Jump to content
HWBOT Community Forums

WHATS UP WITH 3d11!!??


Recommended Posts

  • Replies 105
  • Created
  • Last Reply

Top Posters In This Topic

Prolly not all realize, that its much harder to bench 11 high that aq or any other old benchmark.

Need some hard facts?

 

Game 4, Reference GTX 580 , 1255MHz GPU, 2500MHz memory clocks, -100c.

Power draw from PSU (ONLY videocard, +12V rail)

840W, brown colored measurement field, at right.

 

Aquamark, same stuff.

470W, same colored measurement at right.

 

3D03 Nature, same settigns

590W.

 

 

Perfect info on this post. Thanks!

If there is any chance, could you test hwbot heaven xpreset :)

Just for the interest..

 

I have not that much to say on this 3dmark11 thing,

but for sure it is a great benchmark.

Mainly because it is short and heavy,

but not because it would add something to the existing benchmarks.

Link to comment
Share on other sites

I'd like to see 11X preset earning full points, 11P is too similar to current Vantage ranking. So what if you have to buy a license, at least you will have a benchmark which stands apart from the rest. Heaven would be the closest but is still quite different.

 

Stop hating on 05. :(

 

Could see this coming from a mile away :D

Link to comment
Share on other sites

Your not getting it man :D

It's the same as it for vantage. If you run the SAME gpu clocks in quad on an sr2 that you do on a single gulftown that is clocked higher in mhz, the gpu score is ALOT higher. 11 is similar, but not as magnified. Thats all your seeing. Like I said run 11 in quad with a single cpu system and go form 260-270 drivers then you will see the difference in your comparison.

I'm not gonna argue that the physics test is higher on sr2, but the combined test is NOT its lower. The net effect of 1-2k higher physics score is barely anything in total score in 11 im tellin ya, its still very competitive between the two platforms. Its high mhz single cpu vs low mhz dual cpu. Much more balanced than vantage and look how close those scores are. Don't use my scores as compare:D Look at smoke/Andre

 

after some testing it seems pretty clear you are scoring both physics and combined a LOT higher which is what is driving the 4way score most

 

here's me at 5Ghz and ive looked at andres runs which i think he said were at 6.3Ghz and he was only 10-15% higher on last two tests.

 

3DMARK11 P is pretty much like Vantage

 

im running 5Ghz with 1150/2300/1200 4way here

 

clipboard01kmc.jpg

Link to comment
Share on other sites

You don't have to prove that the sr2 is better at 3dmark11 than the board you are using, thats for sure :D but its much closer than you think bro. At first I was getting similar scores with sr2 when I first started testing it, but I got it to score higher in the long run, particulary in the tests you mention. I'm not gonna tell u how to run the benchmark dude.

Link to comment
Share on other sites

Hey Dino don't take this the wrong way man, but something is wrong in your results and they are very misleading. I'm unsure if the gigabyte 3d11 results are really that poor in combined and physics or your scores are messed up, but let me try and make some things CLEAR about the SR2 with regards to 3D11 and vantage cuz I think maybe most people just don't know whatsup due to not having run it themselves.

 

Ok so for testing here I ran 3D11 Performance preset using EVGA Classified SR-2/ dual 5680's and quad SLI on reference GTX580's. All test are at stock clocks (unless otherwise noted) and untweaked without any "magic" other than moving the slider in NV CP all the way to the left. Thats it!

 

First of all, this comparison of game tests on Sr2 vs your single CPU system is BAD. You are comparing DIFFERENT drivers here below:

clipboard01kmc.jpg

 

Please use this screenshot as a more accurate compare which are both on 275 versions drivers (your 275 version is still a bit faster though). This shows 1210/2530 stomping 1150/2400 as it should in 3D11 game tests:

3d11_25470.jpg

 

 

after some testing it seems pretty clear you are scoring both physics and combined a LOT higher which is what is driving the 4way score most

 

Yes my physics and combined is A LOT higher, but it's not because I have 2 cpus man :D

I ran some 3D11 Performance preset here at stock clocks using both TWO cpus and ONE cpu to show this awesome advantage the Sr2 has in 3D11 :D. There are two runs for each instance to show typical margin of error that you can expect in 3dmark11. Its pretty tight and consistent from run to run and another one of its advantages as a solid benchmark:

