Printable View
How strong a processor does the fastest single-chip graphics card need to show its best? And what processor would be necessary for two cards like that in CrossFireX mode? You’ll find the answers in this review.
http://www.xbitlabs.com/images/video.../25_crysis.png
http://www.xbitlabs.com/articles/vid...u-scaling.html
The difference between Core2Duo 4.1 vs i7 4.1 is almost 50% every case, so Crysis does take advantage of SMT.
Such a "bad coded" game, huh?
Interesting to see also that adding a second card offers performance increases below 30%, something wrong with GPU scaling, but CPU scaling seems ok. Drivers, perhaps?
yeah x-fire is still fail on the 5 series :/Quote:
Originally Posted by Luka_Aveiro
Quote:
So, the numbers don’t always provide the full picture
So true
I wouldn't call it a failure, i would just say that in this particular games XFire doesn't scale to good. Just look at the Stalker Clear Sky and Far Cry 2 numbers and you will see a different picture.Quote:
Originally Posted by Jamesrt2004
At least in Stalker the numbers are up by around 90% for CrossFire and core number and speed make almost no difference.
It all depends on drivers, on the game developers ability to code for a platform and other stuff like that.
Comparing i7 (Quad Core) against Core2Duo (Dual core) is pretty much retarded.
They should compare i7 against Core2Quad.
I take this with a huge bag of salt. They always get results that no one ever getsQuote:
Originally Posted by Luka_Aveiro
Indeed,it's Xbitfail after all.Quote:
Originally Posted by skugpezz
no L3 cache speeds listed......
cas 4 with 4 slots filled doesn't happen on Amd, this isn't new either >_>... all 4 stick memory are relaxed slightly.
I agree it is somewhat skewed since they are comparing 2 cores to 4 cores and like you said C2Q would be a better comparison. My question is how does the C2D@5ghz+ (since many extreme OCers push the 45nm Wolfdale through the 5ghz wall) do against the i7 at stock? I'm really surprised the C2D@4.1ghz pulled the numbers it did.Quote:
Originally Posted by jam2k
Quote:
Originally Posted by Luka_Aveiro
I guess you are ignoring the fact that at 1920x1200 where the GPU is finally getting stressed the C2D@4.1ghz is only 2-3fps different than the i7 when the GPU is under maximum stress (Fully IQ@1920x1200). This is the main reason I went with a highly clocked dual core system, many games just don't use all 4 cores effectively and the higher clocked C2D pulls ahead for that reason. Now down the road yes the quads will pull away but right now the C2D@4ghz+ are pretty good CPUs even at the high end market. If you play at 1920x1200 it really depends on your GPU a lot more than your CPU.
Agree, pointing on difference at 1280×1024 is really stupid. Everyone knows i7 is better in clock-to-clok comparison with Phenom II. But normal user has some cheap Phenom II and HD 4870/HD 5770 (at maximum) for 1280×1024 resolution. But almost no difference (single GPU) between Phenom II and Core i7 (clock-to-clock!!) at 1920×1200. For that is almost every 2,8 GHz+ K10/Lynnfield or whatever enough. So what want authors of articles like this point to?? I still don't get it.
id like to see article try to answer questions like, whats the best resolution for me, or how much should i spend on my cpu, or is my duel core getting too old. interesting niche articles that show how one aspect is affected by all the others. too many reviews just do the fastest stuff, at cookie cutter settings, which in the end gives us 100 tests all identical, with different answers.
The 5770s have legendary crossfire scaling. You will find in a lot of tests that using two 5770s in crossfire can give a 100% and over performance increase (I get 23 FPS in heaven benchmark with one 5770, and 48 with both at the same clock speeds).
Core I7 with a pair of 5770s is IMO currently the most efficient setup.
benchmarks are obviously going to use SLI/CF better than games, but no one really cares about benchmark performance except benchers. The question is how much FPS do you get in an actual game between one 5770 and two.Quote:
Originally Posted by bhavv
Seems suspicious.
Coming from Core 2 Duo to i7 I gotta say there is a noticeable overall performance boost, though.
So they are right, at least to some extent.
not at 1920x1200, it's barely noticeable in most games. Now newer games that use Quads may pull the gap further, but it's only 2-3fps at 1920x1200 Full IQ.Quote:
Originally Posted by zalbard
"One more pitfall for the AMD platform in this test session was PCI Express. The Gigabyte GA-MA790GP-DS4H mainboard is based on the AMD 790GX chipset which supports two ATI Radeon cards in CrossFireX only as PCIe x16 + PCIe x8"
It's x8+x8 in CF and not x16+x8.
