Future Upgrade Plans: Real World Performance Advice

[h4rm0ny]

My current set up is a Sabretooth 990FX, Phenom II 1100T Hexcore, 12GB DDR3 CAS 8 RAM (mix of Corsair Dominator and Vengeance paired modules). I have a Radeon 6870 and a Kingston 60GB SSDNow drive for what relevance those have. It's overclocked to a stable 4GHz and seems happy.

I use it for a lot of business work - big spreadsheets, virtualized Linux systems, and graphics rendering. CPU and memory intensive stuff, rather than games. (The 6870 was unjustified extravagance, but I figure I can get away with that here. )

I'm trying to assess whether it would make sense to upgrade the processor to Bulldozer. Actually, as Pile Driver is (I think) not too far away, I'll probably hold off for that.

Basically, I have six real cores happy at 4GHz. I'm not sure if something with 8 "sort of" cores (or 4 sort of 1.5 cores, I'm really not sure how to think of Bulldozer) is going to be better. It looks to me like it's a bit two steps forward, one step back from where I am.

I'm also not sure if there are other changes in the chip design that make it better independent of the number of cores and clock speed.

Also, I *think* that having a Bulldozer / Piledriver CPU might let me overclock / upgrade the memory more than I can with the 1100T. But I don't know if that's true as I'm not very good with memory side of things. Faster memory would probably be a pretty good thing for the graphics rednering and biblical-scale spreadsheets, so if Bulldozer let me improve that side of things, it would be a plus by itself.

So advice please! I bought the Sabretooth partly for its upgradeability in expectation that Bulldozer would be super-dooper, but when it came out, the reaction from people was a bit "meh". From where I am and what I want (crunchy power, not gamey power), is an upgrade (and to what) worth while?

Thanks a lot for any suggestions.

H.

 

[trents]

Core for core and clock for clock the 1100T will out perform the Bulldozer but if you go with a Billdozer 8 core you would get some extra performance because of the 2 extra cores and the fact that, with good cooling, it will overclock higher than the 1100T. But it will not be huge difference and you would need pretty high end cooling (think water) to get there. My feeling would be to wait for 6 months to a year. Not necessarily for Piledriver but for the Bulldozer technology to mature (for more efficient core steppings to be developed). Then BD will scale better with regard to price, power consumption and performance. That should happen just as Piledriver is about to be released.


What are you currently using for a CPU cooler.

 

[h4rm0ny]

Core for core and clock for clock the 1100T will out perform the Bulldozer but if you go with a Billdozer 8 core you would get some extra performance because of the 2 extra cores and the fact that, with good cooling, it will overclock higher than the 1100T. But it will not be huge difference and you would need pretty high end cooling (think water) to get there. My feeling would be to wait for 6 months to a year. Not necessarily for Piledriver but for the Bulldozer technology to mature (for more efficient core steppings to be developed). Then BD will scale better with regard to price, power consumption and performance. That should happen just as Piledriver is about to be released.


What are you currently using for a CPU cooler.

Hmmm. You've said sort of what I thought it might be. That BD might have 33% more cores, but those cores are worth less so the improvement is not what I hoped. I do still wonder whether for some of the applications I use, e.g. graphics rendering, whether the more cores are better than fewer better cores... I just don't know enough about it and maybe few do.

I'm using a Noctua NH-D14 which I think doesn't leave much room to go higher on air. I don't really fancy fiddling around with water. I want to stick with what I have more or less.

What about memory? If I get BD / PD does that open up the door to faster memory? PD is supposed to be out in the next couple of months, yes?

I may just put the money toward a very, very fast SSD instead if you think BD at this time would be a bit of a waste. I do have an SSD, but it's one of the earlier ones and there are faster available now. (Though it would mean re-installing Windows 7 again - ugh!)

Thanks for the reply, btw. Very helpful.

EDIT: Also, I guess, I'm curious as to whether there is much new stuff in there that has more subtle advantages. I heard something about Windows being upgraded to run better on the new architecture for example...

 

[RGone]

What if? I wonder? Have you heard? All that about a processor that is not here yet and has not really had any clandestine testing by review sites or the like so far as any of us know about. Guess and speculation alone is driving PD talk. Hype for PD just as for BD and thus after a year of speculation the BD came onto the scene and most were severly disappointed and only just now after at least a half year are we seeing many coming to the forum with BD. And many are coming in with BD "simply because" AMD dried up the thuban cpu market. Pretty slick, shady...I mean brilliant marketing plan.

I certainly would not go BD with PD maybe 6 to 7 months in the offing. IF PD is some faster and uses less power, then it would certainly be the way to go. But to be the first on your block with PD could be the worst procedure. Wait and see if users smile with PD and if the PD arrives before Win8 or the Linux guys are all ga ga over PD then go for it.

Ram speed has not done very much for BD even though the BD IMC handles faster ram somewhat better than Thuban did. Who can even guess at what PD will do with ram. Looser timings to get faster ram has been mostly a 'wash' unless the ram speed was far greater at the looser timings. DDR1600 at Cas 9 would probably only be just equal to DDR2000 at Cas 11. That makes little sense unless you just like to say you have DDR2000 ram. But then most things are driven by a mine is bigger than yours mind-set.

