Is my brand new Gigabyte G1 gaming GTX970 a lemon (no, it's geat. I had power saving enabled in Windows 10 and the CPU was throttling the results)

[d510r9]

** EDIT **

Windows 10 was running in power saver mode. Under Power Options was a "hidden" additional plan High Performance which I selected. 

New result: 10,681 (+6% average G1 gaming GTX970)

New O/C result: 12,063

 

The card is great, I'm the lemon.

 

** EDIT ** 

 

 

 

I got a Gigabyte GTX 970 G1 Gaming GPU from Amazon today (it's a rev 1.1) and have been running some benchmarks and it looks like it under-performs a regular 970. Running 3dMark (Fire Strike), from what I can tell you would expect these kinds of averages:

• 
  

  Vanilla 970: 9,568

  
  
  
• 
  

  G1 gaming 970: 10,046

  
  
  
• 
  

  G1 gaming 970 OC: 11,427

  
  
  

 

My scores range from 8,141 to 9,155 (at its worst 20% slower than a regular g1 gaming card). On 3D mark it shows my score under-performing a lot of other systems 3dMark 8,828. Running Heaven and comparing my results to similar system builds/settings I also get low scores (1,209) and min/max FPS (7/90).

 

Heaven settings:

• 
  

  Render: Direct3D11

  
  
  
• 
  

  Mode: 1920x1080 8xAA fullscreen

  
  
  
• 
  

  Quality: Ultra

  
  
  
• 
  

  Tessellation: Extreme

  
  
  

 

I don't think another part of the system is a bottleneck as the rest of the components are pretty decent:

• 
  

  Video Card: Gigabyte GeForce GTX 970 G1 Gaming 4GB

  
  
  
• 
  

  Drivers: 358.50

  
  
  
• 
  

  CPU: Intel Core i7-4790K 4.0GHz (and Noctua NH-U12S Cooler)

  
  
  
• 
  

  Motherboard: Gigabyte GA-Z97X-SOC

  
  
  
• 
  

  Memory: Kingston HyperX Fury White 16GB (2 x 8GB) DDR3-1600 Memory

  
  
  
• 
  

  Storage: Samsung 840 EVO 500GB 2.5" Solid State Drive

  
  
  
• 
  

  Power Supply: Corsair RM 850W 80+ Gold Certified Fully-Modular ATX

  
  
  
• 
  

  OS: Windows 10 (64-bit)

  
  
  

 

I read the nVidia experience stuff could drop the card into a non-performance mode/2d or something like that so I did a clean install of only the (latest) drivers.

 

Do you think I have drawn a crappy card, or is this kind of deviation from the average normal? Another theory I had was that first release window cards may be better than a card a year later as they think most enthusiasts buy early? It's from Amazon and they're great at returning no questions asked so no problems to RMA it.

 

I'm not worried about getting a different brand/model if anyone thinks another 970 is a better option. I REALLY hate noisy fans, though when not gaming I can set the bios to not spin say sub 40C). My last build was about 13 years ago (lol!) and it had a Radeon 9700, so really wanted to get an nVidia card this time around.

 

Many thanks for your help/advice

[Dabombinable]

 

I got a Gigabyte GTX 970 G1 Gaming GPU from Amazon today (it's a rev 1.1) and have been running some benchmarks and it looks like it under-performs a regular 970. Running 3dMark (Fire Strike), from what I can tell you would expect these kinds of averages:

• Vanilla 970: 9,568

• G1 gaming 970: 10,046

• G1 gaming 970 OC: 11,427

 

My scores range from 8,141 to 9,155 (at its worst 20% slower than a regular g1 gaming card). On 3D mark it shows my score under-performing a lot of other systems 3dMark 8,828. Running Heaven and comparing my results to similar system builds/settings I also get low scores (1,209) and min/max FPS (7/90).