They should have used a high end 790FX AM3 board for this test and not a AM2+ mainstream mobo.:rolleyes:
i7 - X58 + DDR3 + x16+x16
C2D - X48 + DDR3 + x16+x16
PHII - 790GX + DDR2 + x8+x8
I agree. Why would anybody who has i7 @ 4.1 Ghz and 5879 crossfire wants to play Crysis at 1280x1024 res. with no AA/AAF?
They should have found out the highest "playable" settings (i.e. 1680x1050 2xaa, No AAF) on the lowest clock speeds (i.e. 2.7 ghz) and take that setting as a starting point in their comparison.
cant wait for monitors that support 240FPS (even though that is a very long ways off)
You can google a review for this information, I do not need to provide it :).Quote:
Originally Posted by LedHed
Crossfire gain in games on the 5770 is so good that every review recommends buying them.
already seen reviews and the performance gain in Games is not 100%, 75% at the most.Quote:
Originally Posted by bhavv
Quote:
Originally Posted by Oliverda
Yep, like the article was purposely skewed.. Phenom II, no DDR3, lol.
yes... quite strange, no core 2 quad, phenom II without using 790FX (16x+16x) and using ddr2... also, wouldn't 2x2gb be more appropriate than 4x1gb?...
at the end I don't think it would make a huge difference, but...
the only people who probably have 4x1gb of ddr2 are probably still using a PII 940, like myself, for the last year now. but even i have a 2x 16x PCIe slots, and my mobo is older than any PII out there.Quote:
Originally Posted by Spectrobozo
Here:Quote:
Originally Posted by LedHed
http://www.tweaktown.com/reviews/298...re/index9.html
Probably 95-110% (estimate) performance gain in Far Cry 2.
Im runnin a single 5870 with a 955 at 3.8ghz on a 26inch 1920*1200... according to this... if i ever want to go crossfire im ONLY going to get about a 5 fps increase? wtf :confused:
Actually, the difference in the OP charts is around 10-15 FPS, seeing gains from around 30 FPS to 40+.
The total FPS gain is not how to measure crossfire scaling, but rather the percentage difference, which will also vary from game to game.
Oh, wait, youre using a phenom :p.
i would have liked that review so much better if they had 5870s since its 2x 5770s. in many cases the 5850 by itself was close enough to be equal to the 5770 xfire. and the price of a 5850 is better than 5770 xfire too. its availability that sucks
except far cry 2 sucks - so its basically a benchmark
way to use an old 8x8 ultra low end board for amd but use full 16x16 boards for intel thats not stacked at all
it looks like it dosnt matter solong as u have a good MB
I don't think PCIe 2.0 8x vs 16x has much of an impact. techPowerUp did a review of the PCIe scaling on a 5870, but no word on minimum fpsQuote:
Originally Posted by zanzabar
Example of a 5870 on PCIe 2.0 8x vs 16x on crysis
http://www.techpowerup.com/reviews/A...Scaling/7.html
But it's probably the crossfire overhead eh?
it dosnt matter all that much but when u are compairing platforms that are all native 16x16 but then u take a gimped lowend board for the amd part there is something wrong. especially when there is 1-2 fps from the best to worst in most of the testsQuote:
Originally Posted by kite7
Xbitlabs is the last place I would expect AMD biased against results.
This is quite the surprise.
well, there are no core 2 quad results, but im guessing thats because no one buys core 2 quads anymore. why there are core 2 duo results, i dont know.
anyway, there are phenom results, thats basically gonna be pretty close to a core 2 quad at the same frequency.
if so then why does the C2D out pace the Phenom in almost every test?Quote:
Originally Posted by grimREEFER
the phenom does beat the core 2 duo in most testsQuote:
Originally Posted by LedHed
ATI HD 5870 Scaling Performance: X58 vs P55 Showdown
http://www.anandtech.com/mb/showdoc.aspx?i=3649&p=3
Barely any performance difference between 16X CF and 8X CF (X58 vs P55 respectively) , though I do agree xbitlabs should have went with a 790FX board anyways and more ram for the AMD setup.
It doesn't, but games have one renderthread where the "action" is. DX11 will scale betterQuote:
Originally Posted by LedHed
Anyone wh tries to ru(i)n an 5970 with a dual core is a retard (Hangover pronunciation)!
that entire review would beg to differ:Quote:
Originally Posted by initialised
he only later mentions that the dual core supposedly is more "jerky", but to me that is a message added by sponsors pushing new hardware. Because jerkiness effects min fps, we would be able to see it if it was happening. Like the Phenom shows, sometimes only pulling 6fps, while the rest of the CPUs are around 20fps.Quote:
Now, I’ll try to sum everything up. First of all, the single Radeon HD 5870 does not depend as much on the CPU as it is supposed to. According to my tests (the left part of the diagrams), the Radeon HD 5870 is quite satisfied with an overclocked dual-core CPU manufactured two years ago. And you can even leave the CPU at its default frequency at the high-quality settings and 1920x1200. The only exceptions are Left 4 Dead and Warhammer 40000: Dawn of War II. The former game is not a problem for modern top-end Radeons while the latter, on the contrary, calls for a quad-core CPU, preferably from Intel.
