Tom's Hardware Superposition Benchmark Thread

[RememberThe5th]

It is always power limit, as i said im hitting perf cap on 1973MHz overclock.

Do the shunt mod, increase power limit by 300%

Patience, i have only soldering iron

 

[EquineHero]

Patience, i have only soldering iron

You can get a lot of solder from old power supplies. Just solder a wire across the shunt.

 

[RememberThe5th]

Patience, i have only soldering iron

You can get a lot of solder from old power supplies. Just solder a wire across the shunt.

Well for soldering iron i meant also that i have solder and other stuff, just, just wait a bit, for the summer and ill do and post tutorial , but as i looked today for resistor, oh god, its not normal one, its surface mount, that makes things a bit complicated as the PCB is painted, if you scratch it a bit you can actually see the copper and PCB ( i know because i tried to resolder fan header to this R7 250), well patience is key i don't wanna lose GPU now .
Here goes the R7 250

1.2GHz on both mem and core. 1.15V

Oh and who says you cannot run almost 4GB Vram on 1GB muahhahahaha

 

[EquineHero]

Patience, i have only soldering iron

You can get a lot of solder from old power supplies. Just solder a wire across the shunt.

Well for soldering iron i meant also that i have solder and other stuff, just, just wait a bit, for the summer and ill do and post tutorial , but as i looked today for resistor, oh god, its not normal one, its surface mount, that makes things a bit complicated as the PCB is painted, if you scratch it a bit you can actually see the copper and PCB ( i know because i tried to resolder fan header to this R7 250), well patience is key i don't wanna lose GPU now .
Here goes the R7 250

1.2GHz on both mem and core. 1.15V

Oh and who says you cannot run almost 4GB Vram on 1GB muahhahahaha

It's pointless to add those scores to the chart. They're way to low for its intended purpose, heck even my GTX 460 SE does way better

 

[RememberThe5th]

The point is i can, nah i dont care if you are willing to add or not i just added for fun :3

 

[fagetti]

6 years old gtx 670 overclocked

 

[fagetti]

This 670 is actually connected to PCIe 1.1 x4 port so it would probably perform alot better on 2.0 x8 or x16. There is just no space in my rig for it cause of noctua cooler taking all freakin space. Has someone actually tested can 6 years old card even perform much better on abit faster PCIe port?

 

[RememberThe5th]

Yes it actually should perform better.

 

[fagetti]

I tried the 1080p extreme and it is a <<watch the language please>> lagfest even with +770 on memory. Anyone tested how stock or oc gtx 670 performed in superposition medium? Want to compare difference from PCIe 2.0 x16 to PCIe 1.1 x4.

 

[COLGeek]

Re-stickied. Stay on track and keep it clean or we'll close for good. Thanks.

 

[EquineHero]

I tried the 1080p extreme and it is a <<watch the language please>> lagfest even with +770 on memory. Anyone tested how stock or oc gtx 670 performed in superposition medium? Want to compare difference from PCIe 2.0 x16 to PCIe 1.1 x4.

Seriously, connect it to a PCIe 3.0 interface, it will perform hella better. The 1050Ti averages 12FPS on my end, the 670 should do better. It has WAY more CUDA cores but you're clocking it the same. For example, the average Passmark score for the GTX 580 is 5000. I got 3500 at PCIe 3.0 x4 - and over 6000 at PCIe 3.0 x16. The amount of lanes and the revision severely affect the performance of the card.

 

[Deleted member 217926]

I tried the 1080p extreme and it is a <<watch the language please>> lagfest even with +770 on memory. Anyone tested how stock or oc gtx 670 performed in superposition medium? Want to compare difference from PCIe 2.0 x16 to PCIe 1.1 x4.

Seriously, connect it to a PCIe 3.0 interface, it will perform hella better. The 1050Ti averages 12FPS on my end, the 670 should do better. It has WAY more CUDA cores but you're clocking it the same. For example, the average Passmark score for the GTX 580 is 5000. I got 3500 at PCIe 3.0 x4 - and over 6000 at PCIe 3.0 x16. The amount of lanes and the revision severely affect the performance of the card.

It makes no difference. I had a 2600K on a P67A UD4 before moving to Skylake and I just moved my 980 Ti over. Literally no difference between PCI-E 2.0 and 3.0. A 1080 might see 2-3FPS and a 1080 Ti a few more but you need new, flagship level cards to saturate the PCI-E 2.0 interface. The x4 slot will of course be slower.