 

TWO CPUS:

P15890-cpu2.jpg

 

P15923-cpu2.jpg

 

and ONE CPU..WOW what an advantage! 170 points between low margin 2 cpu run vs high margin one cpu run!:

P15724-cpu1.jpg

 

P15666-cpu1.jpg

 

Now to put things in perspective a little, lets see what the single cpu system gets with roughly 400mhz higher overclock on cpu along with higher memory and uncore (typical system OC differences between single and dual cpu system in 3D11. This is the equivalent of running about 6ghz on gulftown vs the 5.6 that we ran in the EVGA SR-2 3d11 World Record. Again, I'm showing two runs for each instance to show the margin of error between runs:

P16359-cpu1-oc.jpg

 

P16432-cpu1-oc.jpg

 

Hmmm so an overclocked single cpu system is actually faster then a dual cpu system SETTING for SETTING as shown here in 3D11 REAL WORLD typical OC vs typical OC? :D

 

3DMARK11 P is pretty much like Vantage

 

WRONG!!

 

Now lets see the differences between SINGLE vs DUAL cpu on SR-2 in 3Dmark Vantage Performance in comparison to 3D11 at stock, again two runs for each instance to have a little margin of error difference, which can be as much as 6-800 points, here suprisingly it wasn't so great and it scored consistently for me.

SINGLE CPU:

43k-single-cpu.jpg

 

43k-single-cpu2.jpg

 

DUAL CPU:

P48252-cpu2.jpg

 

P48535-cpu2.jpg

 

Wow so 5k point difference between single cpu vs dual cpu at stock. So this is that big SR2 adVANTAGE everyone likes to talk about?? :D Well, the REALITY is that an SR2 cant clock as high as single cpu system can, especially a single cpu system with an ultra binned non realistic VERY non-average cpu. Current EVGA 3DMARK Vantage World record of almost 80k is with two cpus clocked at 5940mhz while the current highest single cpu result its crushing is clocked at a whopping 700mhz+ higher. Fair to use about 5-600mhz difference in overclock and then compare vantage scores??? That is like comparing my WR result at 5.9 with 6.5 gulftown.

So lets take a look a this supposed huge advantage that this 7 pot and very difficult to overclock rig has over its single cpu competitor when overclocked on system side:

Advantage huh? :D

 

SINGLE CPU with 600mhz OC on system side:

P49076-cpu1oc.jpg

 

P49105-cpu1-oc.jpg

Edited by k|ngp|n
Link to comment
Share on other sites

Nice compo :)

Hmm, so 10 times difference between 3Dmark Vantage Performance and 3Dmark11 Performance in percentage way.

 

Easy math:

 

I took average of those low-high pairs in each benchmark with 1P and 2P and what we get?

 

3Dmark11 P : +1.35% increase in total result score in 2P system vs 1P. (15906.5 avg pts vs 15695 avg pts)

 

3Dmark Vantage P: +10.37% increase in total results 2P vs 1P (43846 avg pts vs 48393.5 avg pts)

 

Almost 10 times difference.

 

And I bet difference between benchmarks even increase with higher delta's between CPU frequency diff 1P vs 2P.

We all know that X5000 series CPU's locked multi, what limits SR-2 only to Vantage actually in cases of 3w/4w SLI, from what we see.

Link to comment
Share on other sites

Andre is at 6.3GHz 1250/2500/1250 (266.58) and I'm at 5.1GHz 1150/2300/1200 (275.35)

 

you can see combined and physics tests are not much different. looks like the new driver is much stronger in GTs, not sure how combined will be affected

 

you boys are totally missing the point of my discussion here, it originally started with "if you want to be fair then use 3DMARK11 instead of Heaven etc" to which i said hang on EVGA will have an advantage which you dont deny right. The advantage seems to be in the tweaks though from what you are trying to say.

 

I personally am commenting from end user perspective and as a hwbot member wanting healthy competition. i am not trying to put a foot in the door for gigabyte or anyone here at all just simply discussing relevance of a benchmark if only one platform which no one benches on seems to have the edge. If you reckon that is not the case and that its actually some tweaks you found all good then.

 

Can i ask you, how consistent is the scoring with SR2 at very high bclock? Do you see some higher than usual frame jumps like it does on Vantage or wprime?

