As expected, NVidia has released an additional GeForce card in the form of the GTX650Ti, filling out the GeForce 600 range.

The GK106 GTX650Ti has 768 cuda cores (shaders) and reference models are clocked at 925MHz, sport 1GB GDDR5 and have a TDP of 110W (so one 6-pin connector is sufficient).
Manufacturer variants will include 2GB versions and a range of clocks from 925MHz to 1071MHz (~16% range so far)…

As with other GK cards, these are PCIE3 compliant, but like the GTX650 they do not feature boost (get a fast one)!

Cost:
In the UK these begin at £115 and rise to ~£135 (at good online retailers).
In the USA the starting price is $145, and in the Euro zone they also start around 145Euro.

Performance:
With 768cuda cores these cards fall between the GTX650 and GTX660Ti (GTX660 is OEM), and should give roughly twice the performance of a GTX650.

In terms of performance per Watt, the GTX650Ti should slightly outperform the GTX650 (in-itself), but in terms of performance per outlay (purchase cost) it’s certainly worth the extra $40 for twice the crunching performance!

While it’s generally preferable for larger cards to have 2 or 3 fans, I think one fan should be sufficient for these cards. I would however be concerned about the size of the fan; in my experience small fans fail very frequently, while larger fans generally keep running for years.

Overclocking aside (if even possible), it’s generally the case regarding improved/performance cards (cards with higher frequencies) that they give as good a performance per Watt (at least) as reference models. So when you add the system Wattage overhead (motherboard, CPU, RAM, HDD...), they are better value for money. That said, for the same reason it’s usually more efficient overall to get a ‘bigger’ card.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Update: GTX660 with GK106 and 960 shaders at 980 MHz went retail a few weeks ago.

MrS
____________
Scanning for our furry friends since Jan 2002

Ouch, that's pretty bad! By now you won't be able to use the code either, and you'll never know if that guy is happily running the game now.. or is there anything the game distributor could help you with here? Like invalidate all current installations using this code?

MrS
____________
Scanning for our furry friends since Jan 2002

About 190k RAC using only long-run tasks.

MrS
____________
Scanning for our furry friends since Jan 2002

With my GTX660Ti I didn't notice disturbing desktop lag, although HD video playback may sometimes have dropped below 25 fps for short periods. But it's got a lot more raw horse power than a GTX650Ti.

MrS
____________
Scanning for our furry friends since Jan 2002

Your core-OC still yielded more, percentage wise, than your mem-OC. However, GT640, GTX650Ti and GTX660Ti are less balanced than usual cards in the way that they pack a surprising amount of shader power into a package with "usually just enough" memory bandwidth. In games these cards are love memory OC's, much more so than usual cards. Seems like on these cards GPU-Grid is finally also affected by the mem clock by a measureable amount.

You can take a look at the memory controller load via e.g. GPU-Z. Your numbers should be interesting, with and without OC. I've got a slightly core-OC'ed GT640 at ~66% mem controller load and a slightly core-OC'ed GTX660Ti at 44%.

MrS
____________
Scanning for our furry friends since Jan 2002

Should I try a long run workunit ?

Thanks for collecting and sharing that data. This should be enough for anyone who wants to know how well this card (and some comparably fast ones) performs :)

MrS
____________
Scanning for our furry friends since Jan 2002

Interesting read.

- 54,151 s: 101,850 credits - bonus missed ?

Be sure to get a GTX 650 Ti, as it's massively faster than a GTX 650 for crunching. Oh that sweet naming madness..

MrS
____________
Scanning for our furry friends since Jan 2002

The GTX 650 Ti is twice as fast as the GTX 650, and costs about 35% more. It's well worth the extra cost.

I moved these well off topic posts from the New app is out for testing News thread, to this NVidia GTX 650 Ti Graphics cards (GPUs) thread.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

John, your issue is caused by the new features employed in the NOELIA WU. It's not related to your card at all.

