• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

AMD Phenom II X6 1045T, or Sandy Bridge 2500K?

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
No. In fact, it's a pretty bad choice. By going with a 1055T he'll be spending $12 more on electricity every single month since it'll be running 24/7. He'll make his investment back in three-four months and afterwards it'll be paying for itself. Really don't see why you keep pushing him towards getting an AMD system when the price difference between both systems is a meager $40. The Intel system will also perform higher.
Point is, 2500k wasn't designed for heavy multithreading. This wPrime benchmark proves it. Thuban @ 4.2 beats the crap out of 2500k clocked @ 4.9. 2500k may consume less power but it will surely run hotter at the same time, equal cooling provided.



Obviously, it's up to Larry to make the choice...I am not pushing anybody here. Just hate seeing things one-sided as the majority of 2500k owners make it look like 😉
 
Last edited:
MagicCarpet, I'm not trying to dissuade you but RARELY are 1100Thubans able to overclock stably at 4.2 GHZ and not with less expensive MBs. My 1100T is on an Asrock 970Extreme4 MB which runs 2 460GTX video cards in SLI. I have a problem OC beyond 3.8 due to the limited power Phases of the Asrock MB. To buy a higher End MB that would "guarantee" 4.2 out of the Thuban would mean spending at least $180. The Intel I2500k almost always OCs to 4200-4400 on most MBs. I'd opt Intel. In the interest of full disclosure, I just ran WPrime at 32M on all three machines. On the Asrock Intel 2500K@4.4Ghz it scored at 8.71 seconds; on the Asus Intel255K@4.52Ghz it scored 8.442 seconds and on the Asrock 970 Extreme 4 with the 1100T stock it scored 12.4 seconds and OC to 3.9 Ghz it scored 7.39 seconds. No doubt the Thubans extra cores helped but honestly that 3.9 was DANGER zone for me. The Intel scores could go higher stably. Overall the Intel is noticeably faster. BTW, I'm using the latest WPrime download so the numbers in your chart don't correalate to it.
 
Last edited:
Point is, 2500k wasn't designed for heavy multithreading. This wPrime benchmark proves it. Thuban @ 4.2 beats the crap out of 2500k clocked @ 4.9. 2500k may consume less power but it will surely run hotter at the same time, equal cooling provided.



Obviously, it's up to Larry to make the choice...I am not pushing anybody here. Just hate seeing things one-sided as the majority of 2500k owners make it look like 😉

Not really sure if you're serious. wPrime is not indicative of performance in the real-world or in DC. That same benchmark puts a dual-core Atom as being the same speed as an ULV Core 2 Duo. Not much weight can be placed on it.

In DC an i5 and an X6 are comparable in performance clock-for-clock.
 
Yeah, but thing is, he's using it for DC. Like I've repeatedly said, a 2500K at 4.5GHz is faster than a 1055T at 4GHz in distributed computing. It also consumes less than half the power, meaning you save on electricity anywhere from $5-12 a month depending on how much you keep it running. Also saves on cooling/utility bills if you use air conditioning.
i hate to say it, but you're making a mountain out of a molehill. the PPD of either the 1100T or the 2500K is going to pale in comparison to the PPD of the OP's GPU (assuming he lends his GPUs to the appropriate DC projects)...and he's going to have multiple GPUs at that. so the 10% increase in CPU PPD is actually probably going to amount to significantly less than that once you take into account his gross PPD (the combined PPD of the CPU and GPUs).

take one of my crunching rigs for instance. my single HD 5870 GPU nets ~220,000 PPD in Milkyway@Home, and my 1090T CPU nets ~12,000 PPD between Einstein@Home and Test4Theory@Home. if i had a 2500K and an increased CPU PPD of ~10%, then my CPU PPD would be approx. 13,200. factoring the GPU PPD back into the equation, we see that with the 1090T platform, i net ~232,000 PPD, and we see that with the 2500K platform, i'd net ~233,200 PPD. that's a ~0.5% increase in gross PPD. now granted, an "apples to apples" comparison can never be made b/c there's too much variation in point logging between projects, but i'm sure you see the bigger picture here...