 

Heaven settings:

• Render: Direct3D11

• Mode: 1920x1080 8xAA fullscreen

• Quality: Ultra

• Tessellation: Extreme

 

I don't think another part of the system is a bottleneck as the rest of the components are pretty decent:

• Video Card: Gigabyte GeForce GTX 970 G1 Gaming 4GB

• Drivers: 358.50

• CPU: Intel Core i7-4790K 4.0GHz (and Noctua NH-U12S Cooler)

• Motherboard: Gigabyte GA-Z97X-SOC

• Memory: Kingston HyperX Fury White 16GB (2 x 8GB) DDR3-1600 Memory

• Storage: Samsung 840 EVO 500GB 2.5" Solid State Drive

• Power Supply: Corsair RM 850W 80+ Gold Certified Fully-Modular ATX

• OS: Windows 10 (64-bit)

 

I read the nVidia experience stuff could drop the card into a non-performance mode/2d or something like that so I did a clean install of only the (latest) drivers.

 

Do you think I have drawn a crappy card, or is this kind of deviation from the average normal? Another theory I had was that first release window cards may be better than a card a year later as they think most enthusiasts buy early? It's from Amazon and they're great at returning no questions asked so no problems to RMA it.

 

I'm not worried about getting a different brand/model if anyone thinks another 970 is a better option. I REALLY hate noisy fans, though when not gaming I can set the bios to not spin say sub 40C). My last build was about 13 years ago (lol!) and it had a Radeon 9700, so really wanted to get an nVidia card this time around.

 

Many thanks for your help/advice

 

Go to the Nvidia control panel, manage 3D settings, and change the "Power management mode" to Prefer maximum performance, and see if that makes a difference. Also at around 60-70% fan speeds the card starts to get noisy (aka the loudest part of my rig) if mine is anything to go by

[Majestic]

That combined score is really weak.

 

 

Go to the Nvidia control panel, manage 3D settings, and change the "Power management mode" to Prefer maximum performance, and see if that makes a difference. Also at around 60-70% fan speeds the card starts to get noisy (aka the loudest part of my rig) if mine is anything to go by

 

No, the GPU score is fine. The Combined score is weak.

 

CPU score is weak as hell for a 4790K. I think something is wrong with that. Make sure it's running at the correct clockspeed, i think it's throttling.

[Dabombinable]

 

That combined score is really weak.

 

 

 

No, the GPU score is fine. The Combined score is weak.

 

My GTX 970 G1 Gaming when I last ran benchmarked it scored a lot higher.....

[Synawke]

I have almost the same set up as you, i have a EVGA 970 though and get a score of in the 1300s stock with about a 10/115 fps difference. when its over clocked i get 1400+ with about a 30-140 fps difference

[iamdarkyoshi]

my system benchmarks lower on everything too. I think its windows 10.

[Zayzo]

Same benchmark on Windows 10 with clocks at 1600/2100. (My card) 

What are your cpu temps? 

[Dabombinable]

Same benchmark on Windows 10 with clocks at 1600/2100. (My card) 

First things first, your still using Windows 8? And it seems that the OS after Windows 7 are scoring lower. Or at least 8 and 10 are.

[Zayzo]

First things first, your still using Windows 8? And it seems that the OS after Windows 7 are scoring lower. Or at least 8 and 10 are.

That's a score on 10. And i'm still using 10. 

[Majestic]

http://www.3dmark.com/3dm/8839075?

 

Look, i have the same combined score with a GTX 960 and a 4670K (4ghz, so it's about the same clockspeeds). I think it has nothing to do with win10. Use msi afterburner or something to check GPU load during the combined test. If it's not hitting 99% (like mine) there is something wrong with the CPU during the test. Probably some throttling. If it is, then the GPU is returning to some sort of 2D mode.

[Dabombinable]

That's a score on 10. And i'm still using 10. 

Windows NT 6.2 is supposed to be Windows 8, with NT 6.3 being 8.1, and Windows 10 being NT 10.

[Zayzo]

Windows NT 6.2 is supposed to be Windows 8, with NT 6.3 being 8.1, and Windows 10 being NT 10.

I can run it again if you want. But it was on windows 10. I haven't had windows 8 for so long. I still have the screenshot in my heaven folder. 

[Dabombinable]

I can run it again if you want. But it was on windows 10. I haven't had windows 8 for so long. 

