ASUS Strix Nvidia GTX960 4GB - Same 4GB problem as GTX 970

UnknownzD · Mar 26, 2015

So I have just got the brand new GTX 960 4GB ASUS in hand, with the BIOS submitted to the techpowerup database. However, I have just found out that the 4GB card is designed in the same way as GTX 970 does, with a lower bandwidth provided over 3.5 GB. And here is the result.

Code:

Nai's Cache Size Benchmark
DISCLAIMER:
This Benchmark tries to roughly estimate the L2 cache size in
CUDA by benchmarking memory latencies for differently sized
working sets and different chunks of global memory.
Use it without anything in the DRAM of your GPU or else
the swapping behaviour of the GPU may corrupt the measurement.
If the benchmark produces strange outputs nevertheless,
there is a high proability that this benchmark is not working
as intended. Your GPU is probably just fine. So please stop
making annoying whine posts in any forums, if this benchmark
produces a suspicous output.
Press any key to continue . . .
Device name: GeForce GTX 960
Device memory size: 4096 MiByte
Chunk Size: 128 MiByte
Allocated 30 Chunks
Allocated 3840 MiByte
Benchmarking L2 cache size
L2 cache size of chunk no. 0 (0 MiByte to 128 MiByte): 1024 kiByte
L2 cache size of chunk no. 1 (128 MiByte to 256 MiByte): 1024 kiByte
L2 cache size of chunk no. 2 (256 MiByte to 384 MiByte): 1024 kiByte
L2 cache size of chunk no. 3 (384 MiByte to 512 MiByte): 1024 kiByte
L2 cache size of chunk no. 4 (512 MiByte to 640 MiByte): 1024 kiByte
L2 cache size of chunk no. 5 (640 MiByte to 768 MiByte): 1024 kiByte
L2 cache size of chunk no. 6 (768 MiByte to 896 MiByte): 1024 kiByte
L2 cache size of chunk no. 7 (896 MiByte to 1024 MiByte): 1024 kiByte
L2 cache size of chunk no. 8 (1024 MiByte to 1152 MiByte): 1024 kiByt
L2 cache size of chunk no. 9 (1152 MiByte to 1280 MiByte): 1024 kiByt
L2 cache size of chunk no. 10 (1280 MiByte to 1408 MiByte): 1024 kiBy
L2 cache size of chunk no. 11 (1408 MiByte to 1536 MiByte): 1024 kiBy
L2 cache size of chunk no. 12 (1536 MiByte to 1664 MiByte): 1024 kiBy
L2 cache size of chunk no. 13 (1664 MiByte to 1792 MiByte): 1024 kiBy
L2 cache size of chunk no. 14 (1792 MiByte to 1920 MiByte): 1024 kiBy
L2 cache size of chunk no. 15 (1920 MiByte to 2048 MiByte): 1024 kiBy
L2 cache size of chunk no. 16 (2048 MiByte to 2176 MiByte): 1024 kiBy
L2 cache size of chunk no. 17 (2176 MiByte to 2304 MiByte): 1024 kiBy
L2 cache size of chunk no. 18 (2304 MiByte to 2432 MiByte): 1024 kiBy
L2 cache size of chunk no. 19 (2432 MiByte to 2560 MiByte): 1024 kiBy
L2 cache size of chunk no. 20 (2560 MiByte to 2688 MiByte): 1024 kiBy
L2 cache size of chunk no. 21 (2688 MiByte to 2816 MiByte): 1024 kiBy
L2 cache size of chunk no. 22 (2816 MiByte to 2944 MiByte): 1024 kiBy
L2 cache size of chunk no. 23 (2944 MiByte to 3072 MiByte): 1024 kiBy
L2 cache size of chunk no. 24 (3072 MiByte to 3200 MiByte): 1024 kiBy
L2 cache size of chunk no. 25 (3200 MiByte to 3328 MiByte): 1024 kiBy
L2 cache size of chunk no. 26 (3328 MiByte to 3456 MiByte): 1024 kiBy
L2 cache size of chunk no. 27 (3456 MiByte to 3584 MiByte): 1024 kiBy
Error estimating L2 cache size of chunk no. 28 (3584 MiByte to 3712 M
ably because of swapping!
Latency for the smallest working set: 0.000267 ms
Latency for the largest working set: 0.000268 ms
L2 cache size of chunk no. 29 (3712 MiByte to 3840 MiByte): 1024 kiBy
Benchmarking DRAM
0 MiByte to 128 MiByte: 88.68 GByte/s Read, 86.64 GByte/s Write
128 MiByte to 256 MiByte: 88.68 GByte/s Read, 86.24 GByte/s Write
256 MiByte to 384 MiByte: 88.65 GByte/s Read, 85.85 GByte/s Write
384 MiByte to 512 MiByte: 88.71 GByte/s Read, 86.62 GByte/s Write
512 MiByte to 640 MiByte: 88.68 GByte/s Read, 86.14 GByte/s Write
640 MiByte to 768 MiByte: 88.63 GByte/s Read, 85.81 GByte/s Write
