Massman Posted October 15, 2010 Share Posted October 15, 2010 Hey guys, I had a discussion with Dinos22 regarding the difference in efficiency between R3E and UD9 with multi-VGA and Vantage, so I spend some time testing the boards I have here for scaling. Since this is pretty pointless, I thought I might as well justify the time spent on it by making a nice graph and posting that in a forum for everyone to watch. With a few pats on the back, I might feel a little more happy about doing this ... Conclusion is pretty simple: find a decent CPU and a decent pair of GPUs because the difference in efficiency is insignificant and un-important. Only for 3-Way SLI the implementation of the UD7 and R3E seems to be a tad better, but nothing you can't overcome by increasing the GPU clocks with 10MHz ... (Sorry, no Evga boards in the house so couldn't include SR2 or Classified). CPU: 980X @ 150x27 = 4050MHz UNC: 3.6GHz MEM: 1800 CL9-11-9 GTX480: 800/1050 OS: W7 64-bit Drivers: 259's Quote Link to comment Share on other sites More sharing options...
Massman Posted October 15, 2010 Author Share Posted October 15, 2010 Quote Link to comment Share on other sites More sharing options...
chuchnit Posted October 15, 2010 Share Posted October 15, 2010 Nice graph Massman. I'm kinda surprised at the Xpander results! Quote Link to comment Share on other sites More sharing options...
Massman Posted October 15, 2010 Author Share Posted October 15, 2010 Surprised in what way? Quote Link to comment Share on other sites More sharing options...
chuchnit Posted October 15, 2010 Share Posted October 15, 2010 Surprised in what way? I'm surprised that it is as efficient as it is. I figured the UD9 would take the cake in between the two. I guess I was just expecting more latency with all thats going on with the Xpander. Quote Link to comment Share on other sites More sharing options...
Maxi Posted October 15, 2010 Share Posted October 15, 2010 Nice info Massman, get your hands on some EVGA parts if you can. Quote Link to comment Share on other sites More sharing options...
Massman Posted October 15, 2010 Author Share Posted October 15, 2010 I'm surprised that it is as efficient as it is. I figured the UD9 would take the cake in between the two. I guess I was just expecting more latency with all thats going on with the Xpander. The asus-boys have gotten the hang of getting decent 3D boards. The R2E was pretty much the same: CPC a bit more efficient than the other boards. Quote Link to comment Share on other sites More sharing options...
TaPaKaH Posted October 15, 2010 Share Posted October 15, 2010 9-11-9 for ultimate performance! Quote Link to comment Share on other sites More sharing options...
Massman Posted October 15, 2010 Author Share Posted October 15, 2010 9-11-9 for ultimate performance! No. It wasn't single channel, so it was sub-par performance ... Quote Link to comment Share on other sites More sharing options...
Christian Ney Posted October 15, 2010 Share Posted October 15, 2010 Nice Massman even if it's useless for me , it would be helpful for others Quote Link to comment Share on other sites More sharing options...
chuchnit Posted October 15, 2010 Share Posted October 15, 2010 The asus-boys have gotten the hang of getting decent 3D boards. The R2E was pretty much the same: CPC a bit more efficient than the other boards. Ya Asus makes kick ass boards for 3d. I still thought the Xpander would hamper things more. I guess what I should really be surprised about is the lackluster performance of the $700 UD9 Quote Link to comment Share on other sites More sharing options...
dinos22 Posted October 15, 2010 Share Posted October 15, 2010 (edited) told you that Pt1t's results were bullshit thanks for testing though....good info Edited October 15, 2010 by dinos22 Quote Link to comment Share on other sites More sharing options...
chuchnit Posted October 16, 2010 Share Posted October 16, 2010 Massman, how much bribing would it take to get you to compare UD9 vs RE3 + Xpander at max clocks on ln2? I'd like to see if the efficiency delta is the same at max clocks. I know Duex says he can pull off 2 RTL tighter on UD9 than Classified 4-way. I'm just wondering if the results are the same when both boards are optimized to the max. In other words, all things not being equal. Quote Link to comment Share on other sites More sharing options...