MrS
____________
Scanning for our furry friends since Jan 2002

Going by your runtimes I would say you are getting around 190K/day
http://www.gpugrid.net/results.php?hostid=144671
(86400/32000)*70800≈190,000
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

A reference GTX660 has a boost clock of 1084. My GTX660Ti operates at around 1200MHz, and the reference GTX660Ti is 1058MHz, so you can't say the GTX660 has faster clocks.

I would like to see some actual results from your GTX660, to see if it really is completing Nathan Long tasks in 6h. I doubt it because my GTX660Ti takes around 5 1/2h and it's 40% faster by my reckoning:

I34R5-NATHAN_dhfr36_5-17-32-RND3107_0 4392017 24 Apr 2013 | 11:12:40 UTC 24 Apr 2013 | 19:20:22 UTC Completed and validated 19,505.27 18,807.06 70,800.00 Long runs (8-12 hours on fastest card) v6.18 (cuda42)

A 960cuda core GTX660 should take ~7.5h. The 1152shader versions are OEM.
These GeForce Kepler cards are either GK106 or GK104, but architecturally they are very similar. It's not like comparing high end and mid range Fermi's which were quite different. All the GK106 and GK104 cards are super-scalar.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

This "not using all shaders" is not the best wording if you don't know what it means. As a brief explanation: it's a property of the fundamental chip architecture, every chip / card since "compute capability 2.1" has it, which would be the mainstream Fermis and then all Keplers. This move allowed nVidia to increase the number of shaders dramatically compared to older designs, but at the cost of sometimes not being able to use all of them. At GPU-Grid it seems like the "bonus shaders" can not be used at all.. but this matters only when comparing to older cards. The newer ones are way more power efficient, even with this handicap.

Regarding the actual choice between 650Ti, 650Ti Boost, 660 and 660Ti: Jim is right that higher clock speeds are to be preferred over more shaders.. but in this case all these cards are based on the same architecture and lithography and hence can reach similar clock speeds and similar voltages (=efficiencies). I don't see a penalty here for the bigger GK104 cards. My GTX660Ti runs GPU-Grid happily at 1.23 GHz and POEM at 1.33 GHz. It's been one of the first and from what I've heard more like the norm rather than an exceptionally good card.

Regarding the issue "GTX660Ti is not faster than GTX660" I'd say "show us the numbers". Theoretically the Ti should be 30% faster in the same setup.. and this rule hasn't failed us up to now.

I'd go for the largest of these cards which fits nicely into your budget. But don't go higher than a GTX660Ti, since the GTX670 has the same number-crunching power but higher game performance.. which you'd unnecessarily pay for.

MrS
____________
Scanning for our furry friends since Jan 2002

@Flashhawk: you've also got the significant performance bonus of running Win XP. The other numbers here are from Win 7/8.

@Jim: that is impressive performance, thanks for showing the numbers. As a comparison I'll use your config with only a logical core assisting the GPU, as this is what I'm also running. In this case we've got 6:15 h = 22500 s at 1110 MHz.

I'm running at 1228 MHz and should hence get 20400 s. The difference in shaders is 1344/960 = 1.4, which means I should be getting 14530 s. Well, this is clearly not the case :D

Instead I'm seeing 19030 s for these WUs, just 7% faster at the same clock speed. I can't look the memory controller load up right now, but it was fairly low, so it shouldn't limit performance this much (especially since memory OC didn't really help GPU-Grid historically). I think it seems like these WUs are too small for the larger cards to stretch their legs, i.e. there's too much context switching, PCIe transfers etc. happening.

MrS
____________
Scanning for our furry friends since Jan 2002

Jim1348, is that definitely a 960shader version and not an 1152 OEM card?
Is it a 2GB or 3GB card?

It might be the case that some WU's benefit from the extra memory bandwidth. The bus width of the GTX660Ti is 192bits, the same as the GeForce GTX 660 and GTX650Ti, but because the GTX660Ti has more shaders and SM's it's bandwidth is relatively less. I typically see a 40% memory controller load when running GPUGRid WU's on the GTX660Ti. This is a lot higher than any previous cards I've had.