 

[Rogue Leader]

I tried the 1080p extreme and it is a <<watch the language please>> lagfest even with +770 on memory. Anyone tested how stock or oc gtx 670 performed in superposition medium? Want to compare difference from PCIe 2.0 x16 to PCIe 1.1 x4.

Seriously, connect it to a PCIe 3.0 interface, it will perform hella better. The 1050Ti averages 12FPS on my end, the 670 should do better. It has WAY more CUDA cores but you're clocking it the same. For example, the average Passmark score for the GTX 580 is 5000. I got 3500 at PCIe 3.0 x4 - and over 6000 at PCIe 3.0 x16. The amount of lanes and the revision severely affect the performance of the card.

Thats not true. An x4 slot will drag it down, but there is no way in hell a GPU will get a massive performance difference between PCIe 2.0 x16 and 3.0 x16. Its been tested many times, and usually equates to a few %.

 

[EquineHero]

I tried the 1080p extreme and it is a <<watch the language please>> lagfest even with +770 on memory. Anyone tested how stock or oc gtx 670 performed in superposition medium? Want to compare difference from PCIe 2.0 x16 to PCIe 1.1 x4.

Seriously, connect it to a PCIe 3.0 interface, it will perform hella better. The 1050Ti averages 12FPS on my end, the 670 should do better. It has WAY more CUDA cores but you're clocking it the same. For example, the average Passmark score for the GTX 580 is 5000. I got 3500 at PCIe 3.0 x4 - and over 6000 at PCIe 3.0 x16. The amount of lanes and the revision severely affect the performance of the card.

It makes no difference. I had a 2600K on a P67A UD4 before moving to Skylake and I just moved my 980 Ti over. Literally no difference between PCI-E 2.0 and 3.0. A 1080 might see 2-3FPS and a 1080 Ti a few more but you need new, flagship level cards to saturate the PCI-E 2.0 interface. The x4 slot will of course be slower.

I tried the 1080p extreme and it is a <<watch the language please>> lagfest even with +770 on memory. Anyone tested how stock or oc gtx 670 performed in superposition medium? Want to compare difference from PCIe 2.0 x16 to PCIe 1.1 x4.

Seriously, connect it to a PCIe 3.0 interface, it will perform hella better. The 1050Ti averages 12FPS on my end, the 670 should do better. It has WAY more CUDA cores but you're clocking it the same. For example, the average Passmark score for the GTX 580 is 5000. I got 3500 at PCIe 3.0 x4 - and over 6000 at PCIe 3.0 x16. The amount of lanes and the revision severely affect the performance of the card.

Thats not true. An x4 slot will drag it down, but there is no way in hell a GPU will get a massive performance difference between PCIe 2.0 x16 and 3.0 x16. Its been tested many times, and usually equates to a few %.

My GTX 1070 was being severely bottlenecked by my old Xeon's PCIe 2.0 restriction. Sure I had a good framerate in everything but ever since I've upgraded to my 4790K I've seen a large improvement in all of my games.

The Witcher 3 went from 20-45fps at 5760x1200 to 45-60FPS at 5760x1200. I know my CPU wasn't a bottleneck cause it was only around 80% on all cores - I mean have you seen my Passmark bench? I've set a new model record three times with that CPU.

Certain applications don't mind PCI revisions but those like games that require as much bandwidth as possible can show a large difference between revisions.

Another example with Passmark - My 1050Ti did about 7500+ on an i5 3470, which is PCIe 3.0. However, on my Xeon, it only did about 5000 due to the PCIe limit.

 

[EquineHero]

<<watch the language please>>

https://www.youtube.com/watch?v=hpigjnKl7nI

But anyway. The GTX 670 should be able to outperform the GTX 1050Ti because of the memory and core speeds. GTX 1050Ti is normally clocked at 1.3GHz and 3500Mhz on core and memory - putting your 670 at that it should kill a 1050Ti with both hands tied behind its back blindfolded cause it has almost double the amount of CUIDA cores.

 

[Deleted member 217926]

The Witcher 3 and Fallout 4 were two of the games I tested with. I just moved my Steam SSD ( 500GB 840 Evo ) over to the new build. No difference in performance.

 

[Rogue Leader]

My GTX 1070 was being severely bottlenecked by my old Xeon's PCIe 2.0 restriction. Sure I had a good framerate in everything but ever since I've upgraded to my 4790K I've seen a large improvement in all of my games.

The Witcher 3 went from 20-45fps at 5760x1200 to 45-60FPS at 5760x1200. I know my CPU wasn't a bottleneck cause it was only around 80% on all cores - I mean have you seen my Passmark bench? I've set a new model record three times with that CPU.