768 MiByte to 896 MiByte: 88.65 GByte/s Read, 86.65 GByte/s Write
896 MiByte to 1024 MiByte: 88.71 GByte/s Read, 86.10 GByte/s Write
1024 MiByte to 1152 MiByte: 88.65 GByte/s Read, 85.85 GByte/s Write
1152 MiByte to 1280 MiByte: 88.71 GByte/s Read, 86.65 GByte/s Write
1280 MiByte to 1408 MiByte: 88.72 GByte/s Read, 86.34 GByte/s Write
1408 MiByte to 1536 MiByte: 88.64 GByte/s Read, 85.87 GByte/s Write
1536 MiByte to 1664 MiByte: 88.71 GByte/s Read, 86.65 GByte/s Write
1664 MiByte to 1792 MiByte: 88.58 GByte/s Read, 86.31 GByte/s Write
1792 MiByte to 1920 MiByte: 88.62 GByte/s Read, 85.94 GByte/s Write
1920 MiByte to 2048 MiByte: 88.71 GByte/s Read, 86.64 GByte/s Write
2048 MiByte to 2176 MiByte: 88.56 GByte/s Read, 86.30 GByte/s Write
2176 MiByte to 2304 MiByte: 88.66 GByte/s Read, 85.94 GByte/s Write
2304 MiByte to 2432 MiByte: 88.70 GByte/s Read, 86.61 GByte/s Write
2432 MiByte to 2560 MiByte: 88.56 GByte/s Read, 86.27 GByte/s Write
2560 MiByte to 2688 MiByte: 88.66 GByte/s Read, 85.94 GByte/s Write
2688 MiByte to 2816 MiByte: 88.71 GByte/s Read, 86.62 GByte/s Write
2816 MiByte to 2944 MiByte: 88.59 GByte/s Read, 86.29 GByte/s Write
2944 MiByte to 3072 MiByte: 88.61 GByte/s Read, 85.98 GByte/s Write
3072 MiByte to 3200 MiByte: 88.70 GByte/s Read, 86.60 GByte/s Write
3200 MiByte to 3328 MiByte: 88.56 GByte/s Read, 86.29 GByte/s Write
3328 MiByte to 3456 MiByte: 88.63 GByte/s Read, 86.06 GByte/s Write
3456 MiByte to 3584 MiByte: 14.00 GByte/s Read, 15.35 GByte/s Write
3584 MiByte to 3712 MiByte:  7.57 GByte/s Read,  8.42 GByte/s Write
3712 MiByte to 3840 MiByte:  9.48 GByte/s Read, 10.50 GByte/s Write
Benchmarking L2 cache
0 MiByte to 128 MiByte: 278.67 GByte/s Read, 284.78 GByte/s Write
128 MiByte to 256 MiByte: 278.70 GByte/s Read, 284.94 GByte/s Write
256 MiByte to 384 MiByte: 278.77 GByte/s Read, 284.93 GByte/s Write
384 MiByte to 512 MiByte: 278.76 GByte/s Read, 285.01 GByte/s Write
512 MiByte to 640 MiByte: 278.85 GByte/s Read, 285.13 GByte/s Write
640 MiByte to 768 MiByte: 278.78 GByte/s Read, 285.19 GByte/s Write
768 MiByte to 896 MiByte: 278.85 GByte/s Read, 284.98 GByte/s Write
896 MiByte to 1024 MiByte: 278.72 GByte/s Read, 284.93 GByte/s Write
1024 MiByte to 1152 MiByte: 278.73 GByte/s Read, 284.89 GByte/s Write
1152 MiByte to 1280 MiByte: 278.83 GByte/s Read, 284.85 GByte/s Write
1280 MiByte to 1408 MiByte: 278.78 GByte/s Read, 284.88 GByte/s Write
1408 MiByte to 1536 MiByte: 278.69 GByte/s Read, 284.79 GByte/s Write
1536 MiByte to 1664 MiByte: 278.76 GByte/s Read, 284.80 GByte/s Write
1664 MiByte to 1792 MiByte: 278.82 GByte/s Read, 284.73 GByte/s Write
1792 MiByte to 1920 MiByte: 278.64 GByte/s Read, 284.82 GByte/s Write
1920 MiByte to 2048 MiByte: 278.82 GByte/s Read, 284.80 GByte/s Write
2048 MiByte to 2176 MiByte: 278.71 GByte/s Read, 285.25 GByte/s Write
2176 MiByte to 2304 MiByte: 278.85 GByte/s Read, 284.73 GByte/s Write
2304 MiByte to 2432 MiByte: 278.58 GByte/s Read, 284.64 GByte/s Write
2432 MiByte to 2560 MiByte: 278.56 GByte/s Read, 285.08 GByte/s Write
2560 MiByte to 2688 MiByte: 278.82 GByte/s Read, 284.89 GByte/s Write
2688 MiByte to 2816 MiByte: 278.77 GByte/s Read, 285.10 GByte/s Write
2816 MiByte to 2944 MiByte: 278.95 GByte/s Read, 284.64 GByte/s Write
2944 MiByte to 3072 MiByte: 278.75 GByte/s Read, 285.56 GByte/s Write
3072 MiByte to 3200 MiByte: 278.65 GByte/s Read, 285.10 GByte/s Write
3200 MiByte to 3328 MiByte: 278.78 GByte/s Read, 284.69 GByte/s Write
3328 MiByte to 3456 MiByte: 278.83 GByte/s Read, 284.72 GByte/s Write
3456 MiByte to 3584 MiByte:  7.31 GByte/s Read,  8.55 GByte/s Write
3584 MiByte to 3712 MiByte:  7.32 GByte/s Read,  8.54 GByte/s Write
3712 MiByte to 3840 MiByte: 24.47 GByte/s Read, 28.34 GByte/s Write
Press any key to continue . . .