One problem in assessing the impact of bandwidth is that these tasks can vary by some margin. Just looking at 8 WU's I've seen a 7% variation in runtime on my own system. It's also hard to know what influence the operating system has. It might just be 11% from XP to W7, as it use to be, or it could be more for some WU's.

The 11% difference isn't sufficient to move from 14500sec to 18800sec though. flashawk's GTX670 (on XP) is about 30% faster than my GTX660Ti on W7x64. So, some other factor, other than OS is involved and the likely candidate is bandwidth. Another consideration might be CPU usage. I might look at this at the weekend.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

My GTX660Ti is made by Gigabyte, and comes with a dual fan. It's a descent model, but fairly standard; most boost up to around 1200MHz.

The theory has been that because the GTX660Ti has the same number of shaders as a GTX670 it should be just as fast. However, that doesn't seem to be the case. So perhaps the memory bandwidth is having more of an influence.

My CPU is an i7-3770 @4.2 GHz and I've 2133MHz system memory and a SATA6 drive, so it should not be bottleknecked anywhere else. I do use the CPU, but I usually have 2 threads free.

It's difficult to compare GPU's on different operating systems but the fact that the GTX660 also performs so well comparatively leads me to think that the GPU memory bandwidth is more of an issue than was previously thought.

- It just occurred to me that the drop from 32 to 24 ROP's could be the issue for the GTX660Ti rather than or as well as the memory bandwidth; the GTX670 has 32 Rops, but the GTX660Ti only has 24 Rops. The GTX660 also has 24 Rops - perhaps a better Rops to Cuda Core ratio.
If that is the case then we might also see a similar difference between the "GTX650Ti Boost" and the "GTX650Ti"; the Boost also has 24 Rops, but the 650Ti only has 16 Rops.

Going back over some old posts it looks like the new apps use the PCIE a bit less than before but use the GPU memory a bit more. The new apps also favor the super-scalar cards, making the old comparison charts redundant even for CC2.0 vs CC2.1 architectures.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Some quick measurements / observations:

- running my stock config (with GPU-OC and everything) I'm seeing 38% memeory controller load and 86 - 90% GPU utilization

- deactivating 8 Einstein WUs on the i7+HT increases power consumption a bit and GPU utilization to 92%

- increasing the memory speed by 130 MHz increases power by ~1%, but nothing else changes (from 1.5 GHz to 1.63 GHz, without multiplying by 4 for the DDR clock rate)

- decreasing the memory speed by 250 MHz lowers power by about 5-6%, increases the memory controller load to 41%(+/-1%) and increases GPU utilization to 93%

It seems like "waiting for memory" does not count as idle time for the GPU and that some of it is happening here. Been running a regular Nathan.

MrS
____________
Scanning for our furry friends since Jan 2002

Heat isn't an issue; dual fan card, open case with additional large case fans, 137W used - GTX660Ti presently at 55°C and would rarely reach 60°C.

The problem isn't lack of memory either; the 3GB versions don't perform any better running one GPUGrid task at a time.

I noticed that even when running two GPUGrid tasks at a time the GPU memory controller load only rose by about 1% even though the GPU utilization rose from around 90% to around 98%. I suspect that the reported memory controller load does not accurately reflect saturation - the full extent of the bottleneck. The way the memory is accessed could be part of the issue. It certainly seems to struggle to go much past 40%.

I'm still not fully aware of how the ROP's are used. Since the shaders were trimmed down, a lot now takes place on the GPU cores, so having relatively fewer ROP's could be key to the relatively poor performance. The drop to 75% of the ROP's from the 670 to the 660Ti could be the problem. It's either the memory bandwidth/controller load or the ROP count, or both. I suspect the ROP count is limiting the memory controller load, but I don't know enough about the ROP usage to be confident.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Ah, I forgot to comment on the ROP suggestion yesterday. Well, ROP stands for "render output processor". I've never heard of these units being used in GP-GPU at all, which seems logical to me, since the calculation results wouldn't need to be assembled into a framebuffer the size of the screen. And blended/softened via anti-aliasing.. which you really wouldn't want to do with accurate calculation results :D

And if there's a bottleneck in memory bandwidth then this might not need to show up as 100% memory controller utilization. For example the need for new data from memory might not be distributed evenly over time.

