Printable View
Many thanks onethreehill :up:
Anyone know of any tests comparing a Q6600 OCed (~3.2-3.4GHz) vs i5-750? Currently sifting through the review list... I picked up a cheap LGA1156 board but still on the fence as to whether to use it.
What is impressing me is the power consumption, my GTX260+Q6600 B3 3.3GHz idles 180W from the socket, plus the heat is amazing probably more due to the B3 stepping than anything else...
Time to promote Madshrimps: http://www.madshrimps.be/?action=getarticle&articID=951
I :banana::banana::banana::banana:ed up by choosing vantage as 3D measurement tho. But, doesn't matter, you can find 3D numbers in the hundred other articles.
Thanks!Quote:
Originally Posted by massman
Review in my website :)
Intel Core i7 870 & Intel Core i5 750 LGA1156 : First review
http://www.vmodtech.com/main/article...-first-review/
MSI P55-GD80 : The Mini Review
http://www.vmodtech.com/main/article...e-mini-review/
Review : GIGABYTE P55-UD4P
http://www.vmodtech.com/main/article...byte-p55-ud4p/
ECS Black Series P55H-A : The Review
http://www.vmodtech.com/main/article...-a-the-review/
MSI P55-GD65 - Overclock Results
http://www.vmodtech.com/main/article...-gd65-oc-test/
Review : ASUS P7P55D Deluxe
http://www.vmodtech.com/main/article...p7p55d-deluxe/
Intel DP55KG EXTREME BOARD : Overclock Results
http://www.vmodtech.com/main/article...clock-results/
Thank you,
Quite as expected. Same in singelcard, with the X58 pulling ahead with multicards.Quote:
Originally Posted by onethreehill
Dude, the ONLY Core i5 cooled with LN2 review out there :clap:Quote:
Originally Posted by massman
And the only one with ONLY useless syntetic benchmarks in it.Quote:
Originally Posted by jmke
Unique in every way :cool:Quote:
Originally Posted by STaRGaZeR
Seriously, it doesn't make the conclusion wrong. If you want other tests, go and read other reviews; it's good to form a well-balanced opinion :up:
i5 750 is 159.99 at micro center...:O
Great price. I wish I could buy in micro center.Quote:
Originally Posted by breakfromyou
onethreehill
thank you!
Thanks. Good review.Quote:
Originally Posted by gosh
I see that for home use (image editors, archivers, games, video encoding) i5-750 is actually better then Ph-II 965.
I think that Intel has priced i5 750 against Phenom II 955 because in real use they are equal in WindowsQuote:
Originally Posted by kl0012
is there any review that shows overclocked performance scaling? of i750 and i860?
Intel Core i7-860 Lynnfield Processor Review - Best value in processors?
http://www.pcper.com/article.php?aid=781
Clock for clock: Core i5, Core i7, Core 2 Quad, Phenom II X4 Performance
http://www.legionhardware.com/document.php?id=857
i5 / P55 Lab Update - Now with more numbers
http://anandtech.com/mb/showdoc.aspx?i=3639
can't find the info did they turned turbo off!? If that haven't than clock-for-clock point of the article is missed.Quote:
Originally Posted by onethreehill
thanks onethreehillQuote:
Originally Posted by onethreehill
clock for clock at 3.6 ghz was what i was looking for and yeah surprise!!!!
wish they had turbo info there though...
Quote:
The most surprising match up was between the Core i5 750 and Core i7 860, which I mentioned was our original intention for this article. When clocked at the same frequency there is little difference between these two processors, and when pushing them as high as 3.60GHz there really is no difference. Of course the only difference between the two is the Hyper-Threading support of the Core i7 860, which really failed to help it in the majority of our tests.
Madshrimps has OC scaling (only in synthetic benchmarks) and turbo mode scaling. Don't know about the othersQuote:
Originally Posted by geo
lolQuote:
Originally Posted by onethreehill
On one side you have:-
http://www.legionhardware.com/Bench/..._Gaming_03.png
& in the other side you have:-
http://images.anandtech.com/graphs/p...3214/20030.png
:p::p::p::p:
I am waiting for the dam 5870 x2 to come out and then test both i7 750 and PhII 955/965 with a single 5870 x2!!!!