And as a result, I am going to RMA the card because I just need to pay $100 more to get a more juicy 970 card instead of this crappy card. Actually I would expect it to be running with full speed for all memory segment for this new release, but apparently it is not.

MxPhenom 216 · Mar 26, 2015

I dont know why anyone would by a 4gb 960 anyways. That extra 2Gb is not helping the card in any way.

UnknownzD · Mar 26, 2015

Because I would expect that GTX 960 does not have such a problem as a class-level lawsuit was performed against Nvidia for their GTX 970 release. But apparently I am wrong and they have just done this again. I know that the bandwidth of this card is limited, but I would expect that it will not suffer the laggy problem as GTX970 does during memory intensive situation. I am very disappointed with the marketing claims made by Nvidia.

MxPhenom 216 · Mar 26, 2015

UnknownzD said:
Because I would expect that GTX 960 does not have such a problem as a class-level lawsuit was performed against Nvidia for their GTX 970 release. But apparently I am wrong and they have just done this again. I know that the bandwidth of this card is limited, but I would expect that it will not suffer the laggy problem as GTX970 does during memory intensive situation. I am very disappointed with the marketing claims made by Nvidia.

I hope you realize that the 4gb 960 are not by spec from Nvidia. They are 2GB cards cards by reference spec from Nvidia. Any more memory added to the card is done by their vendors.

It seems to me the normal reference 2GB 960s have no memory issues.

UnknownzD · Mar 26, 2015

Oh seriously? I have never expected that the card was developed without the help from Nvidia at all.

MxPhenom 216 · Mar 26, 2015

UnknownzD said:
Oh seriously? I have never expected that the card was developed without the help from Nvidia at all.

Board partners are allowed to do what they want with their cards at discretion of Nvidia. For instance, no one is allowed to do non reference designs of the Titan line of cards, but the normal lower tier Geforce cards, board partners can do whatever they want. circuit boards, power delivery, memory capacity, coolers, etc.

UnknownzD · Mar 26, 2015

Well I doubt it, because allowing the board partners to do whatever that they want may damage the brand image made by Nvidia at the end. Anyway, I have recognized that this may be an ASUS-specific problem but, according to the consumer law here, you should explicitly state every down side of a product, without the intention to hide it at all. I really hate to see that the card does not run in full speed for all memory segment.