Could anyone run some comparable tests with the current GPU-Grid app and probably the Nathan long-runs? Preferrably within the same PC. We'd need an exceptionally low-bandwidth GPU like the GT640 or GTX660Ti compared to a regularly balanced one like.. pretty much any other. Over and downclock the memory by a noticable amount (about 10% should be fine.. not sure if my memory could take this as OC, though) and observe the change in performance. Ideally no CPU tasks would be run along with this, so the measurements aren't disturbed by changing CPU load. Run a few units in each configuration, average the runtimes and compare: is the performance change with memory clock significantly higher on the low-bandwidth card?

MrS
____________
Scanning for our furry friends since Jan 2002

The GTX660Ti has one less Rop cluster (of ROPs, L2 cache and Memory) than a GTX670. So it drops from 32 ROPs, with 512KB L2 cache and quad channel memory to 24 ROPs, 384KB L2 cache and triple channel memory.

I don't know how you can separate these cluster elements but given what MrS said I respect that any slow-down isn't ROP count based. I reckon that NVidia thinks 384K is sufficient to support the reduced memory bandwidth, but I'm not sure if you can't really separate the two. We know that the GTX670 performs better for the more memory intensive games. So that leaves the memory bandwidth looking like the culprit. The 3GB cards aren't any faster either.

I would go along with the idea of uneven memory requirements. Something is stopping the memory controller load going past ~41%. I'm currently using 38% but that's with the CPU usage at 100%. When I disabled CPU tasks and ran two WU's it only went up to ~41% and that was with 99% GPU utilization.

GoodFodder, I think we ran some >1GB Long WU's a few months back.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

I suspect the performance improvement of the gtx 660 over say the gtx 650ti is due to the larger cache size (384K v 256K) rather than bandwidth (gtx 670/680 incidently has 512K).

As some people suggested it appears that the GTX660Ti is relatively better (slightly) when it comes to Noelia's WU's, compared to Nate's.
My GTX660Ti was 25% faster than a GTX660, however for most of the run I only crunched one CPU WU, which enabled the Noelia task to use 94% of the GPU. Using the CPU less accounts for at least 4% but probably >6% compared to crunching CPU tasks on 6 CPU threads. My CPU is also at 4.2GHz, so that might make a little difference compared to stock clocks, possibly bringing it to 9 or 10%. If you run a WU without any CPU tasks running you would find out.

Might have been better to post in the NOELIAs are back! thread, but I might rename this one as we have drifted into a GPU runtime comparison thread.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

The 650TiBoost is basically a 660 with less shaders. While the shaders are historically the most important component they have been trimmed down significantly from CC2.0 cards and CC2.1 cards to CC3.0 cards. This means that there is more reliance on other components especially the core but even the CPU. So the relative importance of the shaders is open to debate. It certainly seems likely that the GPU memory bandwidth and probably the different cache memories have more significance with the GeForce 600 GPU's. I think the new apps have exposed this (they generally optimize better for the newer GPU's), making initial assessments of performances outdated. There seems to be some performance difference with the different WU types, but that would need to be better explored/reported on before we could assess it's importance when it comes to choosing a GPU. Even then it's down to the researchers to decide what sort of research they will be doing long term. I'm expecting more Noelia type WU's, as they extend the research boundaries, but that's just my hunch.

The Boost is definitely worth a look, but whether its closer to a 660 or a 650Ti remains to be seen. I would speculate that it performs slightly better than a 650Ti but not that close to a 660; the Boost has less shaders to feed so the memory won't be that important (proportionately less so). A comparison of the 650Ti to a 650TiBoost should reveal the importance of cache (at least the difference between 256 and 384K). As with any GPU, it comes down to the price, performance and running cost.

