cancel
Showing results for 
Search instead for 
Did you mean: 

Image Rendering went down 50K

Atovv
Level 7
Hello Guys

I have an issue with the Image Rendering in RealBench, my best score is 206K, but I was doing 195 - 198 K range.

I was trying to level up, so I installed my OS again. I was running Windows 8,1 Pro and now Ive installed same OS.

No hardware change or anything like it ... but know, Im doing 150K best score 160K.

I really dont know why my score went down.

Any tips or advice?
8,332 Views
17 REPLIES 17

Gobe
Level 8
A few of us are scratching our heads on the image editing scores. I've max'ed the field at over 270,000 yet Arne, with the same basic equipment and NVMe setup is next best at around 252,000. 18,000 points is a HUGE difference and we just can't come up with a good reason for it. I do have DRAM that runs XMP at 3200 and 14-14-14-34, but I've turned it down and didn't lose much of anything.

Not that I'm complaining...

I'd be taking a close look at Task Manager to see if anything has somehow snuck into your diagnostic mode and might be stealing performance in your new OS installation. Definitely make sure Intel Speed Step/C-States are disabled. I know that just the other day, my AV started showing up in diagnostic mode where it previously didn't (I think I installed an update to the AV engine). I have to kill it manually as it really drags the benches down.

Atovv
Level 7
Hey Thanks I fix my issue. I had on Energy Setting on Balance Mode (Default) when re-installing the new OS. Now it in Performance Mode, and my CPU is at min. 100%, so its running as it should in the benchmark.

I want to get thru the 200K barrier in Image Rendering, I just cant. Dont know why.

5960X @ 4.7 GHZ, 4 Titan X @ 1510 mhz and 16GB (4x4GB) Kingston Predator @ 2667 mhz and timing @ 13-14-14-28-2T

With this setup I should be getting better Image Rendering Scores, but nope.

Any clues o tips, just for pushing further the Image Rendering Scores.

Atovv wrote:
Hey Thanks I fix my issue. I had on Energy Setting on Balance Mode (Default) when re-installing the new OS. Now it in Performance Mode, and my CPU is at min. 100%, so its running as it should in the benchmark.

I want to get thru the 200K barrier in Image Rendering, I just cant. Dont know why.

5960X @ 4.7 GHZ, 4 Titan X @ 1510 mhz and 16GB (4x4GB) Kingston Predator @ 2667 mhz and timing @ 13-14-14-28-2T

With this setup I should be getting better Image Rendering Scores, but nope.

Any clues o tips, just for pushing further the Image Rendering Scores.

Not looking at it at the moment, but when you say "image rendering" you mean the first test? If so, I've found it to be highly correlated to (in order):
1. disk speed
2. cache speed
3. memory speed

disk speed when running it "cold". Cache speed when running it for a 2nd or 3rd time as now the cache is warm and disk performance becomes less critical (though still bounding - I assume on the write side). A 950Pro and a cache as far above 4.0GHz as you can manage seems to be the key.

Arne_Saknussemm
Level 40
This is intriguing...

This test does respond well to memory tuning...cache I'm not sure...

My RAM timings

14-15-15-15-1T / 4-6-260-8320-12-4-16-3-8-6-14 / 0-1-2-2-2-1-1-0-0-0-5-5-4-1-0-2

Maybe something down in second or third timings?

I ran 4.7 cache on that run still 18000 points off Gobe...I want those points! LOL

RAM quantity?

Gobe...mind posting up your timings and cache frequency? if not bench secret...

Arne Saknussemm wrote:
This is intriguing...

This test does respond well to memory tuning...cache I'm not sure...

My RAM timings

14-15-15-15-1T / 4-6-260-8320-12-4-16-3-8-6-14 / 0-1-2-2-2-1-1-0-0-0-5-5-4-1-0-2

Maybe something down in second or third timings?

I ran 4.7 cache on that run still 18000 points off Gobe...I want those points! LOL

RAM quantity?

Gobe...mind posting up your timings and cache frequency? if not bench secret...

Is gobe still running with the 960? I wonder if we've hit a point where _something_ actually renders faster in software for lack of mondo GPUs or if he's actually doing raid0 950 pro (he has 2 in his specs), so his disk has electrolytes per plants craving 😉

I hadn't seen much impact with 2800 cas 12 vs 3000 cas 14, but then, the CAS improvement should help with random access, so maybe that's why I didn't notice it. I traded speed for random latency and ended up in roughly the same place?

Perhaps more importantly (or not) is why I don't get any love from the realbench leader-board. I was #3 or so by top-score for a while and #5 more recently. I never saw my name in lights. 😉 (had to select all scores).

Gobe
Level 8
Hey guys.. working in Michigan till Wednesday night so I'm not able to take a look at things until I get back home.

(Arne, you want the 18,000? You can't handle the 18,000!) 😄

My timings are 14-14-14-34 and I've not messed with anything secondary or tertiary yet. I don't get any benefit from dropping down to 13-13-13-30. I've managed to flirt with 12-12-12, but haven't really made any conclusions. I've even loosened the timings and not really lost anything. Go figure.

I've just recently managed to break the 270K barrier, but that was up at 4.9 GHz with the ice bucket. I can get around 268K consistently and that's whether I have my cache at 4200 or 4700. Actually, with the CPU frequency up over 4.8 Ghz, I hit 268K pretty much regardless of various settings. If I have a bench secret, it's a damned good secret, because even I don't know what the secret is. I'll try dropping from 32GB quad to 16 GB dual and see if anything happens.

cekim, I'm running dual 960 now and it makes no difference to the image editing whether I run one or two. My motherboards don't boot 950 Pro NVMe, so I don't run in RAID0, but I did set up for boot off a regular SSD and set up the 950 Pro's in RAID0 in Windows and ran the benchmark on the RAID0 and I got absolutely zero benefit from it. If anyone has any ideas for ways I can test, I'll be happy to give it a try.

Gobe wrote:
Hey guys.. working in Michigan till Wednesday night so I'm not able to take a look at things until I get back home.

(Arne, you want the 18,000? You can't handle the 18,000!) 😄

My timings are 14-14-14-34 and I've not messed with anything secondary or tertiary yet. I don't get any benefit from dropping down to 13-13-13-30. I've managed to flirt with 12-12-12, but haven't really made any conclusions. I've even loosened the timings and not really lost anything. Go figure.

I've just recently managed to break the 270K barrier, but that was up at 4.9 GHz with the ice bucket. I can get around 268K consistently and that's whether I have my cache at 4200 or 4700. Actually, with the CPU frequency up over 4.8 Ghz, I hit 268K pretty much regardless of various settings. If I have a bench secret, it's a damned good secret, because even I don't know what the secret is. I'll try dropping from 32GB quad to 16 GB dual and see if anything happens.

cekim, I'm running dual 960 now and it makes no difference to the image editing whether I run one or two. My motherboards don't boot 950 Pro NVMe, so I don't run in RAID0, but I did set up for boot off a regular SSD and set up the 950 Pro's in RAID0 in Windows and ran the benchmark on the RAID0 and I got absolutely zero benefit from it. If anyone has any ideas for ways I can test, I'll be happy to give it a try.


Admit it, you put pine tar on your PCIe pins don't you? LOL

cekim wrote:
Admit it, you put pine tar on your PCIe pins don't you? LOL


No, but if I thought it might help, hell yea, I would!

Gobe wrote:
No, but if I thought it might help, hell yea, I would!

I saw that voltage you inflicted on your chip, I have no doubt. 😉