AMD "Piledriver" refresh of Zambezi - info, speculations, test, fans

[r.p]

7.77 vs 7.42 points is only a very small improvement. If this new gen CPU can't do at least 9 points it's hardly worth upgrading imho.

hmmm ... cinebench 11.5 is a little bit like super-pi, it has no real relevance anymore.
... but if you look at the 3dsmax/vray results, it's an improvement of ~17% ...

[SUPERKAMES]

Hope I didn't come off harsh... It was pretty late.

When does the NDA lift? Any chance you could share some x264 FHD benchmarks?
http://downloads.guru3d.com/x264-FHD...load-2825.html

If you can't I understand, don't want to get you in trouble...

Thanks for sharing!

@4.5ghz I saw some thing like this



@All : I'm not PR for AMD ,I'm an architect ,and just want to show you something anh I wish that you can test them yourself .
For more 3sdmax/Vay or mentalray results:http://www.behardware.com/articles/8...es-on-am3.html
Maybe after NDA , they can test that for you again . I have no benefit to show the wrong things.

[aGeoM]

Hi guys,

I just finished this Cinebench runs.


CINEBENCH R11.5

[CPU] - Cinebench R11.5 - AMD 8120FX@4500 - 7.39.jpg


HYBRID CINEBENCH R14/11.5

[CPU] - Cinebench R14 - AMD 8120FX@4500 - 7.77.jpg


Edit


CINEBENCH R11.5 @4890

[CPU] - Cinebench R11.5 - AMD 8120FX@4900 - 7.96.jpg


HYBRID CINEBENCH R14/11.5 @4890

[CPU] - Cinebench R14 - AMD 8120FX@4900 - 8.46.jpg


Hope this helps somehow.

[Nintendork]

It's pretty obvious than we'll the same improvements already shown with trinity + some extra from the l3 cache.

Zeus
When intel is the de-facto standard(from Core2 to IB it's a predictable and evolutionary path with strengths different from AMD's) and the competition releases a new arch with some nice instructions one would expect than software that take advantage of them will show what the new cpu has to offer.

[Zeus]

Fully agreed mate but that doesn't mean they can slacken off on older instruction sets. Intel's strength seems to be that it can do both where AMD only wants you to see how good benches with new instructions run to hide their weak(er) side.

It would be the same thing as to say how well it runs on Windows 8 while most still run Windows 7 for example.

I don't mean to turn this into an AMD vs Intel fight as we've had enough of that in the past but it's an observation that cannot be ignored.

[FlanK3r]

I agree with thinks, the newest software is better optimized (R9 Cinebench to R10 and to R11.5 now or new R13). But we can not say, the R11.5 is very old or archaic. I think, performance of FX in Cinebench R11.5 is not bad. Is it about 80% 4c/8T Sandy Bridge 2600K. We can to say, superpi is archaic benchmark for modern CPUs. The main problem is x87 and FX, there is different story in architecture than with Phenoms II, doesnt mean PII is stronger in single performance. Really not. Clock for clock maybe yes a bit, with stock speeds is single performance ratio of FX better than PIIs.
Otherwise I agree with Zeus, Intel CPU are strong everywhere, single and multi. AMD only in multi (ratio multi scaling is better at AMD than at Intel!). The way for AMD is making Steamroller much, much better in FP point and single thread performance or let classic Bulldozer uarch back and focus at something new (APU is good too, but the problem will be still software support for iGPU acceleration)

[Daveburt714]

@4.5ghz I saw some thing like this

An encoding rate of >2fps is a nice improvement on an app that I (and many other people) actually use.

After seeing this, the 5ghz OC on a little closed loop water settup, and the aggressive pricing they've got my money!

Eh, who am I kidding. They had my money anyway...
I'm an AMD junkie!

Thanks for running that for me!!

[tbone8ty]

I'm hoping for 5ghz on my nzxt havic 140...stable in prime95 for 30 min

If so I buy

[Zeus]

Dave, i'm really looking forward to see how will do on this new platform. I promise that if it beats my current set up i will return to the AMD camp. But to be honest i doubt it very much.

Agree with you Flank that in multithreaded benches Bulldozer didn't do too bad but single threaded was so bad it almost made me cry. It's been the first time was unable to beat my SuperPi 1M result from my former set up.

With my current set up i have already done 1M at 6.848 sec and the cold weather is yet to come. When will we see that on AMD you think? Best 1M with my Bulldozer was low 16 seconds and 14.xxx on PII so it probably is a long way.

Prices for the new AMD cpu's (or are they apu's?) don't seem too bad though.

