• 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.

MSI Cyclone GTX460 1GB (upgrade from 8600GT)

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
So the point of my post - I have an 8600GT now...do I need to driver sweep my system before installing the 460GTX? Or can I get away with nothing more complicated then shutting down, swapping cards, and rebooting?
I'd at least uninstall your current drivers, swap the cards and then reinstall latest drivers. Just to be sure.
 
Disclamer: All video mods must overclock there gpu's or they get a vacation. 🙂
I want to see 900 core at 1.053 volts. 🙂

Heh, good luck with that... 😀

The Cyclone I installed in my brother's PC, maxed out 860mhz core @ 1.053v. It would do 900 @ 1.087v and run Furmark & 3DMark just fine, but failed during my Crysis:Warhead 3 hour loop.
 
Yeah, it was a little underwhelming. Chances are I could have left it at 880ish and it would have been fine.

But hey, when you're clocking the core at ~30% more than reference, it's still 🙂

True.

Nice overclock on the 5770 by the way I've gotten my 5750 to 915 core stable and 1375 memory ,stock voltage, not bad for $85, 9 months ago. 🙂
 
Install the gtx460, install the latest drivers , I believe they unistall the old drivers automatically. edit: Just looked yes they do have that option.

http://downloads.guru3d.com/Forceware-260.63-Win-Vista-|-7-(64-bit)-download-2618.html


Disclamer: All video mods must overclock there gpu's or they get a vacation. 🙂
I want to see 900 core at 1.053 volts. 🙂

:thumbsup: Thanks Happy!

That was an easy install.

Right now I've got it just running stock. It is quieter than the 8600GT it replaced. Course I haven't really done much to get it hot yet (no MSI burn).

I'm benching a TMPGEnc transcode job that took ~4.5hrs w/o the help of CUDA. Going to see what the boost is before I overclock the card.
 
I use TMPGEnc for a lot of DVD transcodes as well as creating youtube HD family vids.

Perhaps the biggest use I get with TMPGEnc is that I will take 2 or 3 DVD's and transcode them to a lower bit-rate and fit them onto a dual-layer DVD.

Depending on how you optimize the transcoding with filtering, TMPGEnc can usually give you a near equivalent IQ provided you give its transcoder engine at good 3200 KB/s bandwidth. Below that and you'll see IQ degrade, above 3500 KB/s and IQ is identical to my eyes and tastes.

But you have to get rid of the video noise if you want to transcode at 3500 KB/s and not lose IQ. Video noise sucks up bandwidth as the transcode engine puts all its efforts into attempting to faithfully recreate the pixel noise instead of putting its effort into reproducing the edges and transitions of moving things we do care about.

(there's two kinds of noise to reduce, still-picture and temporal...if you have a crappy video camera like I do then you get a lot of pixel noise from frame to frame and this can be offset some by the temporal noise reduction filter)

So here are the results from a quick comparison of the same transcode job I ran with my old vid card (8600GT) and with the new GTX 460 (at stock OC).

8600GT Transcode Time: 17,103 seconds (04:45:03 )

8600GTQ6600StockTwilightZoneTranscode.jpg



GTX 460 Transcode Time: 13,546 seconds (03:46:46 )

GTX460Q6600StockTwilightZoneTranscode.jpg



The GTX460 sped up the transcode time by nearly a full hour and took just 80% of the time as the same transcode job without using CUDA.

IQ is identical with and without the CUDA assist.

I'm knew CUDA would help some, but I had no idea I'd gain back an hour of usable computer time on a 4-5hr job like this.

At this point I couldn't be happier with my $200 upgrade. I'm going to tinker with overclocking and see what kind of improvements I get versus the tradeoff of adding more noise/heat to my office.
 
I'm happy you enjoy your purchase (and sorry if I spoil it somehow 😛), but:

- your new card has more than 10 times the shader power
- your new card has more than 5 times the bandwidth

And you gained... a bit over 20%? I'm guessing the transcoding is done on the shaders, so why was the gain so miniscule compared to the hardware increase? Obviously I wouldn't expect a 10x increase in speed or something... but ~20%? 😛
 
Last edited:
I'm happy you enjoy your purchase (and sorry if I spoil it somehow 😛), but:

- your new card has more than 10 times the shader power
- your new card has more than 5 times the bandwidth

And you gained... a bit over 20%? I'm guessing the transcoding is done on the shaders, so why was the gain so miniscule compared to the hardware increase? Obviously I wouldn't expect a 10x increase in speed or something... but ~20%? 😛

I kind of thought the same thing, but I really have no idea how well shaders scale with transcoding.

Idon'tcare, try a modest overclock first. Do 800mhz on the core and see if that helps at all. Heat and noise should absolutely not be an issue with that card's cooler.
 
I'm happy you enjoy your purchase (and sorry if I spoil it somehow 😛), but:

- your new card has more than 10 times the shader power
- your new card has more than 5 times the bandwidth

And you gained... a bit over 20%? I'm guessing the transcoding is done on the shaders, so why was the gain so miniscule compared to the hardware increase? Obviously I wouldn't expect a 10x increase in speed or something... but ~20%? 😛

Actually there could very well be a linear correlation to shader count/clockspeed but I wouldn't be able to measure that.