MxPhenom 216 · Mar 26, 2015

UnknownzD said:
Well I doubt it, because allowing the board partners to do whatever that they want may damage the brand image made by Nvidia at the end. Anyway, I have recognized that this may be an ASUS-specific problem but, according to the consumer law here, you should explicitly state every down side of a product, without the intention to hide it at all. I really hate to see that the card does not run in full speed for all memory segment.

that is why I said discretion of Nvidia. NVIDIA has the ultimate say on if the vendors can do non reference designs or not. But if they can, its really up to the vendors on what they want to do. After all they are the ones selling NVIDIA GPUs, and making NVIDIA money.

rtwjunkie · Mar 26, 2015

UnknownzD said:
Well I doubt it, because allowing the board partners to do whatever that they want may damage the brand image made by Nvidia at the end.

This has been Nvidia's lenient policy for many years. Check with Nvidia if you don't believe it. Why do you think you get so many different types of the same model cards? As an example, remember the GTX 660? Not that long ago it was the sweet spot between performance and price. MSI's TF OC was one of the best, per W1zzard's own testing. The PCB on it? They used the PCB from the GTX 680. Hardly the spec Nvidia envisioned, because the native PCB spec was a much smaller board. It may actually have been part of the reason that card outperformed other 660 cards that were clocked the same and higher.

UnknownzD · Mar 26, 2015

Well I honestly thank you for your help, and notified me something like this.

However, I still can't accept that the GRAM cannot be running under full speed even when 4GB of memory is available, and for such a reason, I may go over to GTX 980 instead. I am still waiting for the reply from the local distributor.

RealNeil · Mar 26, 2015

rtwjunkie said:
This has been Nvidia's lenient policy for many years. Check with Nvidia if you don't believe it. Why do you think you get so many different types of the same model cards? As an example, remember the GTX 660? Not that long ago it was the sweet spot between performance and price. MSI's TF OC was one of the best, per W1zzard's own testing. The PCB on it? They used the PCB from the GTX 680. Hardly the spec Nvidia envisioned, because the native PCB spec was a much smaller board. It may actually have been part of the reason that card outperformed other 660 cards that were clocked the same and higher.

This is why I always wait for Partner GPUs to come out. They're the reference design with sweet little tweaks.

AsRock · Mar 26, 2015

Kinda figured that it be some thing of the same deal like the 970. RMA because of ?, the 3.5GB thing OM if so as your on about getting 970 so your still going have the 3.5GB.

MxPhenom 216 · Mar 26, 2015

AsRock said:
Kinda figured that it be some thing of the same deal like the 970. RMA because of ?, the 3.5GB thing OM if so as your on about getting 970 so your still going have the 3.5GB.

On the 960 the card is going to run out of GPU grunt before memory capacity even being an issue though.

UnknownzD · Mar 26, 2015

That's why I am considering to choose GTX 980 instead.

What I meant is that the car is definitely not worth it, as I was buying the car to avoid the problem experienced by GTX 970 owners, while hoping that the 4GB GRAM would help to improve the textures by having a larger GRAM, but apparently I am wrong.

MxPhenom 216 · Mar 26, 2015

UnknownzD said:
That's why I am considering to choose GTX 980 instead.

What I meant is that the car is definitely not worth it, as I was buying the car to avoid the problem experienced by GTX 970 owners, while hoping that the 4GB GRAM would help to improve the textures by having a larger GRAM, but apparently I am wrong.

also that benchmark utility has been debunked and the author of it has admitted to that it in no way is anything like a real world application and how those use memory and "swapping"

UnknownzD · Mar 26, 2015

So I have just received a phone call from local distributor, and they said that ASUS has refused to perform any upgrade for customers experiencing such a problem, even if the last 500MB is not running with the GDDR5 standard, which is a violation of the consumer law over here, and therefore I guess I will never buy anything from ASUS again. Apparently EVGA and AsRock are producing better product in the recently year, and I guess I will move on to EVGA instead after selling this card out.

I am not blaming on the product design made by ASUS, but rather blaming on the service quality provided by ASUS, as they refuse to admit that this is somehow their problem as well. May go over to AMD as well if a DX12 card from AMD is out.