Massman Posted October 16, 2010 Author Share Posted October 16, 2010 Hmm. I still have a few drips of LN2 in the tank and already did a few runs on the R3E. [hwbot=1038703]submission[/hwbot] Only thing that's different is the CPU. I no longer have this B1 ES and the one I have is a bit weaker than the one used for those runs. Also, I'm not entirely sure those runs were optimized properly (in terms of driver settings) ... I think any comparison will be open for criticism. In fact, I think the only situation where there would be no criticism on methodology would be when turns out both boards are the same, which is an assumption we can already make now . I can hook up the Single Stage for a few 4.8G-5G runs. Would that be interesting for you? Quote Link to comment Share on other sites More sharing options...
chuchnit Posted October 17, 2010 Share Posted October 17, 2010 Hmm. I still have a few drips of LN2 in the tank and already did a few runs on the R3E. [hwbot=1038703]submission[/hwbot] Only thing that's different is the CPU. I no longer have this B1 ES and the one I have is a bit weaker than the one used for those runs. Also, I'm not entirely sure those runs were optimized properly (in terms of driver settings) ... I think any comparison will be open for criticism. In fact, I think the only situation where there would be no criticism on methodology would be when turns out both boards are the same, which is an assumption we can already make now . I can hook up the Single Stage for a few 4.8G-5G runs. Would that be interesting for you? That's fine actually because I think we all know cpu clocks aren't going to vary that much on single stage or ln2 from board to board. I am more curious if say you tune each setup to its max on each board. I mean subs, RTL's, pcie, uncore; etc. I know you are busy so its not a big deal if you can't find the time. I am just wondering if we see one clearly stand out without all things being equal? Quote Link to comment Share on other sites More sharing options...
dinos22 Posted October 17, 2010 Share Posted October 17, 2010 That's fine actually because I think we all know cpu clocks aren't going to vary that much on single stage or ln2 from board to board. you sure about that? Quote Link to comment Share on other sites More sharing options...
Massman Posted October 17, 2010 Author Share Posted October 17, 2010 I'll run the test if Dino is willing to accept a possible outcome where the R3E is able to produce higher clocks ;-) Quote Link to comment Share on other sites More sharing options...
Massman Posted October 17, 2010 Author Share Posted October 17, 2010 Now running the tests on the UD9. Having some weird slowdown issues now at clocks that are supposed to be completely stable. Quote Link to comment Share on other sites More sharing options...
Monstru Posted October 17, 2010 Share Posted October 17, 2010 WHAT different 3D efficiency issues? I've tested mult-GPu on UD9 since May, and I also tested RE3 back then and I saw no difference... http://lab501.ro/placi-video/nvidia-gtx480-performanta-in-configuratii-multi-card Quote Link to comment Share on other sites More sharing options...
Massman Posted October 17, 2010 Author Share Posted October 17, 2010 We were discussing the results of Pt1T's testing of UD9 and R3E where UD9 was behind the R3E for quite a bit. In my current testing, I see the same thing happening: Note: with R3E the results vary more than with UD9. Sometimes 151FPS, sometimes 153FPS in GT1, for UD9 it's pretty stable results. I'm now testing the Xpower for some reference 3-Way numbers. In 3-way, the UD9 is slightly ahead of the R3E (although it was different at 4GHz). Quote Link to comment Share on other sites More sharing options...
SF3D Posted October 17, 2010 Share Posted October 17, 2010 I would like to know, that have someone really bought UD9? In Finland 0 boards have been sold so far. Price is 517€, so I can see the reason there. If that board is loosing some very valuable performance in 4-way, I can not see any reason to even test it. Buying one would be even more ridiculous. Quote Link to comment Share on other sites More sharing options...
Massman Posted October 17, 2010 Author Share Posted October 17, 2010 Quite a lot of mainboards have been sold, as far as I know. At least, relative big amount ... it still is high-end board. But don't jump to conclusions yet. I will re-test UD9 with other settings so maybe I can find why it's slower. Quote Link to comment Share on other sites More sharing options...
Massman Posted October 17, 2010 Author Share Posted October 17, 2010 Also, as far as I can see the ASUS 4-Way SLI (Xpander) is still not fully available in retail Quote Link to comment Share on other sites More sharing options...
SF3D Posted October 17, 2010 Share Posted October 17, 2010 Did I say anything about ASUS hehe The fact is, that RIIIE cost 370€ in here, which is huge amount of money. 4-way setups are mainly for top end benching, so basic players are not the target group. So, adding Xpander type extension board is good idea if we keep money in our minds. The main idea of RIIIE is to keep cost as low as possible and still offer best possible performance. Nf200 chips on board does not fit in to that idea. I am not taking any sides, but about this particular subject, it is my honest opinion that RIIIE + Xpander is better choice. The ones who does not need or want 4 way ability, will get really good board and they don't have to pay extra (Nvidias fault, not GB's in UD9 case) Quote Link to comment Share on other sites More sharing options...
Massman Posted October 17, 2010 Author Share Posted October 17, 2010 I agree on pricing ... and performance-wise although I'm stunned the difference is this large. At 4G, the difference is almost non-existing, at 4.8G suddenly huge gap. Must be some other problem, I think. Oh, and ... ehr. I don't think R3E+Xpander is in fact a choice . Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.