now consider the fact that distributed computing GPU tasks often consume a partial CPU core (typically anywhere from 0.04CPUs to 0.5CPUs). with the OP's rig running some CPU tasks and some GPU tasks, i highly doubt he'll find the perfect balance of CPU and GPU crunching that'll allow an exact 100% CPU load (and neither under-utilize nor over-utilize the CPU). that being the case, over-utilizing the CPU is only going to slow down his crunch times and PPD productivity. so he'll likely have to slightly under-utilize the CPU to prevent that from happening, which means it won't be running at 100% load very often (if at all), and therefore won't be consuming nearly the 250W you projected earlier in the thread. even if his CPU runs a consistent load of 95%, i don't think that'll be a problem as far as power consumption goes. while it would seem logical to assume that the relationship between CPU load and power consumption is a linear one, my experience tells me otherwise...that is to say, my 1090T CPU seems to consume FAR less (not proportionally less) power at an average 95% load than it does under 100% load, according to my kill-a-watt meter. if you're not sure what i mean about the difficulty of pegging CPU usage at exactly 100% (no more, no less) while trying to crunch a combination of CPU and GPU tasks, i can elaborate on that...

the moral of the story? the OP's CPU PPD might be ~10% better if he chooses the 2500K over the 1045T, but his gross PPD will hardly see an increase b/c he also crunches with multiple GPUs. also, he probably won't save nearly as much on his electric bill as you project if he utilizes and loads the CPU strategically. there's absolutely no denying that the 2500K is the more efficient CPU with respect to both power consumption and PPD. but the $40 extra the OP would have to spend on the intel platform doesn't seem so meager anymore when you 1) consider that the difference will not actually be covered by electric bill savings in only 3-4 months, and 2) consider that the performance increase will be borderline negligible when looking at gross PPD (combined CPU and GPU PPD).

all that being said in the spirit of further stimulating a productive conversation, if i were in the OP's shoes, i'd take the intel 2500K route. but obviously my biggest reasons wouldn't be PPD and power consumption - it would be future upgradability...and let's face it, the upgrade path for Sandy Bridge looks much more appealing and certain than does any upgrade path for Bulldozer/AM3+...



Since you are going to microcenter GET THE 2500K (sorry for the shout). I have 2 2500Ks AND an AMD 1100T with 2 460s running sli. They run noticeably faster in the 2500K rigs.
even though one of the OP's system requirements is SLI, i think it should be last on the list in terms of importance. the only time it'll matter to the OP that 2 GTX 460's in SLI is noticeably faster on a 2500K platform than an 1100T platform is during the 1% of the time that he's actually gaming. the other 99% of the time those GPUs will be crunching, not gaming, and will therefore be run individually (not in SLI) 99% of the time.
 
Well, today's the trip to Microcenter. I just checked their web site, and they only have 1 left of the Asrock 990FX Extreme4 board. However, they have several of the P67 Extreme4 and Z68 Extreme4.

I would prefer to stick with the AMD, I think. No risk of having "bent pins" from the factory, and having MC refuse to RMA it because of that. However, there is the issue of "throttling", that I've read about on both Gigabyte 990XA-UD3, and ASrock 970 Extreme4 and 990FX Extreme4. I am unsure if it applied to Thubans, though, or just FX chips.
 
i hate to say it, but you're making a mountain out of a molehill. the PPD of either the 1100T or the 2500K is going to pale in comparison to the PPD of the OP's GPU (assuming he lends his GPUs to the appropriate DC projects)..


Not in F@H, Fermi only gets around 18k PPD on the high end cards (470+) while an i5-2500k on ubuntu can eaisly put out 40k+ PPD with big advantage, however things seem to be changing on that front as far as needing 12 cores for big adv I've heard. My 1090T at 4.3Ghz only pulled about 22k.

It could probably still be done, the only concern is the time limit on the new requirements... Not even sure that went through as I stopped folding a while ago.

My i5 will pull as much PPD as 2 470s @ 900 core, and a 9800GT while using stupidly less power... It's one of the reasons I stopped folding.


There is nothing my 1090T could beat my i5-2500k at, nothing. There were some things it came close to my i5 in, like encoding, but it was never a faster chip in anything multithreaded. I ran my 1090T between 4.3GHz and 4.5GHz, and run my i5-2500k between 4.8GHz and 5.2GHz.

The only chip AMD has that can compete with the i5-2500k is the 8120/8150, but those are slower in gaming for the most part than Phenom II. The biggest problem with them is their gaming performance and their desire to drain your wall of wattage at over twice the rate overclocked as an i5-2500k while only performing slightly better best case.
 
Last edited:
Well, today's the trip to Microcenter. I just checked their web site, and they only have 1 left of the Asrock 990FX Extreme4 board. However, they have several of the P67 Extreme4 and Z68 Extreme4.

I would prefer to stick with the AMD, I think. No risk of having "bent pins" from the factory, and having MC refuse to RMA it because of that. However, there is the issue of "throttling", that I've read about on both Gigabyte 990XA-UD3, and ASrock 970 Extreme4 and 990FX Extreme4. I am unsure if it applied to Thubans, though, or just FX chips.