Is the i5 750 not more equal to the ph2 965?Quote:
Originally Posted by gosh
performance wise, but price wise its next to the 955.Quote:
Originally Posted by gallag
Read #212Quote:
Originally Posted by gallag
I don't get it? What am I not seeing? In that graph the i5 750 is about 1% off the ph2-965, In others it is slightly ahead. Am I missing something?Quote:
Originally Posted by gosh
Same difference with 955 vs 965 and you get 2 x PCIe x16 with phenom, cheaper motherboards even if the price has gone down with i5Quote:
Originally Posted by gallag
So there is only 1% between the ph2 965 and the ph2 955? And why start twittering on about pcie and mobo prices? Its like you are selling ph2 or something?Quote:
Originally Posted by gosh
Don't take what gosh says seriously. :rolleyes:Quote:
Originally Posted by gallag
Ah that review again... While the final graph does not say to much it is clear that i5-750 is better then Pii-965 for home use.;)Quote:
Originally Posted by gosh
http://img406.imageshack.us/img406/1202/ixbtreview.jpg
No! you picked graphs where i5-750 won. There are other graphs...Quote:
Originally Posted by kl0012
Also even if the test is much better compared to other tests it doesn't test multitasking or low fps in gaming. Both strong areas for phenom
http://ixbtlabs.com/articles3/cpu/in...7-lga1156.html
I tought I've made it clear - home use!Quote:
Originally Posted by gosh
Not much people will use Maya, Lightwave or SpecJVM at home :shrug: And those who will probably will go with i7.
well, check this page: http://www.anandtech.com/mb/showdoc.aspx?i=3639&p=3Quote:
Originally Posted by kl0012
What should I check? I dont use Lightwave at home (especially with Cinema 4D, Photoshop and Word/Exel running at background).Quote:
Originally Posted by gosh
There are more realistic multitasking scenarios:
http://www.bit-tech.net/hardware/cpu...d-cpu-review/6
http://www.hexus.net/content/item.ph...=19979&page=10
And about the games - I have no idea why they have a problem with GTX275.
Don't let yourself sucked into a circle of pointless argumentation with gosh, he never acknowledges anything.
He does that every time, and when it look like that he gets "pawned" he either is switching subject or plays dumb.
If you don't want to know how the hardware works then you shouldn't read my postsQuote:
Originally Posted by Hornet331
Gosh, most, if not everyone here knows how the hardware works (although there are differences in the technical level of understanding). What you fail to acknowledge is the truth because of your obvious bias.Quote:
Originally Posted by gosh
Since you love quoting Anand...
http://www.anandtech.com/cpuchipsets...oc.aspx?i=3634Quote:
Final Words
I'll start this conclusion with what AMD must do in response to Lynnfield. The Core i5 750 is a great processor at $196, in fact, it's the best quad-core CPU you can buy at that price today. In nearly every case it's faster than AMD's Phenom II X4 965 BE, despite the AMD processor costing almost another $50.
Here is another example,
Benchmarks:
http://images.anandtech.com/graphs/a...5733/20002.png
http://images.anandtech.com/graphs/a...5733/20003.png
http://images.anandtech.com/graphs/a...5733/20004.png
http://images.anandtech.com/graphs/a...5733/20007.png
http://images.anandtech.com/graphs/a...5733/20008.png
http://images.anandtech.com/graphs/a...5733/20009.png
http://images.anandtech.com/graphs/a...5733/20010.png
http://images.anandtech.com/graphs/a...5733/20011.png
http://images.anandtech.com/graphs/a...5733/20012.png
http://images.anandtech.com/graphs/a...5733/20013.png
http://images.anandtech.com/graphs/a...5733/20014.png
http://images.anandtech.com/graphs/a...5733/20015.png
http://images.anandtech.com/graphs/a...5733/20017.png
http://images.anandtech.com/graphs/a...5733/20018.png
http://images.anandtech.com/graphs/a...5733/20019.png
http://images.anandtech.com/graphs/a...5733/20020.png
http://images.anandtech.com/graphs/a...5733/20016.png
http://images.anandtech.com/graphs/a...5733/20021.png
http://images.anandtech.com/graphs/a...5733/20023.png
http://images.anandtech.com/graphs/a...5733/20024.png
http://images.anandtech.com/graphs/a...5733/20025.png
Power Consumption:
http://images.anandtech.com/graphs/a...5733/20026.png
http://images.anandtech.com/graphs/a...5733/20027.png
No bias here, I chose every result from Anand's Athlon II X4 review. You keep implying that i5 750 = PII 955 BE when that is CLEARLY not the case.
