AMD's Piledriver And K10 CPU Architectures Face Off

[humbug]

Good old Tom's....

http://www.tomshardware.com/reviews/piledriver-k10-cpu-overclocking,3584.html

I found the results a little surprising, in that Vishera competes a little better to the old K10 than I thought.

Its about time someone did this, though I would have liked the Phenom II x6 and FX-8350 to have been in it.

 

[humbug]

I've always liked the 6350/6300 for the price it's a good cpu.
But i've never understood why the fx4s have always been strong money.

I agree, for £100 its an awesome CPU, in multi threaded work loads it blows the i3 out of the water while being just as fast in low threaded work with an overclock.

The FX-4/6 and 8 are all the same chip all be it with non functional cores disabled or just disabled depending on demand, just like the P-II x4 960T was actually a P-II x6 with dud cores or 2 cores disabled. I guess they can't go much lower than £100 to turn a meaningful profit on them, my guess is also that the FX-4 is not as mass produced as the FX-6 and 8 as its unlikely that more than 1 module is dud.

 

[humbug]

Still find it disgusting that FX cores are weaker clock for clock than my old Phenom II.

The fact that I'd have to run a 6350 @ 4.5Ghz to notice even a small improvement (less than 10%?) over my ancient Phenom II @ 3.8Ghz in stuff that isn't well multi threaded is a bit of a poor joke TBH.


Steamroller better deliver that reported 30% IPC improvement, that's all I'm saying...

I completely agree Vishera is not good enough full stop, but its not all that bad, the FX-4 @ stock is anything from 5% slower to 15% faster than the P-II x4 @ 4Ghz.
Now, when you consider that @ 4Ghz the P-II x4 is about as high as you can overclock it while the FX-4 still has 600 to 800Mhz left in the tank it is actually quite a lot better.

And, with the FX-6 only being £100 while 'overall' completely destroying the P-II x4 makes this even better.

The FX-6 is not an i5-K bar a few select things, but consider also @ £170 that not to far from twice the price.

As for PD, no one know what the performance increase will be, but i'm willing to bet its a solid 20% (per core / clock) if at that they are still cheaper than Intel they would be a dammed good CPU full stop.

 

[humbug]

Mature Phenom II stuff can reach 4.2-4.4 to be honest.

PD has a much more comprehensive instruction list.

I used to be able to get my 1090T to 4.3Ghz with a 3Ghz CPU-NB,- 8 hour Prime95 Stable. On air.
This is an early 2010 model, I don't know about the later 2012 models but I have never seen anyone match mine mine was a Gold pressed Platinum CPU incrusted with Diamonds.

Edit- I was offered quite a lot of money for it on Overclockers.com more than a year ago.

 

[humbug]

My 1055T did 4.375GHZ at 3.2GHZ CPU NB.
That was a 95W.

And the only reason I couldn't go further was it was sucking too much power and my quad rail couldn't handle it.

And my cooling.

Yeah that's the problem with multi Rail PSU's, if it has 20 Amps on one Rail and a power hungry device hocked upto that Rail your going to have limitation issues.

I was using an XFX 750 Watt PSU (Still have the same one) which has 62 Amps on a single 12v Rail, all I needed to do was bump up the Board Power Line to 130% and away she went....

"And my cooling"

Which was?

 

[humbug]

Well I was under a full custom loop, but at that type of heat output the cooling was reaching its peak.
I can't imagine a 4.3GHZ Thuban under air, not without throttling.

I don't have the overnight Prime results Screen Shot anymore so I can't prove it.

 

[humbug]

You'd need a load graph/core graph as well for the entire time.
I've left stuff priming and it's "passed" but when looking careful noticed it'd throttled.

You've already seen my Thuban results like 20 times

I believe you, I have no reason to doubt you.

Anyway, I'm not going to let what was a perfectly pleasant conversation end on a negative note, I do know what i'm doing, I have been at this since the 1990's, I know what happens and know what proof monitoring is needed.