JL87 · Mar 26, 2015

Take it for what you will, but this link has a response from the author of Nai's Benchmark in which he responds saying that his code was not meant to test the Vram as it's been used by people on the internet to expose the 3.5 m issue with the 970s.

https://devtalk.nvidia.com/default/...t-and-nais-benchmark-is-somewhat-suspicious-/

MxPhenom 216 · Mar 26, 2015

JL87 said:
Take it for what you will, but this link has a response from the author of Nai's Benchmark in which he responds saying that his code was not meant to test the Vram as it's been used by people on the internet to expose the 3.5 m issue with the 970s.

https://devtalk.nvidia.com/default/...t-and-nais-benchmark-is-somewhat-suspicious-/

This is all I needed out of that link. Thanks for it!

Hi

@Skybuck

I'm the "German programmer", who initially has programmed this benchmark. As I've already stated in this thread several times and immediately after posting the download: This benchmark isn't eligible for benchmarking the VRAM-bug. You're just benchmarking the swapping behaviour of the global memory space in CUDA. However some people spread this benchmark all around the internet, without reading, understanding or knowing its issues. Even some major news sites disgraced themselves by using this benchmark without investigating it further.

There's also a more recent version of the source code hidden within the spoiler down the page. Furthermore this benchmark isn't very well designed, since I've written it within 20 minutes and I've never dreamed of it becoming so notoriously "popular". Thus the quality of the code is very poor. Also I'm very sorry for the unjustified uproar, which this benchmark is causing.

To 1. : Those high number are caused by errors if the kernel launch fails. A failed kernel launch has a runtime of about 0 ms. Since the bandwidth is calculated by size/time the estimated bandwidth becomes very high. Error handling would avoid this problem, but I was too lazy to program it. Thus your high bandwidth suggests, that there is an error. Maybe the windows watch dog? Maybe wrong project settings?

To 2. : Code is ok.

To 3. : A larger problem size would indeed increase the accuracy. But increasing the problem size would increase the runtime. A higher runtime might cause the kernel to fail because of the watch dog. I was again too lazy to avoid this.

To 4. : Simple hack, to measure the read bandwidth. The compiler cannot determine whether "Temp" will be written back or not. Thus he cannot omit the load instruction.

"Therefore I will also not be running his executable just in case he is trying to infect systems."
Damn! I wanted your pc for my botnet, too!

xorbe · Mar 26, 2015

If the 4GB card has segmentation (3.5+0.5), then the 2GB should have segmentation also (1.75 + 0.25). The bandwidth numbers look good to my eyes, are you sure, OP? ie 86GB/s on 128 bit interface is about 172GB/s on 256 bit interface or about 980 levels on that same test. Also, 660 was unbalanced, so it's nothing new for that product line if so (except for hiding the fact, if the accusation is true.)

UnknownzD · Mar 26, 2015

I have done the test more than dozens time already, and it did occur on my card.

Although the program was not designed for such a purpose, however an intense swapping between memories would be enough to see the read and write speed of each memory segment. In addition, the test result is totally reproducible over a dozen of times, so please explain to me when the result is highly reproduib like this. Moreover, if the reading across first 3.5 GB is over 200GB/s for more than 10 tests, while the reading for the last 500MB is always smaller than 20GB/s during these tests, I believe that the chances that a statistical error has occurred is actually quite small.

'Was not meant to be' doesn't mean it doesn't have such a ability to test the memory speed.

xorbe said:
If the 4GB card has segmentation (3.5+0.5), then the 2GB should have segmentation also (1.75 + 0.25). The bandwidth numbers look good to my eyes, are you sure, OP? ie 86GB/s on 128 bit interface is about 172GB/s on 256 bit interface or about 980 levels on that same test. Also, 660 was unbalanced, so it's nothing new for that product line if so (except for hiding the fact, if the accusation is true.)

I guess you have never looked at the reading for last 500MB section, right? Also in what sense makes you think that the 2GB version will have a same PCB design as 4GB version?

MxPhenom 216 said:
This is all I needed out of that link. Thanks for it!

Next time please paste everything from the same post to here, and avoid hiding something instead.

@ Anybody else:
As a CUDA programmer I'm also kind of interested, why the bandwidth drops so much. I assume that it is caused by the undocumented swapping behaviour of the virtual global memory space. But what are the precise explanations for those drops? My investigations suggest that the global memory swapping is one way associative. They also suggest that a page fault doesn't cause the GPU to upload the page from CPU DRAM to GPU DRAM. Thus the GPU copies the data over the PCI-E for each access again and again, just like pinned memory. Is this correct?

