350 will give better performance results so I can go for the upgrade, I'm not really that bothered if it doesn't beat the Intel CPUs though.

[KeyboardDemon]

Deleted due to double post. Sorry

[King Jay]

Bulldozer was such a let down, can't believe how much AMD hyped it up. Hopefully the Piledriver will give Intel much better competition, not holding my breath though.

[CAT-THE-FIFTH]

It seems the Hexus article seems to have "forgotten" a tiny website called Toms Hardware has already tested Piledriver based hardware:

http://www.tomshardware.co.uk/a10-58...w-32463-2.html

Toms Hardware did two clock for clock tests with the A10-5800K and FX8350,with Turbo and power saving features disabled.





The first test is single threaded. In the second,half the modules of the FX8150 were disabled and is run using 4 threads,meaning the CPU is more like an FX4100 series CPU which has slightly better IPC than the FX8100 series,due to more available L3 cache per core.

The IPC improvements were more akin to around 15% and the A10 lacks L3 cache,which Vishera has.

On top of this the Piledriver cores in Vishera will additional tweaks too when compared to the ones in Trinity.

Both Trinity and probably Vishera will run at higher clockspeeds than Bulldozer too,and have a more refined version of Turbo core too. Hence considering AMD are on the same node,a 20% to 30% improvement in both single thread and multi-threaded performance with Vishera does not seem far off.

The Piledriver chitchat thread has most of the latest information and rumours in it.

[KeyboardDemon]

Both Trinity and probably Vishera will run at higher clockspeeds than Bulldozer too,and have a more refined version of Turbo core too. Hence considering AMD are on the same node,a 20% to 30% improvement in both single thread and multi-threaded performance with Vishera does not seem far off.

That would be awesome, even 10% would make me happy, my 965be out performs the FX8150 in some of the benchmarks here, what I really want is for the FX8350 to feel like a significant upgrade, an upgrade that I don't need, it's just me wanting some bragging rights.

[Kumagoro]

Problem is haswell wont be far off once piledrivers are actually out and I think a lot of core 2 era peeps will jump up to that.

Amd got stuffed taking intels small pay off. Its tactics in the p4 days really worked out well for them. They should have been made to pay at least 10 if not 20 billion.

[CAT-THE-FIFTH]

Problem is haswell wont be far off once piledrivers are actually out and I think a lot of core 2 era peeps will jump up to that.

Amd got stuffed taking intels small pay off. Its tactics in the p4 days really worked out well for them. They should have been made to pay at least 10 if not 20 billion.

TBH,Haswell is going to out closer to mid 2013,so Vishera will probably have at least six months,before that hits. Also,Haswell,is not going to be a massive jump in CPU performance IMHO,and the focus will be on further CPU TDP reduction,as it will mean more TDP for the IGP. Haswell is going to be more about the IGP. AMD Kaveri,will have GPU section which appears to have the same theoretical processing power as an HD7750.

Regarding the settlement,I think AMD could have got more,but then Intel released them from their need to have a fab if they wanted to make X86 CPUs. This probably was the main thing they were aiming for IMHO.

[badass]

It seems the Hexus article seems to have "forgotten" a tiny website called Toms Hardware has already tested Piledriver based hardware:

http://www.tomshardware.co.uk/a10-58...w-32463-2.html

Toms Hardware did two clock for clock tests with the A10-5800K and FX8350,with Turbo and power saving features disabled.

The first test is single threaded. In the second,half the modules of the FX8150 were disabled and is run using 4 threads,meaning the CPU is more like an FX4100 series CPU which has slightly better IPC than the FX8100 series,due to more available L3 cache per core.

The IPC improvements were more akin to around 15% and the A10 lacks L3 cache,which Vishera has.

On top of this the Piledriver cores in Vishera will additional tweaks too when compared to the ones in Trinity.

Both Trinity and probably Vishera will run at higher clockspeeds than Bulldozer too,and have a more refined version of Turbo core too. Hence considering AMD are on the same node,a 20% to 30% improvement in both single thread and multi-threaded performance with Vishera does not seem far off.

The Piledriver chitchat thread has most of the latest information and rumours in it.

That's only 2 tests. C'mon you can't extrapolate that to be an average!

I only see the AM3+ Piledriver CPU's useful as upgrades to existing AMD systems TBH. Outside of that, I'd probably go for the greater single (and in general bar a few specific exceptions, multithreaded) performance of a current Intel CPU.

However it's a good improvement over Bulldozer.

The Piledriver CPU's to be interested in IMHO are Trinity.

[CAT-THE-FIFTH]

That's only 2 tests. C'mon you can't extrapolate that to be an average!

I only see the AM3+ Piledriver CPU's useful as upgrades to existing AMD systems TBH. Outside of that, I'd probably go for the greater single (and in general bar a few specific exceptions, multithreaded) performance of a current Intel CPU.

However it's a good improvement over Bulldozer.

The Piledriver CPU's to be interested in IMHO are Trinity.

Oh,wait so lets go and say 3% then,Intel wins!!

