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

[HEXUS]

Along with other members of the Vishera family.

Read more.

[kingpotnoodle]

Clocked 10% faster, 6% faster at same clock speed - sounds pretty humdrum to me. A 3GHz Piledriver should be equivalent of a 3.4-3.5GHz 'dozer? And still burning 125W on the high end... still doesn't sound competitive without slashing margins pretty thin.

[scaryjim]

... A 3GHz Piledriver should be equivalent of a 3.4-3.5GHz 'dozer? And still burning 125W on the high end...

Erm, no. A 3GHz Piledriver is the equivalent of ~ a 3.2GHz 'dozer.

The high end Piledriver chips (125W TDP) will be clocked at 4GHz or higher, and be the performance equivalent of a 4.2GHz (or higher) 'dozer (i.e. ~ 14% performance enhancement).

If you want to consider lower down the scale, the 95W FX-8100 has a 2.8GHz base speed - an equivalent 95W Piledriver will be clocked at 3.1GHz (like the 125W FX-8120) but perform like a 3.3GHz 'dozer, with a 30W lower TDP.

Besides, within the same power envelopeactual silicon power draw can vary significantly - the TDP is only there as a cooling guide for quick ratification of existing third-party coolers. Trinity seems to have much better power characteristics than Llano, which bodes well for desktop Piledriver...

[billybear]

A 10% increase in power is nothing when Intel is beasting AMD in benchmarks by a significant margin. This means the CPU performance gap for AMD will only grow wider.

[Brewster0101]

A 10% increase in power is nothing when Intel is beasting AMD in benchmarks by a significant margin. This means the CPU performance gap for AMD will only grow wider.

Not interested in AMD beating Intel in the performance stakes - and I don't think AMD care either. What we want is a competitive CPU that can do everything fast. The fact Intel CPU's might be faster on benchmarks doesn't mean they are any faster in real world applications.

100 Frames per second or 150 FPS, who cares....

[J3FFW1SH]

A 10% increase in power is nothing when Intel is beasting AMD in benchmarks by a significant margin. This means the CPU performance gap for AMD will only grow wider.

Not interested in AMD beating Intel in the performance stakes - and I don't think AMD care either. What we want is a competitive CPU that can do everything fast. The fact Intel CPU's might be faster on benchmarks doesn't mean they are any faster in real world applications.

100 Frames per second or 150 FPS, who cares....

I have to agree, bought an FX-8120, clocked it @ 4Ghz and I have never seen it hit 100% could probably left it stock speeds and still been happy!

[Bagpuss]

So Pikeydriver CPU's will be approx 6% faster, clock for clock than Derpdozer..?

Now, given that Derpdozer CPU's are roughly 10-12% slower than Phenom II's, clock for clock..how is this still anything but a fail from AMD.?

I mean, an 8350 overclocked to 4.5Ghz which is probably what most of them will top out at, with non exotic cooling, will still only be 5%, maybe 10% if your lucky, faster than a 2yr old Phenom II x6 @4Ghz.

..that's just crap really.

[scaryjim]

Now, given that Derpdozer CPU's are roughly 10-12% slower than Phenom II's, clock for clock..how is this still anything but a fail from AMD.?

a) Bulldozer theoretical IPC was actually a shade higher than Phenom II, they just weren't able to make use of the IPC on tap. There's one particular test of theoretical INT throughput where the FX-4100 beats a same-clocked Phenom II X4. Float was a different matter, of course... The question is whether these improvements actually increase IPC, or just help Piledriver use the IPC on tap more effectively...

b) Because Piledriver can actually hit the clock speeds Bulldozer was intended to run at. Stock clocks for the high end 8-core parts are rumoured to be at or above 4GHz, and lower core-count parts could actually end up well above that. Fitting a 10% clock increase into the same TDP is pretty damn good you know...

[Jedibeeftrix]

i'd like to see a 95W eight-core piledriver.........

[Jimbo75]

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

[Bagpuss]

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

Which was already a whole 40-50% faster than Phenom II clock for clock....

[Jimbo75]

Which was already a whole 40-50% faster than Phenom II clock for clock....