See how folks appreciate the PD. Or get a BD and see how it works for real as you use it, since most testing done by review sites will say little about performance as you use your computers anyway.

 

[h4rm0ny]

What if? I wonder? Have you heard? All that about a processor that is not here yet and has not really had any clandestine testing by review sites or the like so far as any of us know about. Guess and speculation alone is driving PD talk. Hype for PD just as for BD and thus after a year of speculation the BD came onto the scene and most were severly disappointed and only just now after at least a half year are we seeing many coming to the forum with BD. And many are coming in with BD "simply because" AMD dried up the thuban cpu market. Pretty slick, shady...I mean brilliant marketing plan.

I certainly would not go BD with PD maybe 6 to 7 months in the offing. IF PD is some faster and uses less power, then it would certainly be the way to go. But to be the first on your block with PD could be the worst procedure. Wait and see if users smile with PD and if the PD arrives before Win8 or the Linux guys are all ga ga over PD then go for it.

Ram speed has not done very much for BD even though the BD IMC handles faster ram somewhat better than Thuban did. Who can even guess at what PD will do with ram. Looser timings to get faster ram has been mostly a 'wash' unless the ram speed was far greater at the looser timings. DDR1600 at Cas 9 would probably only be just equal to DDR2000 at Cas 11. That makes little sense unless you just like to say you have DDR2000 ram. But then most things are driven by a mine is bigger than yours mind-set.

See how folks appreciate the PD. Or get a BD and see how it works for real as you use it, since most testing done by review sites will say little about performance as you use your computers anyway.

This is a shame. I actually thought that PD was supposed to be out very soon, but you guys seem to be implying we're looking at six months away.

I get what you're saying about the hype - but I thought with PD we were going to be seeing the maturation of the BD architecture, that it would be a BD MkII which solidified any improvements. I'd read elsewhere that OS's were going to be taking advantage of new stuff in BD/PD over the older Phenom II's that would lead to other small performance gains.

I certainly don't want this just so I can say I have bigger RAM. Few of my friends are into computers so if I want to show off, there are more efficient forms of bling I can spend my money on. But if DDR3 2000 is no better in real terms than DDR3 1600, then why would anyone buy it?

Thanks for the reply. This is useful stuff.

H.

 

[Knufire]

There are specific applications where the faster RAM offers a noticeable benefit. Some research tasks are specifically memory intensive, but the major advantage is in benchmarking.

 

[trents]

If you want to see big gains in everyday real life performance your idea about going with an SSD drive is the ticket. Just choose your SSD carefully. They aren't all equal.

 

[h4rm0ny]

There are specific applications where the faster RAM offers a noticeable benefit. Some research tasks are specifically memory intensive, but the major advantage is in benchmarking.

Do you know if graphics rendering would benefit from that? I do 3D rendering and it's one of the most time-consuming tasks I have to do. The process can easily eat up 7-8GB of memory. I don't know if speed or latency is the most important thing. And as the CPU is pretty much maxed out during renderings, I don't know if that means its processor bound rather than memory bound anyway...?

 

[h4rm0ny]

If you want to see big gains in everyday real life performance your idea about going with an SSD drive is the ticket. Just choose your SSD carefully. They aren't all equal.

The drive I have is this one (pretty sure. It's definitely a Kinston 64GB and I bought it about two years ago):

http://www.kingston.com/datasheets/sv200s3_us.pdf

The one I might potentially get is either of these:

http://www.awd-it.co.uk/corsair-for...-solid-state-hard-drive-cssd-f120gbgt-bk.html

http://www.awd-it.co.uk/ocz-agt3-25sat3-120g-agility-3-120gb-solid-state-drive-sata-iii-2.5-ssd.html

I have all my data on separate, RAID 1 dual-disk arrays. I don't know how much benefit I will see from a newer SSD. I can buy if it will help.

I'm almost certainly going to be buying more memory soon because the graphics work is nearly eating all that I have. I need to bump up to at least 16GB. I'm not sure I wouldn't mind more, but that would probably be very expensive (I haven't looked yet). But as I'm probably replacing the memory, that's one of the reasons I asked about the newer AMDs handling of memory.

 

[trents]

yeah, bulldozers will handle 1866 ghz okay but I don't know about more than 16 gb of ram. Especially might be problematic if you try to overclock. Some motherboards won't even recognize that much.

 

[h4rm0ny]

yeah, bulldozers will handle 1866 ghz okay but I don't know about more than 16 gb of ram. Especially might be problematic if you try to overclock. Some motherboards won't even recognize that much.

Hmmm. I've recently started to get my head around overclocking, but memory is still a bit of a dark art to me. I was thinking of getting something like one or two of these:

http://www.overclockers.co.uk/showproduct.php?prodid=MY-313-CS&tool=3

As I've recently started to max out the 12GB that I have with some of the rendering projects, I think a boost might be useful..