not where it matters, look at the 1920x1200 Full IQ tests (where owners of 5870's will be playing).Quote:
Originally Posted by grimREEFER
5870, 1920x1200 Full IQ settings(the very bottom graphs), 4.1 GHz, Top c2d, Bottom Ph II x4Quote:
Originally Posted by LedHed
10dddd0d1dddddd
01dddd1d0dddddd
C2D=2 wins 2 losses, rest draws
Ph II x4=2 wins 2 losses, rest draws.
5870 CFX, 1920x1200 Full IQ settings, 4.1GHz,
1100d00d1dd000d
0011d11d0dd111d
C2D=3 wins,7 losses, rest draws.
Ph II x4=7 wins, 3 losses, rest draws.
I don't know how you managed to come up with that. Even a skim through is enough to see Ph II wins. Well, it's quadcore vs dualcore after all.
Draws are when the fps difference is only 1 fps. 2 fps isn't a draw. The GPU seems to be the bottleneck more frequently on PhII than C2D though.
Why didn't they use C2Q anyway? Why didn't they test Phenom II about the jerkiness test either? Anandtech reported "smoothness" on Core i7 and Phenom II compared to C2Q and C2D ages ago.
*Takes cover behind flameshield*
Techpowerup reviews tested PCI-e bandwidth with a single card setup, not a crossfire setup. That's why we have CFX dongles, because the bandwidth required for inter-card communication on PCI-e is simply not enough. Legionhardware tested 4870x2 CFX before on 8x 8x setup and it shows performance drop.
I conclude the review is fail (overall. It's good on some matters).
I guess you missed the whole point of terrible min fps (as in game stutter) for the Phenoms compared to the C2D or the i7 at 1920x1200 full IQ.Quote:
Originally Posted by blindbox
Haha I just saw a C2D commercial; I don't think I've ever seen one except when retailer mentions it in specs (but that is normally sale paper).
doesnt matter , apples vs apples ftwQuote:
Originally Posted by kite7
Well there's only 7 games there that shows minimum fps data. 3 among them Phenom II wins, and 2 draws. That makes two C2D wins. There's no conclusion here, just not enough data.Quote:
Originally Posted by LedHed
You can't simply say min fps is the factor of smoothness. You'll need graphs for this. I can have an average of 45 fps but my game is a 60fps~30 fps saw graph or I can have an average of 45 fps constant with a single moment dip to 30 fps.
Something that some reviewers still need, user experience description.
min FPS can sometimes be a simple HDD limitation. graphs show smoothness, or a standard deviation chart might if anyone cared to make those anymore
The core2duo has over 50% power consumption than i7.Quote:
Originally Posted by Luka_Aveiro
@ load, idle, average...? :rolleyes:Quote:
Originally Posted by damtachoa
you think i care about power consumption when I happen to use a C0 e8400 1,46v to 4ghz under H2O?
guess what... I don't.
lol
How many times it has to be repeated that you can not get more then 100% in CF or SLI. There is no over 100% increase period , not possible .Quote:
Originally Posted by bhavv
It just boggles me that people could believe that 5+5 in some instances could equal 11.
if u have rounding or a nonlinear score u can get over 100%Quote:
Originally Posted by railer
or if u have a game with frame smoothing a target fps that is set to low for the single card then it set alot higher or properly for the multi card u can get over 100%
How? lets say hypothetically i have 2x5870 with each scoring 20K in 3dmark. Explain to me how its mathematically feasible to get 41K in CF. And lets drop artificial "rounding" when 3dmark for some reason is "set" to give me the most 19K with one card. I might be missing something but i just cant compute this in my head.
3d mark is a linear scale, but some things arnt. so say that u get a scoring method that was (cpu^2)*(gpu^2)/cpu*gpu. so if u had a cpu score of 2 then a gpu of 2 then u would have a score of 4 but if u had 2 for cpu and 4 for gpu then u would have a score of 16. and frame smoothing and target fps limmit the card so if its set up wrong it wont be the full potential of the gpu, UT3 games have this problemQuote:
Originally Posted by railer
I am getting ~90% in Dirt 2.Quote:
Originally Posted by LedHed
DX11 coding is a big part of it.
edit*
Probably why they "forgot" to test it.
Also did they use dx11 for Battleforge?
If not that means that out of 3 (that I know) dx11 games that are out atm,they tested 0 ?
X-fail.
Yeah, funny, huh?Quote:
Originally Posted by railer
Especially remembering the MSI and Hydra thread...