Don't worry, Unigine Heaven is just being Unigine Heaven

[Zayzo]

Don't worry, Unigine Heaven is just being Unigine Heaven

I think the OPs issue is a cpu issue. If its on the stock cooler its probably throttling. I've seen countless threads on here where people with the 4790k are getting crazy high temps on the 4790k with stock coolers. 

[d510r9]

Go to the Nvidia control panel, manage 3D settings, and change the "Power management mode" to Prefer maximum performance, and see if that makes a difference. Also at around 60-70% fan speeds the card starts to get noisy (aka the loudest part of my rig) if mine is anything to go by

 

 

Thanks mate, just changed it - I thought I avoided this default by installing the drivers directly.

I just re-ran 3dmark. It's a bit higher (8,739) but still pretty low.

 

Next step will look at the CPU settings like Majestic was saying.

[d510r9]

First things first, your still using Windows 8? And it seems that the OS after Windows 7 are scoring lower. Or at least 8 and 10 are.

Yeah I'm running Windows 10. Maybe that's the issue?

[iamdarkyoshi]

Yeah I'm running Windows 10. Maybe that's the issue?

my 390x on windows 10 performs about as well as an hd7950 with windows 8 on passmark...

[Dabombinable]

Yeah I'm running Windows 10. Maybe that's the issue?

Probably. As with every previous versions of Windows it is loaded with bugs and issues after launch. TBH people should only be upgrading from Windows 7 and 8.1 about a month or 2 before the free upgrade period ends-that way you get a version of it with far fewer issues.

[d510r9]

http://www.3dmark.com/3dm/8839075?

 

Look, i have the same combined score with a GTX 960 and a 4670K (4ghz, so it's about the same clockspeeds). I think it has nothing to do with win10. Use msi afterburner or something to check GPU load during the combined test. If it's not hitting 99% (like mine) there is something wrong with the CPU during the test. Probably some throttling. If it is, then the GPU is returning to some sort of 2D mode.

 

Ok, just running Heaven in the background and can see the GPU does hit up to 99%, jumps around a fair bit but looks to hit 99% pretty frequently.

 

The CPU sits around 43C, min 24C, max 68C and utilization jumping between 20-70%.

[Dabombinable]

Ok, just running Heaven in the background and can see the GPU does hit up to 99%, jumps around a fair bit but looks to hit 99% pretty frequently.

 

The CPU sits around 43C, min 24C, max 68C and utilization jumping between 20-70%.

Yeah. Its definately got to be Windows 10 if the temps are fine and the CPU and GPU utilizations are the way they should be.

[d510r9]

I think the OPs issue is a cpu issue. If its on the stock cooler its probably throttling. I've seen countless threads on here where people with the 4790k are getting crazy high temps on the 4790k with stock coolers. 

 

I'm actually using a Noctua NH-U12S Cooler (got it mainly for the quiet fans). It's seems to be keeping CPU temps down. While running Heaven it sits around 45C.

[d510r9]

 

That combined score is really weak.

 

 

 

No, the GPU score is fine. The Combined score is weak.

 

CPU score is weak as hell for a 4790K. I think something is wrong with that. Make sure it's running at the correct clockspeed, i think it's throttling.

 

 

Sorry I'm a bit clueless on some of these things. This is a dump from the report file:

 

 

Processor 1 ID = 0

Number of cores 4 (max 8)

Number of threads 8 (max 16)

Name Intel Core i7 4790K

Codename Haswell

Specification Intel® Core i7-4790K CPU @ 4.00GHz

Package (platform ID) Socket 1150 LGA (0x1)

CPUID 6.C.3

Extended CPUID 6.3C

Core Stepping C0

Technology 22 nm

TDP Limit 88.0 Watts

Tjmax 100.0 °C

Core Speed 799.8 MHz

Multiplier x Bus Speed 8.0 x 100.0 MHz

Stock frequency 4000 MHz

Instructions sets MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, EM64T, VT-x, AES, AVX, AVX2, FMA3

L1 Data cache 4 x 32 KBytes, 8-way set associative, 64-byte line size

L1 Instruction cache 4 x 32 KBytes, 8-way set associative, 64-byte line size

L2 cache 4 x 256 KBytes, 8-way set associative, 64-byte line size

L3 cache 8 MBytes, 16-way set associative, 64-byte line size

FID/VID Control yes

 