*NOOOOOOOOO!* 😛

The risk of bent pins is probably one-in-a-million with each CPU. Remember that when it comes to bent pins it's actually AMD that has more problems since the pins are on the CPU itself where you can make more contact with them and they're pretty fragile. Intel motherboards come with the pins on the socket and they're even more unlikely to have problems because you have less chances of touching the pins themselves. In any case, I think it's pretty irrelevant. I've handled more than 20 Intel + AMD CPUs and motherboards and have never had problems with bent pins.

I've also given thought to this, and Intel definitely seems like the way to go. What was said by another user about totals is pretty irrelevant, too, since a single GTX 460 will give you lower PPD than an i5 or a Phenom II X6. Two of them are needed to match the CPUs. If you have three of them running F@H you know you'll need a very powerful PSU and your electricity bill will rise by a lot, and what I said about power consumption still holds true. Given your usage, you'll save $12 on electricity each month or $120 a year if you go for the i5 while having higher performance. Another thing to take into account is that an X6 running at 1.45V will suffer more degradation than an i5 at 1.30V, and while you'll have 50% more cores with the X6, each of those cores is 50% slower.

To me, if the difference is only $40, it seems like an easy victory for the i5. But it's your call. Again, you'll make your $40 investment back in three-four months and afterwards the CPU will be paying for itself.
 
thought i would post this here the phenom II 960 quad core i have heard about possibilitys of unlocking it to a hex core how good are the chances of that ?
 
The risk of bent pins is probably one-in-a-million with each CPU. Remember that when it comes to bent pins it's actually AMD that has more problems since the pins are on the CPU itself where you can make more contact with them and they're pretty fragile. Intel motherboards come with the pins on the socket and they're even more unlikely to have problems because you have less chances of touching the pins themselves. In any case, I think it's pretty irrelevant. I've handled more than 20 Intel + AMD CPUs and motherboards and have never had problems with bent pins.
I was referring to cases in which the mobo has bent pins, for whatever reason, coming from the factory. And then getting it home, inspecting it, seeing it has bent pins in the socket, taking it back to the store, and having them say "screw you".

I've read several stories to that effect. Neither stores, nor ASrock, will RMA a board with bent pins.

To me, if the difference is only $40, it seems like an easy victory for the i5. But it's your call. Again, you'll make your $40 investment back in three-four months and afterwards the CPU will be paying for itself.
Well, what really put me over the edge, was the mobo features. The 990FX Extreme4 has literally everything I want. The Intel boards, while close, did not. (Unless you go to the Z68 Professional, for $260.)

Btw, I actually picked up TWO X6 1045T CPUs. One is destined for my K9A2 Platinum board, which is my quad-GPU cruncher. Currently it only has a dual-core CPU, I wanted to bump that up, so that it didn't restrict the GPUs.

Edit: I went AMD, got the 990FX Extreme4 board, last one they had left.
 
Last edited:
I was referring to cases in which the mobo has bent pins, for whatever reason, coming from the factory. And then getting it home, inspecting it, seeing it has bent pins in the socket, taking it back to the store, and having them say "screw you".


I've read several stories to that effect. Neither stores, nor ASrock, will RMA a board with bent pins.

Same thing happens if you get an AMD CPU with bent pins. :whiste:

Well, what really put me over the edge, was the mobo features. The 990FX Extreme4 has literally everything I want. The Intel boards, while close, did not. (Unless you go to the Z68 Professional, for $260.)

Btw, I actually picked up TWO X6 1045T CPUs. One is destined for my K9A2 Platinum board, which is my quad-GPU cruncher. Currently it only has a dual-core CPU, I wanted to bump that up, so that it didn't restrict the GPUs.

If you're happy with it, fine. I don't think you made the best choice, but at least you got something decent. I don't really think having IDE would put me off getting a CPU and platform that were much better. Anyway, enjoy.

Edit: I went AMD, got the 990FX Extreme4 board, last one they had left.

^^
 
I'm glad there are still avid AMD fans buying their stuff. If simply just to keep the company afloat while we enjoy our superior Intels at reasonable prices.
 
I'm torn. I really am.

Initially, I was looking at a 2500K at MC. But the boards I wanted, one of the ASrock Fata1ty series, were either too expensive, or not carried/not in-stock at MC.

Here's what I really want:
SLI/CF support (the more video cards the better)
Front-panel USB3.0 header
IDE and floppy ports (I could live without them, but it's a bonus if they are available.)
SATA 6G for an SSD (the more ports the better)

With the AMD rig, with a 990XA-UD3 board, I would get my SLI support, with a third PCI-E x16 slot for a third GTX460, for PhysX or CUDA purposes.
I would get my front-panel USB3.0 header.
I wouldn't get IDE nor floppy.
I would get SIX SATA6G ports.
Filling both PCI-E x16 primary slots, would leave one PCI slot open.