 

clipboard01v.jpg

Link to comment
Share on other sites

Vantage P is too similar to 11P, yes there is CPU scaling in Vantage but the main point is graphics performance based on the test resolution, it is much more magnified and future proof. If you really think you'll get off more running P vs X go for it...but why not go for X over P if you have a choice?

Link to comment
Share on other sites

Andre is at 6.3GHz 1250/2500/1250 (266.58) and I'm at 5.1GHz 1150/2300/1200 (275.35). You can see combined and physics tests are not much different. looks like the new driver is much stronger in GTs, not sure how combined will be affected

 

I'm pretty sure both the Gigabyte and ASUS scores are not 100% efficient and have problems. I agree about the driver having the most impact on GT's. My physics and combined are actually lower in the current WR at 25470, than in previous score that you linked on orb showing that they have lesser impact on final score compared with game tests. 3D11 is all about game tests. Quad is not much different than single card in what impacts scaling, this is another example of why it is not the same as vantage. You can see in the single vs dual cpu comparisons on SR2 for 11 that the game tests go down when scond cpu is added, even at same exact gpu and cpu clock. This is typical sr2 behavior and can be seen in vantage as well. This offsets the slight increase in physics and combined scores (dual cpu) and is a big reason there is not much difference in final score single vs dual cpu for 11. Futuremark did a good job to make a healthy competition for all you end users Dino :D

 

You boys are totally missing the point of my discussion here, it originally started with "if you want to be fair then use 3DMARK11 instead of Heaven etc" to which i said hang on EVGA will have an advantage which you dont deny right. The advantage seems to be in the tweaks though from what you are trying to say.

 

So how does that give EVGA an advantage then if its more about how you run it and not what you run it on? Doesn't that alone make it more suitable for something like HWBOT? The SR2 is better at 3D11 then single cpu board, I won't argue that. Most of the point of this is that when typical overclocks are used on system side, its pretty fair fight. A lot of how much better it is depends on end user though. These screenshots I've shown highlight how close the systems can be when all things are the same from one cpu to two.

Why instead of and not just use both benchmarks? Isn't more variety better man? Makes things more interesting for me personally. Heaven is good, but seriously its just one long game test with no dynamic load changing or shifting from different areas of the system, they are not really similar benchmarks other than they both use DX11. How many dx10 benches does hwbot currently host for WR and global points?

 

Can i ask you, how consistent is the scoring with SR2 at very high bclock? Do you see some higher than usual frame jumps like it does on Vantage or wprime?

 

I have no idea what your talking about and have never seen something consistent with this kind of behaviour. When pushing high bclock the frame rates and wprime times DROP big time, because the data signals between IOH and cpus are getting messed up and data is being thrown back constantly. This slows things down BIG TIME, not causes "higher than usual jumps". Who told u that :D

Anyone who has benched the setup seriously knows its not the case. Getting the baseclock high on this bastard and having the frames/times correct (meaning up where they are supposed to be) is very tricky and requires a lot of painstaking work.

 

Look you asked me "what it was I was trying to say", so instead of keep saying the same things over and over and have you post comparisons that don't jive with reality and are misleading, I'd rather post up some real world /no hype compares so the rest of these guys can judge for themselves what impact having a second cpu alone makes in 3d11 and how it compares to the same impact in 3Dmark vantage. Like you said, not everyone benches the Sr2 setup so you guys really just don't know.

I'm sure TiN will thank you for locking up half of my day yesterday preparing this as I was not able to harass and torture him like i usually do every day :D

Edited by k|ngp|n
Link to comment
Share on other sites

Vantage P is too similar to 11P, yes there is CPU scaling in Vantage but the main point is graphics performance based on the test resolution, it is much more magnified and future proof. If you really think you'll get off more running P vs X go for it...but why not go for X over P if you have a choice?

 

Its not that similar Marky, but I'm totally down with Xtreme.

Just add the benchmark for global/wr points is all I'm asking preset doesn't matter. :D

 

@El gappo :D

Link to comment
Share on other sites

duck-san has experienced higher than normal CPU test scores in Vantage as he reported on XS a while ago, i remember even Gatman chiming in saying they are not records if its based on what seem to be buggy scores

 

youngpro saw board giving buggy boosts in wprime at times during his benching particularly when pushing bclock

 

you are obviously assuming what the scores "should" be on X58 platform also right and if what you say is true you should be able to smash your scores with the Classified easily too right, why not bench Classified? teamau score may be buggy but i unless there are a lot of scores around to compare thats hard to say, we've only done one run at 1150 in pretesting and never got a chance to bench properly as one of the cards died so it is possible for sure.