When I use the 8600GT the transcoder (TMPGEnc) disables Cuda because my job would run slower when using it versus just using my Q6600 for the entire job.

So for all I know, if I could force TEMPEnc to use the 8600GT and CUDA (it is supported, but the cpu would have to be darned slow for it to be the better choice) then the job would take 24hrs or something silly.

All I know is that I pulled out my vid-card, popped in a fairly inexpensive (for my budgets) new one and I get an extra hour of productivity out of my rig. That makes me quite happy.

What I don't know at this stage is what else could be limiting my transcode time...maybe my harddrive is too slow now, or CPU is now the limiting factor in speedup when using CUDA.

I did notice that GPU-Z was showing only about 40-60% GPU utilization during the transcode.

Oh and CUDA only handles a small portion of the entire transcode job, some of it can only be handled by the CPU. So with an infinitely fast CUDA processor my transcode time isn't going to go to zero anyways, it might reach a limit at say 3hrs for this specific job (I don't know).

I kind of thought the same thing, but I really have no idea how well shaders scale with transcoding.

Idon'tcare, try a modest overclock first. Do 800mhz on the core and see if that helps at all. Heat and noise should absolutely not be an issue with that card's cooler.

Thanks for the advice tviceman...I was realizing this morning that unlike OC'ing cpus I have zero experience in OC'ing vid-cards. Total newb in this area in terms of software to use (combustor for stess testing, right? and what about setting OC profiles?)...gonna do some forum searching later.
 
Thanks for the advice tviceman...I was realizing this morning that unlike OC'ing cpus I have zero experience in OC'ing vid-cards. Total newb in this area in terms of software to use (combustor for stess testing, right? and what about setting OC profiles?)...gonna do some forum searching later.

MSI afterburner for overclocking, kombostor for stability, and OCCT to validate your overclock. OCCT is great if you are going to be doing the stuff you are doing because it has an error check mode so it can make sure your video card is processing 100% error free at a given clock speed / voltage.
 
:thumbsup: awesome, downloading the tools now 🙂

A couple of you whiz-bang OC'ing gurus should get together and make a combined AMD/Nvidia "Overclocking your AMD or Nvidia Video Card for Newbies" thread.

Something short and simple, links to the tools used/needed and maybe a rough guideline of what to expect from the effort for a few cards.

"On GTX460 you should get to around 800MHz OC with minimal effort and minimal voltage bump, getting to 900MHz though will require a good non-reference board and some boosting of the voltage".

(and comparable stuff regarding AMD cards of course)

I'd sticky that.
 
MSI afterburner for overclocking, kombostor for stability, and OCCT to validate your overclock. OCCT is great if you are going to be doing the stuff you are doing because it has an error check mode so it can make sure your video card is processing 100% error free at a given clock speed / voltage.

tviceman, couple of questions regarding Kombustor.

I see there are many rendering options for the stability test. DX9 or DX10 or DX11. As well as the AA mode choices.

Is there a "preferred" combination or a generally accepted combination in the forums for using Kombustor and producing results that one can rely on?

I went with the default - DX9 w/no AA - and set it to full screen while Afterburner and GPU-Z were running in the background.

Using afterburner I OC'ed to 800MHz (no voltage changes) and ran Kombustor for ~10min. Here's the result:

800MHzstockVccDX9Kombustor.jpg


Temps leveled off at 74C with fan speed of 3360rpms (85%). It was only barely noticable that the fan speed had increased from the standard desktop 2D idle (2 monitors here) fanspeed and temps.

Does that voltage (0.9750V) look good?

Provided I am comfortable with the noise/heat what are the "safe" voltage ranges for my card?
 
Maybe I just have a bum GTX460 when it comes to OC'ing.

The max OCCT stable clockspeed I could get at stock voltage (0.975V) was 790MHz.

At this MHz and voltage the GPU temp would level off at 77C.

Anything higher than 790MHz required a bump up in voltage to 1V.

The max GPU clockspeed that was OCCT stable at 1V was 815MHz and the GPU temp would level of at 83C.

I tried higher clocks with higher voltage combinations. 1.025V, 1.05V. I could not get OCCT stable. Even trying 835MHz at 1.025V and 1.05V would result in OCCT errors and then the room started smelling like hot electronics.

A couple of time the GPU must have hit some kind of current or thermal max trip and would reset itself to operate at only 400MHz until I rebooted the computer.

So it looks like I've got either a bum GTX460 as far as overclocking is concerned or my PSU is not up to the task.

My PSU is the Corsair 620HX and my APC UPS reports that even at full load with OCCT the PSU is only pulling 378W from the UPS.

One thing I noted about the MSI cyclone when I unpacked it was that the heatsink fins were not "level" or parallel to the PCB. One side is parallel but the other side is angled slightly down towards the PCB.

There was no visible cracks or marks on the heatpipes or heatfins to suggest it was bent by accident so I assumed it was intentionally angled down by design to add extra cooling to the capacitors that sit under the heatsink fins on that side.

At any rate, even at 815MHz and 1V I am happy with the performance of the card, both in terms of its lack of noise and lack of heat as well as in terms of absolute performance with Cuda.
 
Back
Top