With an Intel 2500K rig, with a P67 Professional board, I would get my SLI support, again, with a third PCI-E slot for a third GTX460.
I would get my front-panel USB3.0 header.
I would get IDE and floppy.
I would only get TWO SATA6G ports (off of the Intel), and another FOUR off of marvell.
Filling both PCI-E x16 primary slots, would block off all PCI slots.

But here's the problem.
With the AMD solution, that would be $130 + tax (CPU), - $30 (sale of free combo mobo), $140 (Newegg, for mobo) = $240

With the Intel solution, that would be $180 + tax (CPU), lose out on $50 off mobo because MC doesn't stock it, and $197 for mobo at Newegg = $377

So is a 50% increase in price worth it, for only a 10% increase in PPD?

That's where I'm troubled. Sure, the Intel is faster, in an absolute sense, but is it worth the extra money? That's where I'm having an issue seeing that.

I don't care if gaming is 10FPS faster, if it is already running at 60FPS.

Just buy AMD and be done with it . You will be helping a struggling design house if you buy AMD. Thats a good thing and it could earn you your angle wings . LOL.
 
Agreed. good replies guys...

Any quad core CPU can do BF3 just fine as long as your video card is not bottlenecked,, gl
 
Meh, there's no point in telling him since he bought already. If he's happy with the inferior AMD setup, then what can we do?
 
Not in F@H, Fermi only gets around 18k PPD on the high end cards (470+) while an i5-2500k on ubuntu can eaisly put out 40k+ PPD with big advantage, however things seem to be changing on that front as far as needing 12 cores for big adv I've heard. My 1090T at 4.3Ghz only pulled about 22k...

...My i5 will pull as much PPD as 2 470s @ 900 core, and a 9800GT while using stupidly less power... It's one of the reasons I stopped folding...
i know...hence my disclaimer:
i hate to say it, but you're making a mountain out of a molehill. the PPD of either the 1100T or the 2500K is going to pale in comparison to the PPD of the OP's GPU (assuming he lends his GPUs to the appropriate DC projects)...

and did the OP mention somewhere that F@H was going to be the primary DC project of choice? i saw him mention distributed computing several times, but if he mentioned F@H, then i must have missed it. at any rate, a majority of DC projects that support both CPU and GPU computing are far more efficient on the GPU than they are on the CPU. while there might be more, F@H is the only DC project i know of in which a select few CPUs can significantly outperform most GPUs. i'm not even sure if the reason is that GPUs are genuinely not as efficient as CPUs in F@H so much as it is the nature of the points system and the limitations on the many types of WU's that can be run on the CPU and GPU respectively. if GPUs genuinely are not as efficient as CPUs in F@H, then F@H data would have to be very different from most DC project data, which typically includes massive amounts of instruction level parallelism, and therefore crunches much more efficiently on GPUs. either way it doesn't matter...as i was saying before, F@H is but one of many DC projects that can put a GPU to use. and i doubt the OP was going to put any of his GPUs to work on F@H when we know a CPU can do the same or better all the while drawing far less power.

rather he could put a GTX 460 GPU to work on Einstein@Home BRP4 tasks, complete them in 1/4 the time his 1045T would, and therefore earn 4 times the PPD. if he did it on all 3 GPUs, and the PCIe bandwidth limitations didn't bottleneck things too much (unfortunately they will to an extent), he could probably net ~10 times the PPD his 1045T would otherwise be capable of. he could also put those GPUs to work on SETI@Home Multibeam tasks, which i think run on the order of ~20 times faster on this nVidia GPU than it would on the 1045T.

long story short, in most cases the GPU is a far more efficient cruncher than the CPU. F@H seems to be a rare exception to the rule...


What was said by another user about totals is pretty irrelevant, too, since a single GTX 460 will give you lower PPD than an i5 or a Phenom II X6. Two of them are needed to match the CPUs. If you have three of them running F@H you know you'll need a very powerful PSU and your electricity bill will rise by a lot...
no - its only irrelevant if you continue to assume that the OP is only going to lend his CPU and GPUs to Folding@Home and nothing else...and nowhere in this thread did i see the OP state that. when you take into consideration the whole of the DC community, and not just a single project, you'll find that those GPUs will be far more efficient running not some, but most other projects than they would be folding...