These 'mid-range' GPU's are very interesting when it comes to crunching. They let normal computer users, especially light-gamers, participate in the project (rather than just the enthusiasts). I would include the GTX 660 (OEM)'s as interesting mid-range GPUs, though they are pushing the high end bracket with 1152shaders, and we don't know the price (OEM). If you include the different memory amounts these mid-range GPU's have and the OEM cards, there are 7 to choose from. I would still like to see non-OEM 1152shader GPU's, to fill out the range; one version has a 256bit bus, so it should match the GTX660Ti (192bit). I'm sure manufacturers could make a sweet one.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Pulled a GTX660Ti and a GTX470 from a system to test a GTX660 and a GTX650TiBoost:
[email protected], with the GTX660 in the top PCIE3 slot (0) and the GTX650TiBoost in PCIE3 slot 1. Running 5 CPU WU’s, and two GPUGrid WU’s, VM off.
While running two Nathan WU’s the GPU power reported for both GPU’s stayed around 94%. With the fans on auto the temperatures stabilized at ~62degC. Fan speeds at 56% and 57%. GPU usage stayed around 93 to 94%. The GTX660’s core clock is 1071 while the GTX650TiBoost’s clock is 1124.
The GTX660’s memory clock is listed as 3005MHz and the GTX650Ti’s is 3055MHz. The GTX660 is using 453MB GDDR (~62MB for the system), and the GTX650TiBoost is using 380MB.
Memory controller loads are 31% for the 660 and 26% for the 650TiBoost
At 5% into the run (which is usually a very accurate indication of expected run times for GPUGrid WU’s, while the estimated remaining time isn’t) the GTX660 had taken 18min 10sec. Expected run time ~22000sec
At 5% into the run the GTX650TiBoost took 20min 9sec, so the expected run time is 24200sec.
The GTX660 is about 11% faster than the GTX650TiBoost.
- At 45% this still looked about right.
Considering the shader count and the core frequency (but nothing else) you would expect the GTX660 to be ~19% faster - Shaders; 960/768=1.25, core freq. 1071/1124 =0.95
So it appears that there are additional performance benefits, probably from the shader cache and memory bandwidth to shader count.

Mid-range GPU performances against the first high-end GPU (running Nathan WU’s):
GTX660Ti - 100% - £210
GTX660 - 88% - £153 (73% cost of a GTX660Ti) – 20.5% better performance/£
GTX650Ti Boost 79% - £138 (66%) – 19.6% better performance/£
GTX650Ti - 58% - £110 (52%) – 11.5% better performance/£

At the above prices the GTX660 offers the best performance/purchase price, but is only slightly better than the GTX650TiBoost. Prices are subject to change.

Running costs:
These cards all optimize towards a similar power % (91 to 95), so we should be able to go by the reference TDP’s and come up with a reasonably accurate measure of performance/Watt:
GTX660Ti – 100% – 150W – 100% (performance/Watt)
GTX660 - 88% – 140W – 94%
GTX650Ti Boost – 79% - 134W – 88%
GTX650Ti – 58% - 110W – 79%

Actual Wattage's might change this a bit, but it's roughly what I would expect; the performance/Watt of the higher specked cards would be better. Again the 660 looks quite good.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

The TDP is for reference but it's certainly not a simple consideration. You would really need to list all your components when designing a system if TDP (power usage) is a concern. Key would be the PSU; if your PSU is only 80% efficient, it's lack of efficiency impacts upon all the other components. However, if you are buying for a low end system with only 1 6pin PCIE required then the purchase cost vs the running cost of such a PSU (and other components) are likely to make it more acceptable. For mid-range systems (with mid-range GPU's) an 85%+ PSU is a reasonable balance. For a high end system (with more than one high end GPU) I would only consider a 90+ efficiency PSU (Professional series).

I just ran the two WU's last night. Today I'm finishing off a Noelia (on the 660) and have started a Nathan WU on the 650Ti. The Nathan WU is at 63% and the anticipated runtime is close to the previous WU. I will run a few Nathan WU's to get more accurate results, but so far the Relative performances (in previous post) look reasonably accurate.