Certain applications don't mind PCI revisions but those like games that require as much bandwidth as possible can show a large difference between revisions.

Another example with Passmark - My 1050Ti did about 7500+ on an i5 3470, which is PCIe 3.0. However, on my Xeon, it only did about 5000 due to the PCIe limit.

Something else was holding you back:

http://www.hardwaresecrets.com/pci-express-3-0-vs-2-0-gaming-performance-gain/4/

Passmark is a synthetic benchmark and has 0 to do with gaming loads. If it did a lot more processors would be way more competitive than they are. Even Pcie 2.0 x16 has way more bandwidth than most any game would need.

 

[jankerson]

My GTX 1070 was being severely bottlenecked by my old Xeon's PCIe 2.0 restriction. Sure I had a good framerate in everything but ever since I've upgraded to my 4790K I've seen a large improvement in all of my games.

The Witcher 3 went from 20-45fps at 5760x1200 to 45-60FPS at 5760x1200. I know my CPU wasn't a bottleneck cause it was only around 80% on all cores - I mean have you seen my Passmark bench? I've set a new model record three times with that CPU.

Certain applications don't mind PCI revisions but those like games that require as much bandwidth as possible can show a large difference between revisions.

Another example with Passmark - My 1050Ti did about 7500+ on an i5 3470, which is PCIe 3.0. However, on my Xeon, it only did about 5000 due to the PCIe limit.

Something else was holding you back:

http://www.hardwaresecrets.com/pci-express-3-0-vs-2-0-gaming-performance-gain/4/

Passmark is a synthetic benchmark and has 0 to do with gaming loads. If it did a lot more processors would be way more competitive than they are. Even Pcie 2.0 x16 has way more bandwidth than most any game would need.

Exactly.

I ran 3DMark Fire Strike on my old i7 870 and I7 7700K with a RX 480 and the GPU score was the same in both systems. (Within 100 points.)

13,807 - I7 870
13,905 - I7 7700K

 

[fagetti]

Tell me what benchmark should i run to compare my gtx 670 in Pcie 1.1 x4 compared to other revisions?

 

[Rogue Leader]

Tell me what benchmark should i run to compare my gtx 670 in Pcie 1.1 x4 compared to other revisions?

I didn't get that from your post, you are running at 1.1 x4? The x4 part will kill it especially.

 

[fagetti]

Yep, Asus p7p55d lga 1156 motherboard has only 2 pci express ports. One with 2.0 x16, it wont fit there because of large cpufan. I could turn noctua fan in another position but then i cant fit in my ram which has big heatspreaders. Another port what i use currently has pcie 1.1 x4.
Cant benchmark before next weekend but i remember getting "performing way above expectations above 97th percentile" so im getting to top3 percent of nearly 20k of gtx 670 benchmarks with pcie 1.1 x4, not bad still.

Equinehero: 1050ti is already beaten. It has average bench on userbenchmark at 39%
Im getting over 45% even with pcie 1.1 x4 so it can only get better from here

 

[logainofhades]

My GTX 1070 was being severely bottlenecked by my old Xeon's PCIe 2.0 restriction. Sure I had a good framerate in everything but ever since I've upgraded to my 4790K I've seen a large improvement in all of my games.

I would say it is probably more of the CPU being faster, and higher clocked, than the PCI-E lane speed. A gen 2 Xeon is already about 10-20% slower at same clock speed, vs a 4th gen i7. Factor that on top of a 400mhz clock speed advantage, and that puts the 4790k even further ahead. The games you were playing probably couldn't fully take advantage of all 8 threads of your Xeon either, hence the 80%.

 

[SoNic67]

RX580, 1415MHz (but in test drops back towards stock). I guess the test result page is not updated, since it recognizes my RX580 as RX480. During the test though, the GPU is recognized correctly.
Result: 2537

 

[EquineHero]

RX580, 1415MHz (but in test drops back towards stock). I guess the test result page is not updated, since it recognizes my RX580 as RX480. During the test though, the GPU is recognized correctly.
Result: 2537

Sorry for being a bit slow to update the table, I've been busy.

Table will be updated shortly.

 

[EquineHero]

Sorry for the stupid image, this is an exported frame from a movie I took and I didn't save the score. The test was run on my primary rig to eliminate any possible bottle-necking.

Rule # 11 is being applied to this score for its inclusion in its table.

Table willbe updated shortly.