 

the reason why i voted YAY in the other thread is that Heaven is just not up to par with 11 in terms of being a better benchmark yet. '11 has been changed to most of the demands we had, it is now becoming less buggy and works on SLI systems better. Heaven is a great idea and i like the fact we have an alternative but it needs to be shorter, have subtest and some ability to tweak.....so in other words replacing heaven with X '11 is what i thought would be a good idea. The more tests we add the harder it is with guys that want to seriously bench but have LN2 access issues of price problems. This is not an issue in USA, Taiwan, some places in Australia and a few other countries but it is an issue elsewhere. I think it needs to be taken into consideration when more benchmarks are added personally.

Edited by dinos22
Link to comment
Share on other sites

duck-san has experienced higher than normal CPU test scores in Vantage as he reported on XS a while ago, i remember even Gatman chiming in saying they are not records if its based on what seem to be buggy scores

 

youngpro saw board giving buggy boosts in wprime at times during his benching particularly when pushing bclock

 

you are obviously assuming what the scores "should" be on X58 platform also right and if what you say is true you should be able to smash your scores with the Classified easily too right, why not bench Classified? teamau score may be buggy but i unless there are a lot of scores around to compare thats hard to say, we've only done one run at 1150 in pretesting and never got a chance to bench properly as one of the cards died so it is possible for sure.

 

the reason why i voted YAY in the other thread is that Heaven is just not up to par with 11 in terms of being a better benchmark yet. '11 has been changed to most of the demands we had, it is now becoming less buggy and works on SLI systems better. Heaven is a great idea and i like the fact we have an alternative but it needs to be shorter, have subtest and some ability to tweak.....so in other words replacing heaven with X '11 is what i thought would be a good idea. The more tests we add the harder it is with guys that want to seriously bench but have LN2 access issues of price problems. This is not an issue in USA, Taiwan, some places in Australia and a few other countries but it is an issue elsewhere. I think it needs to be taken into consideration when more benchmarks are added personally.

 

Yo Dino! :D Can you tell my girl hasn't been around for awhile mate? haha

Ducks situation was about his gpus being pushed to hard. I talked about that with him actually and another one of our team members was able to reproduce it, its nothing to do with bclock and I actually believe this vantage bug can be replicated on any platform w555 or x58. Its pretty obvious bug tho regardless. While I have a HUGE amount of respect for Duck-san, I agreed with gatman in his conclusion.

As for James, I dunno we talked about prime on sr2 when we met and he never mentioned that to me. Would have loved to discuss that further. I've benched the hell out of wrpime on w555 and never seen this. Maybe this is why I couldn't beat his 1024m time eh? :D Actually I think I really figured that out and may give it another go soon.

 

As for the x58 asus and gigga scores, I'm not guessing or assuming, but going by the ORB details. Can you do me favor and post a full screen of the run you linked to with details bro?

Were getting off base here, its good discussion but this is about hwbot adding 11 for gobal/WR points for PRO league. I just wanted to show some easy comparos in case anyone that had doubts about 11 being an SR2 only benchmark, that is really not the case. Yes the SR2 is better clock for clock / setting for setting. It can be by as much as 2-600 points depending on end user However when comparing 2 fully OC'd rigs pushed within their reasonable capabilities X58 vs W555 is a pretty fair fight for more than a few reasons. Some being the greater OC capabilities of an x58 rig and others being the pure difficulty of running the w555 rig.

Edited by k|ngp|n
Link to comment
Share on other sites

  • Crew

actually Vinbo i used to think the same thing about 05 as you, it certainly has a large "whoever tests the most cpus wins" type element to it, but it also has alot of room to move in tweaking,

 

each to their own, but for me, for a benchmark to be worth running, i need to get some benefit if i invest my time into tweaking it, with both heaven and 11 there isnt much there, sure heaven has a few little quirks we can use to get a few points.. but 11 is fairly tight like vantage.. both have a few things, but nothing in the realms of 3d01, 03, 05, etc..