Your answer still doesn't answer something like this. And because of the ASUS and Nvidia fan boy here trying to ignore the fact that the reading for the last 500MB was surprising low, I have decided to file an official complain to the government department instead, such that a through inspection will be performed over the card, with its label GDDR5.

We will see the result once the officials put the card into test in testing and certification centre.

MxPhenom 216 · Mar 26, 2015

UnknownzD said:
I guess you have never looked at the reading for last 500MB section, right? Also in what sense makes you think that the 2GB version will have a same PCB design as 4GB version?

Its not really PCB design thats the issue. Its internals of the GPU and memory controller.

UnknownzD said:
Next time please paste everything from the same post to here, and avoid hiding something instead.

I was quoting......

UnknownzD · Mar 26, 2015

MxPhenom 216 said:
Its not really PCB design thats the issue. Its internals of the GPU and memory controller.

Data line is still a part of the PCB design, I guess.

xorbe · Mar 26, 2015

MxPhenom is correct -- the card maker can't change the internals of the GPU. Doubling the vram isn't going to change how it works. That's why I said 3.5+0.5 would mean 1.75+0.25 also.

Use a 2nd card for display, then run the benchmark again with the 960 in headless mode. That's why I was looking at the peak bandwidth numbers instead. They are more like half of a 980 versus half of a 970, which suggests no b/w is missing. When the card is actively used for desktop, you don't really have full access to the entire vram. The 970 numbers were produced in headless mode, and the last 512MB was quite even, whereas your numbers are all over the place (and they dip before 3.5GB). I'm not saying you're definitely wrong, but the test needs to be run correctly.

(Shades of Litwicki.)

commando55555 · Mar 26, 2015

Why should asus do anything there is nothing wrong with your card. You brought a low end card you got a low end card whats the problem which will never need more than 3.5 any way bitching for the sake off bitching.

UnknownzD · Mar 26, 2015

xorbe said:
MxPhenom is correct -- the card maker can't change the internals of the GPU. Doubling the vram isn't going to change how it works. That's why I said 3.5+0.5 would mean 1.75+0.25 also.

Use a 2nd card for display, then run the benchmark again with the 960 in headless mode. That's why I was looking at the peak bandwidth numbers instead. They are more like half of a 980 versus half of a 970, which suggests no b/w is missing. When the card is actively used for desktop, you don't really have full access to the entire vram. The 970 numbers were produced in headless mode, and the last 512MB was quite even, whereas your numbers are all over the place (and they dip before 3.5GB). I'm not saying you're definitely wrong, but the test needs to be run correctly.

(Shades of Litwicki.)

Okay, thank you for your reply, and I will further retest the process with another graphic card attached.

commando55555 said:
Why should asus do anything there is nothing wrong with your card. You brought a low end card you got a low end card whats the problem which will never need more than 3.5 any way bitching for the sake off bitching.

Because I have enough money to purchase GTX 970, and the only reason to avoid it is to avoid the memory problem reported by other users. In addition, according to the consumer law here, if you label something on top of the package, stating that the product has certain features to achieve, you have to make sure the product can actually deliver the claimed benefit.

System Name	Ryzen Reflection
Processor	AMD Ryzen 9 5900x
Motherboard	Gigabyte X570S Aorus Master
Cooling	2x EK PE360 \| TechN AM4 AMD Block Black \| EK Quantum Vector Trinity GPU Nickel + Plexi
Memory	Teamgroup T-Force Xtreem 2x16GB B-Die 3600 @ 14-14-14-28-42-288-2T 1.45v
Video Card(s)	Zotac AMP HoloBlack RTX 3080Ti 12G \| 950mV 1950Mhz
Storage	WD SN850 500GB (OS) \| Samsung 980 Pro 1TB (Games_1) \| Samsung 970 Evo 1TB (Games_2)
Display(s)	Asus XG27AQM 240Hz G-Sync Fast-IPS \| Gigabyte M27Q-P 165Hz 1440P IPS \| Asus 24" IPS (portrait mode)
Case	Lian Li PC-011D XL \| Custom cables by Cablemodz
Audio Device(s)	FiiO K7 \| Sennheiser HD650 + Beyerdynamic FOX Mic
Power Supply	Seasonic Prime Ultra Platinum 850
Mouse	Razer Viper v2 Pro
Keyboard	Razer Huntsman Tournament Edition
Software	Windows 11 Pro 64-Bit