May I remind you again,
Not to mention, both the PII and i7s have similar overclocks on conventional cooling, with the slight advantage to i7 based on average overclocks seen on reviews and here on Xtremesystems. Most enthusiasts will overclock their system, which will further widen the gap between i7/i5 and PII.Quote:
Core i5 750 is a great processor at $196, in fact, it's the best quad-core CPU you can buy at that price today. In nearly every case it's faster than AMD's Phenom II X4 965 BE
But of course, you will ignore what I just posted, which is fine really since I don't care but I refuse to let your bias misinform people.
I now see what you mean, Does he even take himself seriously lol.Quote:
Originally Posted by Clairvoyant129
I will ignore your post if you don't know how to explain the results in graphs. That is the hard partQuote:
Originally Posted by Clairvoyant129
Or to make it easier. Explain the difference here:
http://www.anandtech.com/mb/showdoc.aspx?i=3639&p=2
http://images.anandtech.com/graphs/p...3214/19976.png
http://images.anandtech.com/graphs/a...5733/20013.png
Um what difference?Quote:
Originally Posted by gosh
What is there to epxlain?
Cinebench just favours Ci7 more then cinema 4D, tought as you might noticed C4D gains quite a bit form HT, thats why simmilar clocke 920 wins against 750...
Again, you try to generate an issue where no issue is....
Just started a living review of the Gigabyte P55-UD6 here :)
http://www.images.clunk.org.uk/revie...MG_9104-sm.jpg
The Intel Core i7 860 Review
http://anandtech.com/cpuchipsets/showdoc.aspx?i=3641
Onethreehill - Thanks for the link to the Anandtech review of the i7-860 Lynnfield!
It's well done with numerous tests, clearly understandable results and bang-on conclusions.
I'll be starting my own build with an i7-860 from MicroCenter at the introductory bargain price as soon as my preferred Mushkin DRAM becomes available. :up:
Here's a test done by tomshardware to judge Lynnfield's onoard PCI-E performance ------> http://www.tomshardware.com/reviews/...ield,2379.html
IMO the test is awful and they used a 790GX when there is a 790FX with two 16x slots and why not OC the processors?. @2.8Ghz the processors are certainly the bottleneck.
it sounds like you are denying how hardware works more than anything else and your excuse is that you know something about the architecture. well maybe you might want to read up on AMD's less than stellar branch prediction and floating point/SIMD performance with mixed latencies before you post.Quote:
Originally Posted by gosh
"sounds" ?Quote:
Originally Posted by Chumbucket843
Try using core 2 and making small test applications where you time it. you will get VERY good results, better than a fast i7 in most small tests. Then test a more realistic scenario which is very hard to do just writing a small testapp. The result will be completely diffrent.
This isn't as easy as just seeing some graphs where the review may have selected tests to get desired results.
What? How did you come up with that conclusion? Do you have any tests to back up your conclusion? Sources? Oh and AMDZone.com doesn't count.Quote:
Originally Posted by gosh
Anyway, in the end we can conclude that PII needs at least 800MHz-1GHz advantage in clock speed to keep up with Core i5/Core i7.
Make that 2Ghz :rolleyes:
Back in reality,on average the Bloomfield is some 20-25% faster per clock than Deneb. You can make the Mhz numbers for "needed" Deneb from Bloomfied's clocks. The thing is that Bloomfield has a higher than that (average) lead in some workloads while in other it's not so much faster or is equal or even slower in few. So what you get in the end is about 20-25%,which is a healthy lead,a similar lead over Penryn which is some 5% faster than Deneb,which in turn coincides with a rough 20% lead every new intel design brought over previous one since P4 was sent to retirement(P4 not counted as "previous").
I have runed many tests myself and you can ask any developer to try. Or why not study some programming and do it your selfQuote:
Originally Posted by Clairvoyant129
OMG ^^... :rofl:Quote:
Originally Posted by gosh
Is this guy serious?