 

that being said, im not a big heaven fan and would like to see 11 being benched, but why not pick 11 x, that way we have a benchmark that at least for a while is going to be predominantly GPU power.. then we have a genuine benchmark that you need some 4-way handling skills and requires heavy GPU power.. lets face it 4-way vantage is totally driven by CPU power now, you can have 4 average GPU's and a great CPU and take WR..

Link to comment
Share on other sites

you are obviously assuming what the scores "should" be on X58 platform also right and if what you say is true you should be able to smash your scores with the Classified easily too right, why not bench Classified? teamau score may be buggy but i unless there are a lot of scores around to compare thats hard to say, we've only done one run at 1150 in pretesting and never got a chance to bench properly as one of the cards died so it is possible for sure.

 

Your the one thats assuming things Dino :D. I know the benchmark from running the hell out of it, and I KNOW that your combined and physics scores are not good for the clocks you said you ran at man (5ghz/1150/2300) This is not an assumption or reckless speculation. Same thoughts goes for Asus scores..

I'm confident I know the benchmark pretty good.

 

Do you understand where im comin with all of this? or still no?

TBH...the "only becuase of SR2" comments like this are getting tired:

I disagree. 3DMARK11 "P" is "I got EVGA SR2" board guy helper benchmark as you call it

 

Then ur posting compares in my thread with my details in a previous 3D11 WR result that I basically spent 2 months working on against what are with all due respect bro sub par scores that you said you didnt even spend much time on - with much lower clocks for a point of view about how the SR2 is superior? Like I said maybe its your HW, I really dont know. Now thats something I've speculated on.

 

The newer P25470 result I did live has so much work behind it dude, "crushing" it even on an SR2 again with reference cards is gonna be really tough. Lets see what the new classified 580 gets when it launches soon. Who knows :D

 

Look here man, I ran some 3D11 on the EVGA CLASSIFIED 762 today with some air cards. So you say the Asus score is at 6.3ghz and your at 5GHZ and the ASUS score is only 10-15% higher than you on combined/physics?

Well I got you by the same percentage on those two tests at your SAME 5GHZ CLOCK WITH THE CARDS ON AIR:

5ghz-3d11-762.jpg

 

and I got the ASUS score by nearly the same percentage with 400mhz less cpu clock:

5900mhz-3d11-762.jpg

 

22K at 5.9ghz with cards on air is pretty good for single cpu system. At 6.3ghz with cards on this system at 1220+/2550+ its has a very realistic shot at low 25k score maybe even WR.

On older drivers with Sr2 I remember I got a 21.4k result at around 5.4 with same gpu clocks for perspective.

Edited by k|ngp|n
Link to comment
Share on other sites

that's all i wanted to see ;)

 

Cool! :D

 

Also you guys can also clearly see the differences between vantage and 11 now as you see the overclock go higher and higher here. Each game test only went up one frame or less with 900mhz increase in cpu speed for a tiny difference of 50 points in total gpu score!! This is using exact same card clocks on both runs. So basically I got almost NO SCALING in quad in gpu portion of score form the increase in raw mhz. If this was vantageP, H, or even X, how much of an increase in gpu score do you think I'd have gotten from going from 5g's to 5.9g's?? and how much of an impact would that have on total score compared to what you see here in 11?? :D Anyone who runs it knows it would be a HUGE difference.

This is why 3d11 is so different than vantage. Futuremark did right on this one. Vantage and 11 don't scale the same at all. They are very different benchmarks. Scaling is the same for all presets on 11. The guy who has the fastest cpu can't rape the benchmark. 11 is not so much about the highest overclock, but the BEST total system overclock and you have to work very hard on system side as well as get 4 gpus screaming on Ln2 for a killer score for both single and dual cpu systems. It's a big league benchmark that should be for full points in the hwbot PRO LEAGUE. The SR-2 doesn't have the big advantage either, look at how close the physics and combined are with my cards on air on 5.9ghz X58 compared to the 25470 run using 5670mhz sr2 with Ln2 cards. Only 230mhz raw difference in OC but the X58 system can do a lot of things that the sr2 system CANNOT and we all know these are average to easy clocks for 990x. Also in the end on Ln2 the x58 system will have higher GPU tests which carry more weight than that of combined and physics. Its a fair fight

Edited by k|ngp|n
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...