 

Turbo Mode supported, enabled

Max non-turbo ratio 40x

Max turbo ratio 44x

Max efficiency ratio 8x

O/C bins unlimited

Ratio 1 core 44x

Ratio 2 cores 44x

Ratio 3 cores 44x

Ratio 4 cores 44x

TSC 4000.0 MHz

APERF 800.0 MHz

MPERF 4000.2 MHz

IA Voltage Mode PCU adaptive

IA Voltage Offset 0 mV

GT Voltage Mode PCU adaptive

GT Voltage Offset 0 mV

LLC/Ring Voltage Mode PCU adaptive

LLC/Ring Voltage Offset 0 mV

Agent Voltage Mode PCU adaptive

Agent Voltage Offset 0 mV

 

Temperature 0 47°C (116°F) [0x35] (Core #0)

Temperature 1 48°C (118°F) [0x34] (Package)

Power 0 1.06 W (Package)

Power 1 n.a. (IA Cores)

Power 2 n.a. (GT)

Power 3 n.a. (Uncore)

Power 4 7.70 W (DRAM)

Voltage 0 0.70 Volts (VID)

Voltage 1 +0.00 Volts (IA Offset)

Voltage 2 +0.00 Volts (GT Offset)

Voltage 3 +0.00 Volts (LLC/Ring Offset)

Voltage 4 +0.00 Volts (System Agent Offset) 

[Majestic]

Run a Prime95 (version 25.11 64bit) and check your CPU frequency with CPu-Z.

[d510r9]

Run a Prime95 (version 25.11 64bit) and check your CPU frequency with CPu-Z.

 

Hey Majestic, thanks again for your help mate. Prepare yourself for a tall glass of stupid. I'm really new to the whole o/c thing and also a bit of an idiot.

 

The CPU bottleneck you identified led me to download and run PassMark Performance Test 8.0. Before it started it warned I was running Windows 10 in power saver mode (I'm pretty sure I didn't set this and defaulted on install. I would think this would only happen if it thought you were using a laptop but hey ho). Under Power Options is a "hidden" additional plan High Performance which I selected.

 

 

New result: 10,681 (+6% average G1 gaming GTX970)

 

 

I then O/C'd the GPU in MSi Afterburner using some other G1 O/C's as a guide and had no issues with stability/artifacts etc

• Power limit: 112%
• Core Clock: +161 (1,490MHz)
• Memory Clock: +800 (4,307MHz) I still don't understand what the correct reading of this should be CPU-Z and MSi both show 4,307MHz, GPU-Z it shows half that figure 2,153, and somewhere else it showed 8k? 

I also ran auto-tune for my motherboard to O/C the CPU and it selected 4.4GHz (light = 4.5GHz, Medium = 4.6GHz, Extreme = 4.7GHz). An aside, the CPU temp briefly spiked to about 91C, was mostly around 73C. A little bit scary so I don't think I'll be O/C that any time soon again, or is a light O/C safe with my Noctua cooler?

 

 

New O/C result: 12,063

• +26% vanilla 970
• +20% non o/c G1 970
• +6% average o/c G1 gaming
• +8% 980
• +2.2% non o/c G1 gaming 980

So my GPU is great, and I'm a lemon. What a surprise!

 

 

Thanks to everyone for all the quick responses and all the help. Really cool.

[Majestic]

Hey Majestic, thanks again for your help mate. Prepare yourself for a tall glass of stupid. I'm really new to the whole o/c thing and also a bit of an idiot.

 

The CPU bottleneck you identified led me to download and run PassMark Performance Test 8.0. Before it started it warned I was running Windows 10 in power saver mode (I'm pretty sure I didn't set this and defaulted on install. I would think this would only happen if it thought you were using a laptop but hey ho). Under Power Options is a "hidden" additional plan High Performance which I selected.

 

Yeah, but you don't want to use High performance 24/7, as it stops the chip from clocking back down during idle. Balanced is what you want, that still gives you max boost but also power saving options.

 

Yeah, it was either power saving features or disabled speedstepping. It's pretty amazing though what you can achieve with a haswell chip running 1ghz