Yes isn't it just amazing what an extra $40 billion a year revenue and a 2 year lead in process technology brings?

Do you people even realise what it's like for AMD being up against this behemoth who WHEN THEY LOST, BRIBED OEM'S BILLIONS IN ORDER TO AVOID USING AMD CHIPS ANYWAY?

Honestly wtf is wrong with you? AMD is doing a better job with extremely limited resources and if you can't see that you seriously need to educate yourself to the reality of the industry. Do you enjoy intel's penny pinching? 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 $?

Jesus.

[crossy]

i'd like to see a 95W eight-core piledriver.........

Which was already a whole 40-50% faster than Phenom II clock for clock....

95W octo-core would be on my shopping list too. Actually looking at the current AMD processors and - to my untutored eye at least - they appear pretty uncompetitive v's the 1155 stuff from Intel. In my case particularly the AMD TDP's seem pretty high for the relatively modest performance (if the b'mark's I've looked at are representative).

So at the moment if/when I decide to move off my current 1090T (which is serving me very well - despite some benchmarks painting it as performing not much better than a good Core2Duo) I think I'll be defecting to Intel for the first time in a long time, (been an AMD user since the days of the Duron).

[scaryjim]

Which was already a whole 40-50% faster than Phenom II clock for clock....

Questionable stat. At launch the differences between the two were very dependant on which test you were looking at.

Besides, how is that relevant to your point? You were complaining that Piledriver was only 6% higher IPC than Bulldozer, so the comparison against the improvement between two Intel generations is valid. No-one was claiming that Piledriver was going to pulp IB, just that they're a clear improvement over BD. Personally I'd take +10% clockspeed *and* IPC improvements at the same TDP: after all, IB only managed a +3% clockspeed improvement, and despite cutting the TDP by 18W IB actually draws almost exactly the same amount of power as SB, which shows you exactly how much you can read into TDP figures...

... Actually looking at the current AMD processors and - to my untutored eye at least - they appear pretty uncompetitive v's the 1155 stuff from Intel. In my case particularly the AMD TDP's seem pretty high for the relatively modest performance (if the b'mark's I've looked at are representative). ...

Bulldozer certainly had higher power draw than expected and lower performance - for a whole raft of reasons. Piledriver addresses a number of those concerns, but not all of them - some are down to architecture decisions and may never be fully resolved (the unusually low floating-point performance, for instance); others may need fixing in software (the Windows thread scheduler in Win 8 is meant to be much better optimised for Bulldozer/Piledriver, for instance). Power-draw is one that should improve massively - based on Trinity laptop reviews and desktop previews Piledriver is a much more efficient core.

Of course, 95W octo-core processors already exist - the original FX-8100 is such a beast. But it's low-clocked and under-performing: a 95W Phenom II X6 is currently a better multithreaded option.

[Jimbo75]

95W octo-core would be on my shopping list too. Actually looking at the current AMD processors and - to my untutored eye at least - they appear pretty uncompetitive v's the 1155 stuff from Intel. In my case particularly the AMD TDP's seem pretty high for the relatively modest performance (if the b'mark's I've looked at are representative).

So at the moment if/when I decide to move off my current 1090T (which is serving me very well - despite some benchmarks painting it as performing not much better than a good Core2Duo) I think I'll be defecting to Intel for the first time in a long time, (been an AMD user since the days of the Duron).

Depends on what you're going to be doing with it really. I bought a 2500K last month (yes I deliberately didn't buy Ivy) because I'm a gamer and I basically need the extra single threaded performance. If you're using it mostly multithreaded there really isn't much point in switching.

Piledriver will close the gap, but not by enough to count for gamers. If you are a gamer, might as well go ahead and buy that 2500K now.

[Bagpuss]

Which was already a whole 40-50% faster than Phenom II clock for clock....

Questionable stat. At launch the differences between the two were very dependant on which test you were looking at.

Not questionable at all for the stuff i use my PC for, and coming from a Phenom II @3.8ghz, to a 2500k at 3.6Ghz, it's 40% faster..

I ran AMD for 4yrs, no blind fan here, Bulldozer was a flop, and for the foreseeable future, any CPU based on the fundamental design, is nothing but lipstick on a pig...