One thing to note is that most GTX660Ti's are FOC. My 660 is very much a reference model, no FOC GPU or memory. The 650Ti on the other hand is an FOC card with a core clock of 8.8% over reference and a 1.8% memory FOC.
The reason for this is that the GTX660 has one 6-pin PCIE power connector and a TDP of 140W. This is very close to the 150W maximum, so it doesn't give you much headroom for overclocking. The reference TDP for a GTX650TiBoost is 134W, and having less shaders is bound to reduce the power draw. So it lends itself to FOC, as does the GTX660Ti.

Jim, if a GPU has a TDP of 140W and while running a task is at 95% power, then the GPU is using 133W. To the GPU it's irrelevant how efficient the PSU is, it still uses 133W. However to the designer, this is important. It shouldn't be a concern when buying a GPU but when buying a system or upgrading it is.

To support a draw of 133W requires:
@80% efficiency - 160W
@85% efficiency - 153W
@91% efficiency - 145W

The difference of 7W between an 80+ and an 85+ PSU, or 8W between the 85% and 91% PSU's isn't a massive consideration when you have one or two mid-range GPU's. However when you have 2 or more high end GPU's it is:
Two GTX680's (195W TDP) @ 90% power usage on an 80% efficient PSU wastes 70W just supporting the GPU's. Probably closer to 100W for the whole system. This can be more than halved by using a top PSU. The added benefits are reduced heat radiation, and thus reduced noise from not needing the fans to spin as fast. The heat problem increases exponentially when you add a 3rd or 4th GPU, but you need a more powerful PSU anyway; just for the PCIE power connectors.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

In the UK running cost difference for these mid-range GPU's would only be around £10 a year, so yeah, no big deal, and only a consideration if you are getting two or three. In a lot of locations it would be less, and in Germany and a few other EU countries it would be more. So it depends where you are.

In the US the prices seem to be stretched out more; with the higher end GPU's costing relatively more than the lower end, but I expect the GTX650TiBoost to fall in price; it was only released two or three weeks ago, and there might not bee too many to choose from. There is no way I would get one for a few $'s less than a GTX660, which still looks like the best bang for buck.

At present I have both the GTX660 and GTX650TiBoost in the system, and I'm CPU crunching on a modestly OC's i7-3770K (337W at the wall, 91% PSU). I would really need to pull both GPU's, take a reading, add one, only crunch on the GPU, take a reading and then do the same for the other GPU to have accurate figures. A bit much for 2 GPU's that have a TDP difference of 6W, but I might do it in a few days, after I get a few more runs in.
In the mean time I suppose I could make reasonable estimates based on idle power usage (from review forums)...
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

I would suggest a very modest OC, and no more. Antec don't make bad PSU's, especially 600W+ models, but I would still be concerned about pulling too many Watts through the PCIE slot on that motherboard. I suggest the first thing you do is to compete a WU and get a benchmark, then nudge up the clocks ever so slightly. I'll try to resist the urge to OC for a while, to complete more runs and get a more accurate performance table.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Measured Wattages:

0 GPUGrid wU’s + no CPU WU’s – System usage 77W (this includes the GPU idle power, of 7W each)
1 GPUGrid WU’s (on the GTX650TiBoost) + no CPU WU’s - 191W
1 GPUGrid WU’s (on the GTX660) + no CPU WU’s - 197W
2 GPUGrid WU’s + no CPU WU’s - 314W
2 GPUGrid WU’s + 5 CPU WU’s - 338W

Excluding the GPU’s idle power,
The GTX 650Ti Boost used 114W
The GTX 660 used 120W (6 Watts more, as expected going by the TDP)
Adding the idle Wattage,
The GTX 650Ti Boost used 121W
The GTX 660 used 127W

Measured when running Nathan WU’s at 95% power and 95% GPU utilization (no CPU WU’s running). Note that these WU's also use part of the CPU, which does increase the power a bit.

These figures suggest that the power usage and GPU utilization can be multiplied to get the actual Wattage used; 0.95*0.95 represents the actual Wattage used, ie 90% of 134 for the GTX650TiBoost=121W, and 95% power * 95% GPU utilization is 90% of the TDP for the 660; 127W is ~0.9*140W. Though I would want confirmation of that by others before I accept it.

