Furmark Guide

[Victoria Steigerwald]

Forthe simple answer, we recommend using Furmark 2. Once downloaded and installed, you want to up the resolution to your screens and run the stress test. Giving it a good 30 minutes will give you a good idea if your GPU is in a good state.

When exploring how to stress test a GPU, Unigine Superposition emerges as a standout utility, offering a unique blend of realism and technical challenge. This advanced tool is designed to push video cards to their absolute limits, providing a clear picture of how they perform under extreme stress.

Unigine Superposition utilizes state-of-the-art graphics to simulate highly detailed environments, making it a popular choice for testing the capabilities of both gaming and professional-grade video cards. Its ability to create intensive 3D scenes allows users to accurately assess the stability and power of their GPUs, which is especially crucial for those engaged in overclocking or high-performance computing.

You can set a GPU temperature alarm using hardware monitoring software like HWInfo or MSI Afterburner. These tools allow you to set temperature thresholds that trigger an alert if exceeded during the test.

If you do need help with some more GPU troubles, you can find out how to overclock your GPU in our easy guide. Along with other stress test tools and how to test your GPU as a good continuation of getting more out of your system.

PCguide.com is part of the BGFG family of websites. Our mission is to provide you with great editorial and essential information to make your PC an integral part of your life. You can also follow PCguide.com on our social channels and interact with the team there.

2012 is an exciting time to need a graphic card upgrade. New PCI-Express 3 cards have hit the market and include the Nvidia GeForce GTX 600 series, such as the recent GTX 680, and AMD's Radeon HD 7000 Series, including ASUS designed cards such as the Radeon HD 7970 DirectCU II, 7950 DirectCU II and 7870 DirectCU II as well as their TOP pre-overclocked variants. While these cards fill the news and people's upgrade anticipations right now, you don't need PCI-Express 3.0 or even the latest series of graphics cards to overclock though - generally speaking any graphics card can be overclocked in some way. This guide explains the process of how to overclock your graphics card with GPU Tweak, and applies to Nvidia and AMD graphics cards alike.

A good stress test to push your overclock stability is FurMark, while GPU-Z is a must and is included with the ASUS GPU Tweak; however you can also download the latest version separately as an ROG Skinned GPU-Z.

Open your FurMark stress test - Select the profile you wish to run, I suggest you use the profile that will match as close to your gaming resolution. I used the 1920x1080 (1080p) setting. Remember the point of this exercise is to establish knowledge of the baseline stability of your graphic card at stock speeds. You can use any of the pre-set profiles too, as long as you are using the same settings each time we do a stability test.

We are not going to change any voltages in this basic guide until the very end. We now have the stand alone maximum memory frequency for our graphics card. It is actually unlikely that we will achieve this frequency when the GPU core is overclocked as well (due to reasons such as EMI, power draw, and induced heat from the GPU affecting the GDDR5 chips and PCB tracing, but what we do have is a known range within which we can work.

Now the moment of truth: putting them both together. Most of the time it will not be possible to get the max core frequency and the max memory frequency that we achieved separately to be stable at the same time, but we will try and get as close as possible.

Note that if you are specifically told to use HWInfo for game logging, ignore the steps pertaining to Furmark and Prime95 (i.e. follow steps 1, 4, 5 and 6). If told you are doing a stress test, you must download all 3 programs and follow all steps accordingly. None of the programs are optional during a stress test.

All 3 programs need to be running for at least 10 minutes, or until your PC crashes. Start the timer after you get furmark running for the best results. Ensure the Furmark window is on top, and do not use your computer during the 10 minutes.

i have passed through many devices before without issue but i admit i have never passedthrough a gpu - i followed the same prococess checked the vmware articles even loaded some oyuttube videos to ensure i hadnt missed anything but still cant seem to get any load on the gpu when i run something like furmark

anyone have any hints where im going wrong or where i need to be looking, i do see spikes in 100% activity whenver i open a program like nvidia geforce expierance or even furmark its self but as soon as i run any load the gpu is inactive.