Thats right, that is why there are over 30 graphs. So you CAN get a firm understanding of each chip's benefits and pitfalls and see whether certain benchmarks favor what you (the end user) is going to be using their system for.Quote:
Originally Posted by gosh
Benchmark apps are for a reason... :up:
20-25%? Since when was the i7 920/i5 750 slower than the PII 965? I'm guessing you're not counting turbo boost since it's a "cheat" huh?Quote:
Originally Posted by informal
Penryn only 5%? Since you love quoting Anand, read this review and get back to me.
http://www.anandtech.com/cpuchipsets...spx?i=3551&p=8
Unlike you guys (v_rr, gosh and including yourself), I don't have an ulterior motive.
Today 99% can buy any processor about X3 or above and they will not notice any difference in speed.Quote:
Originally Posted by Xoulz
ehh... you know with 20-25% faster you said the same thing as Clairvoyant... maybe a differenz of 100mhz... :ROTF:Quote:
Originally Posted by informal
3.6Ghz deneb should overall match 3Ghz bloomfield(without turbo).Quote:
Originally Posted by Hornet331
why such a low res in this test... it cant hope to show gpu bottlenecking at least 1080p for xfire/sli setups...Quote:
Originally Posted by onethreehill
The Lynnfield Followup: Turbo Mode and Overclocking Investigated
http://anandtech.com/cpuchipsets/showdoc.aspx?i=3640
Butt test and benchmarks are two different things bro... we need deltas, so as a community, we can discuss the finer points of each. :up:Quote:
Originally Posted by gosh
But it would be interesting if it was possible to discuss the technology behindQuote:
Originally Posted by Xoulz
Not even close,Quote:
Originally Posted by informal
http://www.pcgameshardware.com/aid,6...iewed/Reviews/
Now back to reality.
Try again:Quote:
Originally Posted by Clairvoyant129
Plethora of tests,965BE 6% behind 2.93Ghz i7 with SMT and turbo: http://www.computerbase.de/artikel/h...formancerating
Also a lot of tests and comparison between 3Ghz CPUs(Deneb,Yorkfield,Bloomfield): http://www.planet3dnow.de/vbulletin/...#content_start
http://www.bit-tech.net/hardware/200...ition-review/4Quote:
Originally Posted by informal
We are talking about an average performance rating between these CPUs across various types of workloads,not about handpicking the workloads that i7 excels at(3d rendering and video encoding). Nice try though ;).Quote:
Originally Posted by Chumbucket843
http://www.phoronix.com/scan.php?pag...ynnfield&num=1Quote:
Originally Posted by Clairvoyant129
http://ixbtlabs.com/articles3/cpu/in...7-lga1156.html
Quote:
Originally Posted by gosh
AMD is a little bit late. Hope Thuban and new revision will come soon.Quote:
Update: after starting to see a flow of Windows-based reviews today, it looks like there are some more serious Linux + Lynnfield problems at hand, which we are currently investigating.
it is the opposite. Intel now has one processor that is equal to phenom II 955. AMD hadn't any real competition in that area before i5Quote:
Originally Posted by Olivon
:confused:Quote:
Originally Posted by gosh
That is what we are doing, isn't it?
Coincidentally, many are using graphs to illustrate certain points... that somehow you are saying is invalid, because YOU have different results on your butt dyno.
I find that rather odd.
Where?Quote:
Originally Posted by Xoulz
How do cache design from phenom and bloomfield differ and how does that effect performance in different loads?
How does prefetchers on intel work and how does prefetchers effect performance in different loads?
http://www.bit-tech.net/hardware/200...ition-review/5Quote:
Originally Posted by Chumbucket843
I d be suspect of a review that shows the 955 at stock having better numbers than a 965 at stock :/ , I m not really sure what the arguement is here, skimming these past pages... we all know Deneb and Penryn are very close within 5-10% depending on the bench/application/game, and we all know that bloomfield has a much higher ipc than both... So i m not sure where this discussion is going? Are we talking about threshold performance, where we need things to only be a certain speed within a cost window, ie i have 1400$ to spend and i need these tasks to be done, which is the best platform to achieve? Or are we comparing absolute values in which case i m not sure why the discussion isn't bloomfield vs LynnField and more importantly, i wish there was more testing of the pci-e bandwidth contraints compared, ie 4870x2's or gtx295's sli'd / xfired as with the 5870 and eventually the gt300 coming out, and both.. single versions of those cards being more powerful than both the 4870x2/gtx295.. seems pretty relevant that people would want to see how that compares, as the one review i did find that shows the lynnfield vs bloom on xfired 4870x2's shows what seems to be a choke in min FPS and in AVG fps, any links to more of this testing would be appreciated, because I am seriously considering one of these platforms for my next build,...