I will agree to disagree, as it were.

 

[humbug]

@ Dave, those were the days, the P-II x4 960T would unlock to an x6 if the 2 disabled cores were good, I remember people all over the place giddy that they had spent £80 on an x4 960T and ended up with an x6 after just a few BIOS tweaks, Brilliant..... the only down side was, just like with the B55 unlock,- temperature monitoring was disabled.

Sadly AMD have become Killjoy's in that, all the FX chips have the inactive Modules lasered off so they can't be reactivated.

Boooo........

 

[humbug]

My 1100T was better than your 1090T wasn't it? I remember your Physics scores in 3D11 thread were almost 500 points lower, mine clocked to around 4.35Ghz 3.1Ghz NB. I sold that rig to my boss at work though! Should go round there and clock the nuts off it

Mine was a late 2010 model.

I think the 2010 models were defiantly better.

 

[humbug]

Is that on the same clock, or when it is overclocked higher?

The FX-4350 @ 4.2Ghz (stock) is 2% slower to 15% faster than a Phenom II x4 @ 4Ghz in gaming, clock for clock its probably going to be about 5% slower to 10% faster to the FX-4350, those are the 4 cores.

So for the 6 cores (x6 vs FX-6350) it will be the same, just that the FX-6350 runs at 3.9Ghz out of the box. so overall Piledriver is a little faster than Phenom II clock for clock / core for core.

Keep in mind also that 4Ghz is the best the x4 will do on air and about 4.2Ghz on water, while the Piledriver chips will do 4.6 on air and 4.8 on water, so in the end they are about ~15% faster.

I would think with 2 extra cores the FX-8350 would be (30% at stock and 40% overclocked vs overclocked) faster than the x6 with all cores loaded up.

Is it worth the FX-6350 if you have an x4? Yes, if you have an x6? it might be, people upgraded from a 2500K to a 3570K for less performance increase, an FX-8350? for £150 it would be if 8 cores are of use to you.

I would like to see a direct comparison: Piledriver vs Bulldozer, but from what I remember the FX-8150 needed all 8 just to keep up with my x6, I think Piledriver is quite a ways better.

 

[humbug]

Tom's Piledriver review included the 1100T;
http://www.tomshardware.com/reviews/fx-8350-vishera-review,3328.html

Thanks

Its difficult to tell.

Normalised: with Sandy Bridge, Phenom II, Bulldozer and Piledriver at the same clocks....

In Skyrim: if Phenom II is 100%, then Bulldozer is 92%, Piledriver is 98% and Sandy Bridge is 140%.

In WoW, which is a strange one as it appears to use 6 or more cores given that the 3.3Ghz x6 is slightly faster than the 3.7Ghz x4 and yet the 8 thread Ivy Bridge is no faster at all than the 4 thread one.

That one makes no sense at all.

I think Skyrim is a good measure of how these CPU's perform in these low threaded Floating Point DX9 games.

So I think what we can take from that is Bulldozer is the slowest Chip at about 92% performance, next is Piledriver at 98%, then Phenom II as the benchmark at 100% and Sand Bridge at 140%.

Keeping in mind that Piledriver clocks on average about 15% higher than Phenom II and ends up about that much faster.

I would like to see SR gain a solid 20% over Piledriver to put it at 120%, and then overclock to 4.8 / 5Ghz

 

[humbug]

At stock maybe but Phenom X6 at 4ghz I reckon it would be better for games and basically anything which doesn't make use of the new instruction sets (ie encoding).



The Phenom X6 is at stock though and they were very conservatively clocked, 2.8-3.3ghz with all cores in use, at 4ghz they were much better.

It just seems daft to me that they would do a comparison to the K10 architecture whilst ignoring the pinnacle of the K10 architecture, they don't even mention Thuban as if they never even existed.

What they show is the Phenom II chips are not faster, at 4Ghz the Phenom II CPU's perform only marginally better than the FX chips do at stock.