System Name	Ryzen Reflection
Processor	AMD Ryzen 9 5900x
Motherboard	Gigabyte X570S Aorus Master
Cooling	2x EK PE360 \| TechN AM4 AMD Block Black \| EK Quantum Vector Trinity GPU Nickel + Plexi
Memory	Teamgroup T-Force Xtreem 2x16GB B-Die 3600 @ 14-14-14-28-42-288-2T 1.45v
Video Card(s)	Zotac AMP HoloBlack RTX 3080Ti 12G \| 950mV 1950Mhz
Storage	WD SN850 500GB (OS) \| Samsung 980 Pro 1TB (Games_1) \| Samsung 970 Evo 1TB (Games_2)
Display(s)	Asus XG27AQM 240Hz G-Sync Fast-IPS \| Gigabyte M27Q-P 165Hz 1440P IPS \| Asus 24" IPS (portrait mode)
Case	Lian Li PC-011D XL \| Custom cables by Cablemodz
Audio Device(s)	FiiO K7 \| Sennheiser HD650 + Beyerdynamic FOX Mic
Power Supply	Seasonic Prime Ultra Platinum 850
Mouse	Razer Viper v2 Pro
Keyboard	Razer Huntsman Tournament Edition
Software	Windows 11 Pro 64-Bit

System Name	Ryzen Reflection
Processor	AMD Ryzen 9 5900x
Motherboard	Gigabyte X570S Aorus Master
Cooling	2x EK PE360 \| TechN AM4 AMD Block Black \| EK Quantum Vector Trinity GPU Nickel + Plexi
Memory	Teamgroup T-Force Xtreem 2x16GB B-Die 3600 @ 14-14-14-28-42-288-2T 1.45v
Video Card(s)	Zotac AMP HoloBlack RTX 3080Ti 12G \| 950mV 1950Mhz
Storage	WD SN850 500GB (OS) \| Samsung 980 Pro 1TB (Games_1) \| Samsung 970 Evo 1TB (Games_2)
Display(s)	Asus XG27AQM 240Hz G-Sync Fast-IPS \| Gigabyte M27Q-P 165Hz 1440P IPS \| Asus 24" IPS (portrait mode)
Case	Lian Li PC-011D XL \| Custom cables by Cablemodz
Audio Device(s)	FiiO K7 \| Sennheiser HD650 + Beyerdynamic FOX Mic
Power Supply	Seasonic Prime Ultra Platinum 850
Mouse	Razer Viper v2 Pro
Keyboard	Razer Huntsman Tournament Edition
Software	Windows 11 Pro 64-Bit

System Name	Ryzen Reflection
Processor	AMD Ryzen 9 5900x
Motherboard	Gigabyte X570S Aorus Master
Cooling	2x EK PE360 \| TechN AM4 AMD Block Black \| EK Quantum Vector Trinity GPU Nickel + Plexi
Memory	Teamgroup T-Force Xtreem 2x16GB B-Die 3600 @ 14-14-14-28-42-288-2T 1.45v
Video Card(s)	Zotac AMP HoloBlack RTX 3080Ti 12G \| 950mV 1950Mhz
Storage	WD SN850 500GB (OS) \| Samsung 980 Pro 1TB (Games_1) \| Samsung 970 Evo 1TB (Games_2)
Display(s)	Asus XG27AQM 240Hz G-Sync Fast-IPS \| Gigabyte M27Q-P 165Hz 1440P IPS \| Asus 24" IPS (portrait mode)
Case	Lian Li PC-011D XL \| Custom cables by Cablemodz
Audio Device(s)	FiiO K7 \| Sennheiser HD650 + Beyerdynamic FOX Mic
Power Supply	Seasonic Prime Ultra Platinum 850
Mouse	Razer Viper v2 Pro
Keyboard	Razer Huntsman Tournament Edition
Software	Windows 11 Pro 64-Bit

