NVIDIA GeForce RTX 30 Series | RTX 3090 3080 3070 | Discussion

Yes, not to get a result but to see how stable the system is stressing out both GPU and CPU, that’s when I noticed that when I kick Intel utility in the GPU power comes down.

Am I being stupid?

 

Is GPU usage dropping slightly too? Is probably intel burn hogging CPU cycles and affecting furmark . Yes furmark is a GPU stress but all GPU workloads have some small CPU overhead for obv reasons

 

Nah not stupid I just wasn’t sure I was reading correctly all good, I guess what you need to also take into account is the intel burn test is more memory bound as apposed to CPU I think, I could be wrong with this tho

Nevermind looks like I was corrected! Lol

 

I would say your crashing issues was due to only using a single power cable.

Personally I dont consider furmark a good measure of stability. Its more a test of thermals, so your card could be creeping up in temperature and reducing clocks slightly.

For proper OC stability you want something that loads the GPU but also has intervals of low load where the fps shoots up high after a workload and thats where you get crashes.

 

Same with foxtel, the only way to get discounts, never pay full price for that ive gotten $30 discount per month for the last 3 years (have to ask again every 6 months)

 

I wouldn't call it stupid, but it's probably not a realistic load unless you're running some compute task at 100% on both CPU and GPU.

I bought the GPU for gaming, if it's stable in games then I'm happy. If there's an undetectable error then I don't care, only if it's affecting my experience. I use looped 3D Mark as a stability check and then tried it out in Control (+150 can handle hours of play, +160 MHz will crash to desktop after about 10 minutes). I think it's better to test it with a similar load to what you plan on doing with it.

I also use it for Folding@Home which does require it to be error free so I have another profile at +100Mhz and 75% power limit which is far more power efficient than stock but lower absolute performance.

 

This

I would be trying to run intel burn and something like heaven with a higher load or superposition on 4K optimised or even 8k to get it to render higher to stress the GPU

 

holy shit I've just been offered a STRIX OC instead of my TUF OC... I think I might jump on it...

 

just ordered the Bykski Colorful Block from AliExpress.

Block was $93.51 US, and had the following options for delivery

AliExpress Standard Delivery - Free -- expected arrival by 16/11/2020
AliExpress Express Delivery - 7-15 Days - $21 US
EXpress Delivery with DHL - 8-16 Days - $40 US

I went with DHL, as they have handled my EK products in the past, and i can usually pick the package up from their warehouse in Mt Waverley, and save 2-3 shipping! Also provides tracking one item leaves China.

total ended up being $145.48 US

Went to check out with Payl, and it was $298.41 AUD -- Go fiogure, more then double haha

So checked out with Debit Card, was charged $206.18 AUD -- can deal with that, especially since postage was big!

Now lets see if the 8-16 days delivery target is acheived!

 

Hey mate, as mentioned you have now loaded the CPU up so much it can't keep the GPU busy enough to also be 100% used. You could try task manager and lowering the priority of the CPU test if you want to hit max loads.

 

You could try prime 95 instead of IBT, you can choose how many threads to launch so could leave a bit of CPU free for your graphics test or benchmark.

 

I just recently put in an order for a bunch of barrow flexi fittings and things on AliExpress, I was like a kid in a candy store. So many colours, every length and size fitting you can think of for peanuts. I need to go back to get an AM4 CPU block, 3080 ref GPU block and mini res/pump combo.

 

Lol. You're getting ripped.

I was paying $28/month for everything including IQ boxen.
Don't go making a noise on social media, speak to them quietly and save $170/month.

 

saving 170???? my full price bill is only 139

 

Cheers for all the tips guys. It seems that having a separate power cable for each PCI-E header has fixed the issue. I just ran a game for ~25mins with +50 core and +500 mem with no issues. Usually after 10mins or so I'm gone. I had the fans cranked on 100% and GPU temp never went above 60c. Open hardware monitor showed peak GPU power of 350W. CPU was running at stock, so now I gotta try bring that up a bit.

 

Not intending to single you out. There are a lot of other people doing it too.

Why do people daisychain power cables to the GPU?
It turned out that a lot of people who were getting black screen crashes on their 5700xt's were doing exactly the same thing.
There's a reason none of the reviewers were having problems.

The only way you could get away with this without problems is if you have a single rail psu and I'm not sure anyone makes them anymore.

Apart from anything else, there's a high chance of melting the cable/termination and killing your ultra expensive gpu.
350-450w @12v is a LOT of power.

 

2 Reasons - 1 is aesthetics. Having additional PCIE plugs hanging off the card looks crap so people will actually go out of their way to see if 1 cable works. 2 is the assumption that if it has 2 plugs it should support the power of two plugs.

 

If you have never encountered problems with power delivery before then it's understandable. It's neater to have just one cable and there are 2 connectors on it so it makes sense that it should work.

 

All good, if someone learns from it then its a good thing. It wasn't daisy chained as such - each PCI-E connector with my PSU has 2x 8 pin PCI-E connectors hanging off the end of it. To any noob like myself you'd think that means you can use both, but I suppose when you have the latest hardware you start to run into certain bottlenecks.

 

Why's that? I mean in terms of them switching you...