[Manicdan]

i would never use superpi as the comparison for if i should get a cpu

i look at tasks that i actually do, like gaming and rendering, and just a bunch of 2d tasks
so i care about 3-4 threaded perf, idle power, and total multi threaded perf that can be gained on safe 24/7 volts under water. which is why i went with a 2500k. i think for amd to be worth it for me, they need to up their low thread perf by 20% or more. and as far as value goes, i do consider that. so if amd was 10% worse but 30% cheaper, i probably would pick it up. so thats my thought process for purchases. i just wanted to point that out since it would be pointless to me to look at superpi for any reason other than extreme benchmarking (where its more about fun rather than function and price)

[sequoia464]

An encoding rate of >2fps is a nice improvement on an app that I (and many other people) actually use.

After seeing this, the 5ghz OC on a little closed loop water settup, and the aggressive pricing they've got my money!

The pricing that I have seen is $253 for the 8350 and $242 for the 8320 - seems high to me, am I missing something? (unless there is a substantial improvement in performance or power usage)

The reason I say this is that a 2500K can be had for $160 and a 3570 for $190.

I'm using an AMD board now with a Thuban (also have a 6100 that lives in a drawer now) and would love to stay with this board because of the 6 Sata III ports.

[Daveburt714]

The pricing that I have seen is $253 for the 8350 and $242 for the 8320 - seems high to me, am I missing something? (unless there is a substantial improvement in performance or power usage)

The only pricing I've seen so far was from BLT preorder. They've had a history of being ~20% higher with early listings...
Time will tell I quess, but like I said, I'm a junkie so what's $50.

[Oese]

7.77 vs 7.42 points is only a very small improvement. If this new gen CPU can't do at least 9 points it's hardly worth upgrading imho.

4.5%

[Darakian]

For comparison what does a 1100T score in Cinebench 11.5?

[sdlvx]

7.77 vs 7.42 points is only a very small improvement. If this new gen CPU can't do at least 9 points it's hardly worth upgrading imho.

About 15% improvement in V-Ray from what SUPERKAMES posted. That was enough to take AMD from being behind 2600k by around 10% to ahead by about 7%. And that's at the same clocks. We are looking at FX 8350 being 7% faster at rendering at the same clocks than 2600k and it hitting higher clocks. If we see 5Ghz on air on FX 8350 and IB doesn't reach that high, FX 8350 may end up being the fastest CPU in this price range for rendering.

Cinebench r11.5 is old. Rendering software is *constantly* changing and improving. To insist we use an older piece of software is ridiculous. That's like saying if we have two pieces of software that does the same thing, and one uses a better compiler and is faster that it doesn't matter. It matters greatly and if someone renders on FX 8350 in Linux (say Gentoo and they compile with latest GCC with flags to use FX 8350 instructions) it's going to mop the floor with a 2600k.

It is the same with transcoding software. Who is going to use old software?

This is why open source is so great and I don't think it's a huge coincidence that Linux for steam is going to roll out about the same time as Vishera. This will probably end up the same as it always does with Intel fanboys. AMD brings something new to the table and the Intel fanboys say it's useless, then Intel gets it and it's awesome. Just like when we had Netburst vs K8 (I'm talking about clocks), IMC vs FSB, x86_64

[Zeus]

About 15% improvement in V-Ray from what SUPERKAMES posted. That was enough to take AMD from being behind 2600k by around 10% to ahead by about 7%. And that's at the same clocks. We are looking at FX 8350 being 7% faster at rendering at the same clocks than 2600k and it hitting higher clocks. If we see 5Ghz on air on FX 8350 and IB doesn't reach that high, FX 8350 may end up being the fastest CPU in this price range for rendering.

Cinebench r11.5 is old. Rendering software is *constantly* changing and improving. To insist we use an older piece of software is ridiculous. That's like saying if we have two pieces of software that does the same thing, and one uses a better compiler and is faster that it doesn't matter. It matters greatly and if someone renders on FX 8350 in Linux (say Gentoo and they compile with latest GCC with flags to use FX 8350 instructions) it's going to mop the floor with a 2600k.

It is the same with transcoding software. Who is going to use old software?

This is why open source is so great and I don't think it's a huge coincidence that Linux for steam is going to roll out about the same time as Vishera. This will probably end up the same as it always does with Intel fanboys. AMD brings something new to the table and the Intel fanboys say it's useless, then Intel gets it and it's awesome. Just like when we had Netburst vs K8 (I'm talking about clocks), IMC vs FSB, x86_64

I don't mean to turn this into an AMD vs Intel fight as we've had enough of that in the past .....

Now you're the one turning this into an AMD vs Intel fight.

