Massive FPS gains in rf2 using PCI-e 3.0 x16 with higher end cards!

I already referred to the render test in my post.

I have 3.0 equipment. i7-4790 CPU and mainboard with 3.0 support and GPU that also supports 3.0. I wrote the post because despite that, it may not be operating at 3.0.

 

They sure AREN'T fun! Already updated to latest BIOS. No effect.

 

So, you have the card at 99% GPU load during the demanding GPU-Z benchmark, and while the benchmark is running it still doesn't report PCI-E 3.0 (and at 16x) after 3 or 4 seconds?

 

If gpu-z doesn't read 3.0 in load (be it with the render tool in gpu-z or with a game running), then the card is not running in 3.0 mode for some reason.

Have you tried to force it permanently into 3.0 mode through the bios Marc (if possible on yours)?

Also, fyi, gpu-z reports my card is at 1.1 (at the end) when idle and goes up to 3.0 when in load. So you idle of 1.1 is correct but your load of no higher than 2.0 in load is showing a problem.

update: Are you using anything in any of the other PCI-e lanes? I'm wondering that if you are, could it be forcing the gpu x16 lane to go down a gen as a result.

update 2: Are you 100% sure that you saw it say 3.0 at some point? If it did and then went down to 2.0, i can't imagine my possible suggestion about it being forced into 2.0 by another pci-e device (if you have one) could be the cause then.

 

OK, mystery solved. I called Asus tech support. They told me absolutely that GPU-Z should be reporting 3.0x16. Two options to fix: 1) move the card to the other slot (I have SLI); 2) Clear CMOS and reset all MB settings back to default.

1) did nothing, Card still reported 1.1x16.

2) Worked like a charm. I realize now in hindsight that there was another issue that suggested a problem--to do with the cooling fan controls. Anyway, reports 3.0x16 and render test works as expected.

Now for the drum roll again...

No noticeable performance change from when it was reporting 1.1.

So either it was operating at 3, but reporting 1.1, or...as I said above, the move to the new mainboard did zip-all for graphics performance. I got even less than what I would expect by moving up to a 4.2 GHz CPU, so there sure wasn't any GPU increase.

 

Ah-ha....your using SLI. Problem solved.

Your cpu only supports a maximum of 16 pci-e lanes. When using SLI with such a cpu, that means each card has effectively only x8 lanes available. So each card can only run in PCI-e 3.0 x8 mode which is the same as running at PCI-e 2.0 x16.

You need the CPU's that provide at least 32 pci-lanes to have each card running at full PCI-e 3.0 x16 mode i'm afraid.

Apologies for the bad news.

What was your system spec before the upgrade Marc?

 

doh solved

 

I'm not saying he's a dud. Read post 246 and 247, lol. A. His BIOS wasn't even at default settings, and B. He's running SLI on a CPU chipset which only supports 16 lanes so he's not even going to be able to achieve 3.0 @ 16x.

P.S. You love your HDD/SSD wipes don't ya? I think I've wiped my entire system to fix a problem once, maybe twice, in the last 5 or 6 years. 99% of the time it's completely unnecessary and a waste of time.

 

Marc, what was your previous cpu? The only way i can think of that you're seeing identical performance as before is if you were using a cpu with 32 (or more) pci-e 2.0 lanes (e.g. i7-920) which would then make sense that you are getting identical performance with your new system in rf2. If not, then you should still be seeing some improvement in performance because a regular cpu with 16 pci-e 2.0 lanes (e.g. i5-2500k) with SLI cards would have to be ran at 1.1 x16/x16, which performs worse in rf2 than 2.0 x16/x16.

Also, at least in the rf2 live benchmark test, cpu speed does not affect fps performance at all unless our modern cpu's are running slower than something like 2.0Ghz. Anything above that and performance is all gpu dependent (again at least in the rf2 live benchmark test, i can't speak for other scenarios and i don't think they have been definitely confirmed any different yet).

 

So why does my GPU-Z now report 3.0x16 capability and 3.0x16 achieved when running the render test? As far as I know, you only get reduced x8 if you are using SLI, which I am not.

 

BIOS had been updated by the way. And no critical settings were changed, so no, it wasn't obvious that there would be a BIOS issue.

I am not running SLI. I have a board capable of SLI. I get 3.0x16 now as I should and as I reiterated above that Asus confirmed I should with my equipment.

 

That's odd, you shouldn't be able to. Are you sure it's saying "pci-e 3.0 x16 @ 3.0 x16" under load? Also if it says that, does it always say that or does it change to "pic-e 3.0 x16 @ 1.1 x16" when idle.

You should not be able to run at 3.0 x16 on both cards in SLI and your reported no change in performance would agree with this.

Did you mention to asus you are using SLI when you spoke with them?

Also, what was your CPU and graphic card(s) before?

 

Ohhh, so you're not running SLI. I misunderstood you, Marc, when you said "I have SLI" in post #246.

Dr.R, Marc isn't running SLI, it's just the way he worded it. I think Marc meant that he has an SLI capable motherboard.

 

Yes, sorry for confusion.

 

No need to apologize

 

It is not about wiping the system.

It is about restoring your perfect image.

It is about restoring a SSD to factory benchmarks, especially older ones do not do good job of trim ( like mine )

If people here have only done 2 wipes in 6 years well mate all I can say is there is going to be some pretty sick SSDs around !@

You also do not do extensive baseline tests without " renewing systems"


Most every SSD HDD tester worth their salt uses PARTED MAGIC to zero writes drives before they run another batch of tests on SSD's..............

So if you use parted magic to clean drives and restore your images , well you also need trueimage, makes sense. ?






There is reasons people have a aversion to parted magic and trueimage.

They are scared of things they never did nor understand. lol

 

Like ........

Now ya now going to tell me using restore points and incremental windows backups is better then trueimage/parted magic.

lool

Funning mate lopl


btw , serious note you never answered "1600 on a H97 chipset " ram question yet, I think you missed the "plot" there too.

 

When testing hardware, software, OS , really anything.

Once a certain set of benches are complete, , what do they ( pro tresters ) do to restore the OS or hard drives they used ?

Do they format them all and do a fresh OS on every drive ...........no ! lol

They use PartMagic and TrueImage because it is faster and more consistent then a human manually doing 10 OSes.

They do it because 1. it factory resets the SSD and 2. It is way faster then a manual installs ......ergo it saves time )


So in short it saves time when you have bugs or restoring or testing even like yesterday me trying W8.2 for a day again.


Seriously am I going to sit there and totally redo my perfect image W7 sim OS manually taking a few hours ?? or am I going to run trueimage and have it done in 3 minutes. ?


Like be "fair dinkum" mate, the benefit in that situation is smacking you in the face mate .

 

Oh right, ok. So what could be causing your problem?

As well intentioned as your advice is durge, I have to admit I'm with Spinelli on this one that i really doubt reformatting his os is the solution matey.

 

This is what gets my BiPolar going............. hhahahaha looool Weeeeeeeee !


When people do not acknowledge or see the benefit of something in a given situation no matter what hardware or software we talk about, or do not personally use or test .........................then in the same breath use a " throw away like this..........."

W7 was released 5 years ago, so through testing extensively maybe 4 OSes .............................and he does 2 clean drives............. that is it. lol



Sounds reasonable for a "bencher" lool