So i just want to give a little background as to the use, to clear up any license issue, im running this in a home lab i simply want to be able to render video for video editing via the gpu. im not sure if i still fall foul of any license issue in that scenario?

i feel as if maybe svga vmware driver is still in play and simply wont hand over to the nvidia card, i tried disabling it from device manager to see if i could force card to be used as the nvidia but it didnt like that. from what im reading if it had worked, i shouldnt of got a console output at all and i would of had to interface directly with the graphics card on the server. (which is absolutley fine)

i have read another article linked above reddit few comments down, mentions that on the rtx3070 in a similiar scenario the hdmi had to by physically connected, in all my test thus far it has not been. so something i will try tomorrow. also mentions of a vmware paramter i didnt add which i will try: Add parameter: Key = hypervisor.cpuid.v0 Value = FALSE

I recently recycled a few RX580's I bought during the height of 2017-2018 coin mining craziness. Back then RX580's sold for around C$600, turning out at aronud 30-ishMH/s. At that rate, we figured that we'd break even with ETH price above $800. But maintaining many GPUs and have them run with high-enough stability is troublesome. We even got fire alarms, then there was a few instances that the PCIe riser or MOLEX power cable/connector just melted. The best for personal safety we stopped the mining process.

Since GCP (Google Cloud) is kind enough to be out of resource frequent enough, plus it is just impossible to get a hand on RTX 30xx card at non-ridiculous price at the moment, we figured it'd be a good idea to look at AMD's ROCm offering. Deep learning is a lot of fun, but it is not fun if we can't get GPUs. Hopefully this will do for now.

As of March 2021, ROCm support for PyTorch is pretty good. PyTorch even has added a beta-build for ROCm. But AMD magically decided to cut off support for ROCm at Vega. That means non of the pre-Polaris (Polaris, Ellesmere, Baffin, etc.) do not work for sure post ROCm 4.0. I also remember seeing posts about support being iffy at ROCm 3.8+. BUT ROCm is also NOT supported for the latest GPUs (big Navi, at the time of writing). AMD, please spent more time and money on supporting the deep learning community, so researchers are not subject to the mercy of nVIDIA.

If you follow the official AMD guide for ROCm installation, you will see issues installing PyTorch. Along the lines of "some binary cannot be found for current device", referring to the discountinued support. One such post is here. Here, gfx803 is RX580.

@karmic_koala those ICs look like they are specced for both 7gbps and 8gbps, which means they are most likely tightly binned parts so I would not automatically assume they will reach 8gbps if the stock speed is 7gbps, 7.6 - 7.8 is likely what you can expect to get out of 7gbps ICs.

EDIT: Did you ever try changing RASMACTRD, RASMACTWR, and WRPLUSRP? You can further iron out memory errors by raising these timings which despite their appearance are in nanoseconds, try 63, 72, 76 respectively.

EDIT: Did you ever try changing RASMACTRD, RASMACTWR, and WRPLUSRP? You can further iron out memory errors by raising these timings which despite their appearance are in nanoseconds, try 63, 72, 76 respectively.

@ket

thank you for the interesting guide.

I have a RX 570 with 4GB of SK Hynix vram:

I take both power limits and RAM speed could be approached in a slightly different manner.

without stealing your "magic", are there some specific aspects to be aware of?

TIA

in yellow I highlighted the values that are not in line with the guide: power and memory are lower to start with, so your 126W as minimum max power limit should be lowered.

the card has two Vbios(es) and my intention was to have a fully optimized Bios along with the original one.

@elisw to clarify a little more most 70s and 80s will have similar TDP/TDC settings the most likely occurrences of more significant changes will be with cheaply made versions and versions with one PCI-E connector instead of two. In your case you should be fine raising TDP up to 150w or 225w depending on if you have a card with just a 6 pin PCI-E connection or one with an 8pin. As it seems like you probably have a crappy VRM on the card optimising the amount of voltage the card uses will be vital in making sure the card throttles back as infrequently as possible. To have an idea at what the optimal voltage for your card is use GPU-Z to take a look at the cards ASIC quality, the higher it is the less voltage you will need. Typically 1.125v is the optimal voltage, to make the card use that for both manual and auto voltages you would input 65284 using PBE.

@elisw to clarify a little more most 70s and 80s will have similar TDP/TDC settings the most likely occurrences of more significant changes will be with cheaply made versions and versions with one PCI-E connector instead of two.

Ok. I did all the homework (up to the intermediate tuning) and managed fairly decent results. The "new" Hynix, although rated at lower frequency do go well with some OC (and do not suffer from the crp upper memory strap the "old" do). 2150 MHz for the vram seems stable (keep in mind the starting point is 1750, not 2000).

Unfortunately you have to select the Hynix timings (Memory timing level2) in wattman, if you leave it in Automatic it selects the crppy Elpida timings although GPU-Z recognise the chips as Hynix.

standard frequency Hynix timings:

3a8082e126