First off, you are reading selective. I'm not insisting we should use "older" software, all i'm saying is that current hardware should at least support it. Intel does, so why AMD not? I have a hard time calling CB 11.5 old sofware as it has only been released what? Two years ago?

Second, you're comparing an old generation Intel CPU to an AMD one that hasn't even been released yet. Doesn't make sense to me.

Third, you are assuming FX8350 will do 5G on air, how can you be so sure it will do that as i'm still yet to see the first 5G Prime/LinX stable Zambezi cpu on air or water. I have seen a couple of Ivy's doing 5G on water though. I think even mine can do that given enough volts but i'm not willing to push that much volts for 24/7 operation.

I'm not trying to bring AMD down as i have owned more AMD rigs than Intel so your last sentence makes zero sense again. I'm just being realistic and reality is that atm Intel is king of performance. Like it or not. Unless you can prove me otherwise.

[FlanK3r]

IB not much (most of them only 4400=newest batchs up to 4700 MHz average), SB is better for air/WC OC

[Maxforces]

IB not much (most of them only 4400=newest batchs up to 4700 MHz average), SB is better for air/WC OC

but you need add +200MHz to match performance of IB. I have 2 SB stable 4.8 and 4.9 and one fx 8150 stable 4.8Ghz. I can forget about the 5GHz on two platforms 24/7.

[Oese]

Well, its Intel (all time No. 1) vs. AMD (hoping to find another "game changer").

AMD has opted to concentrate on future performance, they were looking for the right architecture for the next few years. Why? Because they dont have a chance beating Intel in their own game. Intel is the one that will let AMD bulldoze and steamroll the new instructions (because they can easily adopt it and then again strike AMD in their "changed game") while sitting back comfortably and watching the market change. So they always have the faster optionfor old software, cause that is what they concentrate on. And guess what? They are slowing down software improvements with that strategy. The software market will wait for Intel to adopt new instruction sets.

It will be like this forever, i cant see this changing...

[alpha0ne]

It will be like this forever, i cant see this changing...

Agreed, Intels foundry business will keep Intel ahead of AMD forever but competition from ARM should start to make a serious challenge within ~ 5 years IMHO .....

[Zeus]

IB not much (most of them only 4400=newest batchs up to 4700 MHz average), SB is better for air/WC OC

In terms of overclocking i think current Zambezi and Ivy Bridge are about equal with Ivy slighly in the lead maybe. There's a thread about overclocking Ivy in the Intel section and you can see a lot of people running their Ivy's at 4.6-4.8GHz.

Most Zambezi's run somewhere between 4.4GHz and 4.7GHz i guess? Still none at 5GHz but there's a few Ivy's running 5G.

Then again, my Zambezi at 4.6G seemed stable but i got the odd BSOD now and then, even under the slightest load. No idea what was causing this as it all seemed rockstable. (prime)

This is the last i've said about Intel in this thread, it's not my intention to turn this thread into a mud fight. I like AMD as much as that other brand. I really do and i really hope Piledriver and Steamroller will rock.

[Tomasis]

it looks promising! if it does 5ghz stable with aircooler on cinebench and games, i buy it

Oeses comment is excellent, it explains why one has to be realistic and expect from product from respective brand. It is way too naive to expect small player to outrun Goliath in order to convince software makers.

However, I see more in use of Linux in future if Valve supports it. Linux is perfect platform for AMD cpus.

I hope see Zeus back in AMD I always upgrade to newer softwares and ignore Super Pi haha.

its not clever to blame on GPU because it is slow on non gpu supported apps compared to gpu supported ones. Similar reasoning on CPU sides

[FlanK3r]

but you need add +200MHz to match performance of IB. I have 2 SB stable 4.8 and 4.9 and one fx 8150 stable 4.8Ghz. I can forget about the 5GHz on two platforms 24/7.

not true... 100 MHz. U can not compare example i7-3770K(3.5GHz) vs i7-2700K (3.5GHz) because turbo at SB-DT working different than at IB-DT. If you did know yet, read some manuals or test it ,-)

[Maxforces]

FlanK3r

it is 150-200MHz most of the time is 200Mhz not 100 and who test with turbo clock to clock

[FlanK3r]

not., again not...

there is clock to clock comparison tab at 4200MHz (turbo off). As you can see, average is only 3.42% better than Sandy-DT.
4200_srovnani.png


AND OC vs power consumption vs Cinebench performance against SB-DT
oc_scaling_sandy_small.png


PS: the difference between Zambezi and Vishera could be simillary as between SB-DT and IB-DT (so about 3-5% average clock to clock)