System Name	Bayou Phantom
Processor	Core i7-8700k 4.4Ghz @ 1.18v
Motherboard	ASRock Z390 Phantom Gaming 6
Cooling	All air: 2x140mm Fractal exhaust; 3x 140mm Cougar Intake; Enermax T40F Black CPU cooler
Memory	2x 16GB Mushkin Redline DDR-4 3200
Video Card(s)	EVGA RTX 2080 Ti Xc
Storage	1x 500 MX500 SSD; 2x 6TB WD Black; 1x 4TB WD Black; 1x400GB VelRptr; 1x 4TB WD Blue storage (eSATA)
Display(s)	HP 27q 27" IPS @ 2560 x 1440
Case	Fractal Design Define R4 Black w/Titanium front -windowed
Audio Device(s)	Soundblaster Z
Power Supply	Seasonic X-850
Mouse	Coolermaster Sentinel III (large palm grip!)
Keyboard	Logitech G610 Orion mechanical (Cherry Brown switches)
Software	Windows 10 Pro 64-bit (Start10 & Fences 3.0 installed)

System Name	Home Brewed
Processor	i9-7900X and i7-8700K
Motherboard	ASUS ROG Rampage VI Extreme & ASUS Prime Z-370 A
Cooling	Corsair 280mm AIO & Thermaltake Water 3.0
Memory	64GB DDR4-3000 GSKill RipJaws-V & 32GB DDR4-3466 GEIL Potenza
Video Card(s)	2X-GTX-1080 SLI & 2 GTX-1070Ti 8GB G1 Gaming in SLI
Storage	Both have 2TB HDDs for storage, 480GB SSDs for OS, and 240GB SSDs for Steam Games
Display(s)	ACER 28" B286HK 4K & Samsung 32" 1080P
Case	NZXT Source 540 & Rosewill Rise Chassis
Audio Device(s)	onboard
Power Supply	Corsair RM1000 & Corsair RM850
Mouse	Generic
Keyboard	Razer Blackwidow Tournament & Corsair K90
Software	Win-10 Professional
Benchmark Scores	yes

Processor	AMD 3900X \ AMD 7700X
Motherboard	ASRock AM4 X570 Pro 4 \ ASUS X670Xe TUF
Cooling	D15
Memory	Patriot 2x16GB PVS432G320C6K \ G.Skill Flare X5 F5-6000J3238F 2x16GB
Video Card(s)	eVga GTX1060 SSC \ XFX RX 6950XT RX-695XATBD9
Storage	Sammy 860, MX500, Sabrent Rocket 4 Sammy Evo 980 \ 1xSabrent Rocket 4+, Sammy 2x990 Pro
Display(s)	Samsung 1080P \ LG 43UN700
Case	Fractal Design Pop Air 2x140mm fans from Torrent \ Fractal Design Torrent 2 SilverStone FHP141x2
Audio Device(s)	Yamaha RX-V677 \ Yamaha CX-830+Yamaha MX-630 Infinity RS4000\Paradigm P Studio 20, Blue Yeti
Power Supply	Seasonic Prime TX-750 \ Corsair RM1000X Shift
Mouse	Steelseries Sensei wireless \ Steelseries Sensei wireless
Keyboard	Logitech K120 \ Wooting Two HE
Benchmark Scores	Meh benchmarks.

Processor	8700k
Motherboard	Asus Rog Maximus X Hero
Cooling	Ekwb Performance P360 kit +PE 240 Rad
Memory	Corsair LPX 16 GB 3000 DDR4
Video Card(s)	EVGA 1080 ti FTW3 (EKWB waterblocked)
Storage	Samsung 970 Pro 512 Samsung 960 Evo 500 NVMe M-2, WD 1TB black
Display(s)	Dell AW3418DW 34.1" (3440X1440) 120Hz G-Sync X2
Case	Fractal Design R6 TG (Gunmetal)
Power Supply	Seasonic Prime 750 Titanium
Mouse	Logitech G600
Keyboard	Logitech G910
Software	Windows 10 Pro

System Name	msdos
Processor	8086
Motherboard	mainboard
Cooling	passive
Memory	640KB + 384KB extended
Video Card(s)	EGA
Storage	5.25"
Display(s)	80x25
Case	plastic
Audio Device(s)	modchip
Power Supply	45 watts
Mouse	serial
Keyboard	yes
Software	disk commander
Benchmark Scores	still running

ASUS Strix Nvidia GTX960 4GB - Same 4GB problem as GTX 970

New Member

Attachments

ASIC Engineer

New Member

ASIC Engineer

New Member

ASIC Engineer

New Member

ASIC Engineer

PC Gaming Enthusiast

New Member

TPU addict

ASIC Engineer

New Member

ASIC Engineer

New Member

ASIC Engineer

New Member

ASIC Engineer

New Member

New Member