...as regards power consumption, how big an issue is it for the OP in the first place? he plans on having at least 2 GPUs so he can SLI (albeit 1% of the time)...and he'll probably have 3 GPUs if he can fit them. if he was planning on using them to crunch while he's not gaming, which would be most of the time, then he must already be aware of the fact that 2 to 3 GPUs under moderate to heavy DC loads is going to consume a generous amount of power and require a powerful PSU anyways...
 
For the record, I used to run F@H on my quad-GPU cruncher, with four 9600SGO 96SP cards on an MSI K9A2 Platinum. PPD was something like 4K per card.

Nowadays, though, I'm more focused on PrimeGrid and Correlizer. PrimeGrid also makes good use of CUDA GPUs, and I think it is much faster on them. At least, I don't think the points system is as screwed up as F@H is.

The funny thing about Correlizer, is, unlike just about every other application, faster on AMD Athlon II/Phenom II than it is on Core2Quad at the same clock speed. Yes, AMD faster than Intel per clock. You heard it here first.

So I figure the X6 should really cut through those WUs.

One thing though, PrimeGrid (and Prime95) are apparently experimenting with implementing AVX support. Which is supported by Sandy Bridge, but not by Phenom II/Thuban. I'm unsure if FX implements AVX, or if they even implement it in the same way that SB does, or whether there has to be a Bulldozer-specific codepath. If AVX support becomes popular, I might just pick up an 8-core FX to stick in there. But I would rather wait for Piledriver, I'm looking forward to 10 cores. 🙂


Edit: Oh, and as far as the power supply goes, I picked up a single-rail ABS 1050W 80Plus Silver PSU from Newegg some time a year or so ago when they were clearing them out. I haven't actually tested it yet, I certainly hope it works. Otherwise I'll have to drop $200 on another 1000+ W quality PSU.

But I think that should be sufficient for three GTX460 cards, overclocked, and a 4.0Ghz Thuban X6 CPU, also overclocked.
 
Last edited:
Heh, "inferior".

Let's put it this way. The X6 CPU was $130, compared to $230 for the Intel alternative, that is what, 10% faster?

So now you're changing the story. You just said the difference was only $40.
It's pretty clear you were looking for an excuse from the get-go to go with AMD, but you should've made that clear from the beginning and not made me waste my time. Anyway, enjoy your setup. Even if it's inferior, it's decent for the price.
 
I see what you did there.

Yet I don't, because the reality is that in this case it IS inferior. Or is calling it "inferior" politically incorrect? Same thing as saying a 3GB GTX 580 is inferior to an HD 7970, which it is. Also, that comment was made in a different thread altogether.

Anyway, there's no point in further derailing the thread. The OP already made the choice.
 
So now you're changing the story. You just said the difference was only $40.
I think that was back when I was considering the Gigabyte 990XA-UD3 board, which Microcenter doesn't stock. So I would have been paying $130 for the hex-core CPU, and another $150 for the motherboard, for $280 total.

If I went with the Z68 Extreme4 Gen3 board at Microcenter, that would have been $135 after combo discount, plus $180 for the CPU, so $315.

That would have been the $40 difference. However, I went with Magic Carpet's recommendation, the Asrock 990FX Extreme4 AM3+ board. Which was $120, after combo discount, plus $130 for the CPU, so $250 total.

So the actual difference was $65.

But I was making a point about the price/performance of the CPUs themselves, irrespective of the motherboard that they were paired with. In that case, the Intel is $100 more expensive.

It's pretty clear you were looking for an excuse from the get-go to go with AMD, but you should've made that clear from the beginning and not made me waste my time.
I admit, I was leaning in that direction initially, as it was my opinion that the X6 was superior for DC than the 4-core Intel without HT. Whether that is true or not, I don't know. I know that Correlizer, runs faster on AMD chips than it does Core2Quad chips of the same speed.

Anyway, enjoy your setup. Even if it's inferior, it's decent for the price.
You keep using that word, as if the only yardstick is yours. I judged the AMD option superior, due to the features that I could get with the motherboard, for the price that I got it for. If I had gone Intel, and demanded the same features, I would have paid $260 - $50 discount = $210, plus $180 for the CPU, so $390 total, which would have been a full $140 more expensive, for the same features, but just to have a 10% faster CPU.

I'm happy with my decision.
 
Yet I don't, because the reality is that in this case it IS inferior. Or is calling it "inferior" politically incorrect? Same thing as saying a 3GB GTX 580 is inferior to an HD 7970, which it is. Also, that comment was made in a different thread altogether.
But the GTX580 isn't inferior. If you are using CUDA apps, it is infinitely superior.

Your view of the world isn't the only one, you know.
 
Back
Top