You fail to realise,iTunes and 3ds Max are some of the applications which don't do well on the FX series - a Phenom II X6 1100T beats an FX8150 in both cases. They are among the worst cases scenarios for the FX8150 and among the best case scenarios for Intel CPUs. This is why Toms Hardware tested them and they are very different types of software.

You also seem to not consider a few other things:
1.)Vishera uses slightly different Piledriver cores to Trinity
2.)Vishera has L3 cache
3.)Trinity is TDP constrained

Hence,I don't see between 20% to 30% improvement in performance on average as not being a reasonable figure. Even if it was a 20% figure,the higher default clockspeeds alone would be responsible for a reasonable chunk of that!

It will be interesting to see how fast the A10 is with a discrete graphics card. Many games like L3 cache. I still think Intel will be better for gaming with expensive cards,but the competitive position should be somewhat better with Vishera against IB than the Phenom II was against SB.

[Biscuit]

Maybe it's time to remind people that Ivy Bridge was a whole 6% faster than Sandy Bridge?

Perhaps i should remind you that IB was a 'tick' in their famous tick-tock cycle, therefore it wasn't meant to have massive increases in performance, it was more about the die shrink. Other examples are Core vs Penryn and Nehalem vs Westmere. Its normal practice for Intel to do this and it works quite well.
The major difference with IB was the improved on chip graphics, although i doubt most of us give a crap about that.

If there is one thing i have learned about AMD, its that you should wait and see untill the product is officially out until you make a judgement. This release is set to be a pretty good improvement over BD, however it still wont be hugely competetive with intel in the consumer space and will probably face massive geek rage as usual.

ATM the bulldozer architechture is a chip with a certain place, it does some things extremely well and if YOU do those things then its worth it, for most of us however the Intel side of things is the better way to go. I reckon if its half decent, i will still support AMD... just because. As a few people have mentioned above in so many words, its not like your genuinely going to notice the difference day to day.

Main thing i hope for is some good micro and mini enthusiast boards, as the last generation seemed to miss that out

[jim]

Of course, for the vast majority of computer users these processors are already vastly more powerful than they need, which is why AMD has spent a lot more time focussing on APUs, and it's pretty clear from the Trinity previews that it's paying off for them.

Indeed. It's quite funny how these days there's still a massive push for top-notch gaming CPUs that stretch ever-further into the stratosphere of performance, but everybody equally knows that CPU power is fairly negligible alongside GPU power for gaming. The only time I push my CPU is when I very occasionally render video, i.e. once a year, and whether it takes 30 minutes or 50 minutes is neither here nor there - it's still a long time. No doubt there are professional renderers and so on where faster performance does equal money, but they can't be a very large proportion of the market.

The number of people with laptops who want a bit of light to very light gaming, 1080p, and as much battery life as possible by comparison is huge.

What AMD has been doing with the APUs is fantastic to my mind, and I think if I was building a new desktop any time soon that wasn't explicitly for gaming I'd be going straight for an AMD solution.

[badass]

Oh,wait so lets go and say 3% then,Intel wins!!

Enough of the militancy!
Seriously. Enough!

You fail to realise,iTunes and 3ds Max are some of the applications which don't do well on the FX series - a Phenom II X6 1100T beats an FX8150 in both cases. They are among the worst cases scenarios for the FX8150 and among the best case scenarios for Intel CPUs. This is why Toms Hardware tested them and they are very different types of software.

You also seem to not consider a few other things:
1.)Vishera uses slightly different Piledriver cores to Trinity
2.)Vishera has L3 cache
3.)Trinity is TDP constrained

Hence,I don't see between 20% to 30% improvement in performance on average as not being a reasonable figure. Even if it was a 20% figure,the higher default clockspeeds alone would be responsible for a reasonable chunk of that!

It will be interesting to see how fast the A10 is with a discrete graphics card. Many games like L3 cache. I still think Intel will be better for gaming with expensive cards,but the competitive position should be somewhat better with Vishera against IB than the Phenom II was against SB.

My reasoning for my previous statement is that for workloads where CPU power really does matter, cost is less of an issue. If you have a workload that can saturate 8 threads, it can likely saturate more. Get a workstation/server type setup with CPU's with even more throughput to get the job done quicker.
Anything that cannot saturate 8 threads is not benefiting from 8 cores. Why not save a packet and go for a better all round CPU/APU, such as an A10 with cheaper motherboards to boot!
If you really do need single/lightly threaded performance, a Core i5 SB or IB CPU will be substantially faster.

Secondly, 4 GHz is 11% faster than 3.6 GHz. Expecting 20 to 30% average performance improvement across the board for the FX-8350 is optimistic at best, downright delusional at worst. The single core turbo maximum frequency has not changed either.
For a 20% average performance improvement, that requires over an 8% IPC improvement. Optimistic I think. Optimistic but possible.
For a 30% improvement, a 17% IPC increase is required. Anyone who at this point in time believes that is either a little too in love with AMD or has been listening to someone who is a little too in love with AMD.