____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

To me it looks like the reported power percentage is calculated relative to the power target rather than TDP. The former is not reported as widely (though it is in e.g. Anandtech reviews), but is generally ~15 W below the TDP. If this was true we could easily calculate real power draw by "power target * claimed power draw".
Some factory OC'ed cards have higher power targets built in, though (e.g. mine).

MrS
____________
Scanning for our furry friends since Jan 2002

You have raised a few interesting points:

Different operating systems perform differently here. Linux can be up to ~5% faster than XP. XP is ~11% faster than Vista and W7. I think W8 is roughly the same as W7 in terms of GPU performance. 2003R2 servers are slightly faster than XP (but only 1 to 3% last time I measured it), and 2008 servers are ~5% slower than XP (but a bit better than W7, except when it comes to drivers).
Obviously GPU utilization is higher with the faster operating systems.

I suppose I should have taken into consideration my PSU efficiency (it's 91%+).

The one big problem with these measurements is that these WU's use the CPU and the GPU. So you are not measuring the GPU running alone. The problem with this is that you can't accurately account for the CPU's power usage; running a CPU WU from another project and taking the CPU power usage from that is not accurate - you can see up to 30W different power consumption running different CPU WU's. How much power the Nathan WU's actually draw is open to debate. I suspect it's a good bit less than the average CPU WU would draw.

For reference:
0 GPUGrid wU’s + 1 CPU BoincSimap WU’s – System usage 91W
0 GPUGrid wU’s + 2 CPU BoincSimap WU’s – System usage 104W
0 GPUGrid wU’s + 1 CPU Ibercivis WU’s – System usage 93W
0 GPUGrid wU’s + 2 CPU Ibercivis WU’s – System usage 106W
0 GPUGrid wU’s + 1 CPU Climate WU’s – System usage 95W
0 GPUGrid wU’s + 2 CPU Climate WU’s – System usage 112W

Yeah, that was too easy - Power target it is. Are there any tools that can tell you what your GPU's Power Target actually is?

I noticed that with MSI Afterburner I cannot unlock the Core Voltage for the GTX650TiBoost, but I can for the GTX660. I can move the Power Limit for both. The last time I played with that it was really inaccurate.
GPUZ is telling me that my power consumption is ~96% of the TDP for my 660 and 95% of the TDP for the 650TiBoost, but that just matches Afterburners power percentage.

I've tweaked things:
660,
Core Voltage +12mV, power limit 109%, Core Clock +78MHz (multiple of 13!), Memory Clock +50MHz; GPU power % now 98%, GPU Power usage ~92% (with 2 CPU WU's), Core clock 1137MHz, GPU Clock 3055MHz.

650Ti,
Core Voltage (cant budge), power limit 109%, Core Clock +78MHz (multiple of 13!), Memory Clock +55MHz; GPU power % now 98%, GPU Power usage ~93%, Core clock 1202MHz, GPU Clock 3110MHz.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

My GTX660 really doesn't like being overclocked. Stability was very poor when it comes to crunching with even a very low OC. It's definitely not worth a 1.3% speed up (task return time) if the error rate rises even slightly, and my error rate rose a lot. This might be down to having a reference GTX660 or it being used for the display; I hadn't been using the system for a bit, and with the GPU barely overclocked, within a minute of me using the system a WU failed, and after 6h with <10min to go! It's been reset to stock.

On the other hand the GTX650TiBoost sticks a modest OC very well, and has returned WU's with the shaders up to 1202MHz (the same as my GTX660Ti), albeit for only a 3.5% decrease in runtime. I dare say a 5%+ performance increase is readily achievable. However, I'm using W7; I would get more than that by just sticking it in an XP rig, and more again using Linux. Also, in XP OCing might not be as beneficial; the GPU would already be running at ~99%. Ditto for Linux.

It's looking like a FOC GTX660 is the best mid-range card to invest in.

____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Factory Over Clocked.
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Actually the higher the GPU utilization, the more a GPU core OC should benefit performance. Because every additional clock is doing real work, whereas at lower utilization levels only a fraction of the added clocks will be used.

MrS
____________
Scanning for our furry friends since Jan 2002