I think the dual vs triple channel and has been maxed out .. we all know heavy memory bandwidth apps are going to need bloom or more memory, and there is a plethora of tests showing the impact and latency advantages/disadvantages..
Yes i am very curious of the ramifications of the pci-e slots in the future, as well. EVGA hosted a p55 discussion panel with their designers two days ago, and in it they seemed to be intentionally vague, and even said things like "with current gen"...sounds to me like they are waffling on the impact, because they really arent sure.
I'm assuming that is in regards to SLI / CF but that vagueness from EVGA isn't a good sign. All this marketing with P55 for doing multigpu and if it turns out to be a poor show with next gen gpus that going to be a kick in the pants for board makers as well as early adopters.
yes they even admit to the fact that you arent gonna get high vga clocks on the boards equipped with the nf-200...which amazes me...what i dont get is this,
When was the last high powered GPU released that was x8 out of the box??? this must be a great trivia question....
so the REAL question is:
if its been so long since x8 was standard, why in the world are we manufacturing x8 SLI boards and slapping big fat CLASSIFIED and ENTHUSIAST and OVERCLOCKING stickers on them?: on this chipset designed for OEM!!!
almost comical
its like they are going backwards!!!
How cute,Quote:
Originally Posted by informal
You see, with those reviews I don't even know what kind of a test platform they used. It's funny how you need to go find obscure reviews in a foreign language.
You love quoting Anand but when I used his results to backup my conclusions, you go find some reviews in a different language and god knows what kind of a OS they used (since Linux favors PIIs and you seem to pick reviews where they used Linux as main OS)... thanks but it's obvious you can't pick any major reviews since they contradict your conclusion.
And lol, don't even bother with Planet3D... must be your favorite site huh?
Exactly.. Well I guess we ll see when the nda lifts for the 5870 how those do in xfire and if they bottleneck x8 pcie2.0 on lynnfield.Quote:
Originally Posted by Chickenfeed
Yeah because only English speaking reviewers know they are doing, what do Germans know about computers. Silly little country with only 80 million people. :rolleyes:Quote:
Originally Posted by Clairvoyant129
don't you think it is strange that most sites are testing same apps?Quote:
Originally Posted by Clairvoyant129
if tests are perfect that you should get suspicios because testing different apps is not perfect.
What I think is the best reviews is those that show up after those done on the release date. Reviews done in other languages often test other applications.
Quote:
Originally Posted by villa1n
Damn... I'm afraid that only ONE HD5870 could bottleneck x8 lanes in lynnfield :(
What exactlly do you mean? Why would you use only 8 lanes with single card?Quote:
Originally Posted by golden_boy
Quote:
Originally Posted by kl0012
Forget it, bad sleep and no coffe :rofl:
I meant that i'm afraid of a single 5870 or 5870x2 could be bottlenecked in lynnfield platform (750) beacause i was thinking in that combo for my next build :D
regards
Quote:
Originally Posted by Clairvoyant129
Yeah silly Germans and their "language" :p:,a "small" and insignificant country with no knowledge of IT/computers,no IT market and poor journalists .
/sarcasm
BTW those two "silly" computer websites just happen to be among the best German computer related websites around with huge community behind them.The reviews used a large pool of applications,covering office ,multimedia and gaming and drew an average from those.You may like it or not,but they done a great job.
Yep, in his mind German websites are somehow less knowledgeable just because they use "foreign" language and their reviews are somehow "obscure"(despite the fact they used massive amount of apps in their reviews). :p:Quote:
Originally Posted by v0dka
Quote:
Originally Posted by gosh
You don't get it do you?
Why do you think those apps were created.. ? What comes first, the chicken or the egg? We as consumers use apps, play games or crunch, etc.. These apps are designed and made to benchmark all aspects of a CPU/GPU's performance.
People are showing you different graphs, to illustrate the point you are trying to make is rather minuscule. As there are other graphs testing the same micro-archetecture/parts that benefits having a L3 cache, etc..
You simply dismiss them, because you are not being honest with yourself.
DID I say they are less knowledgeable? I think I need to spell it out for you since you don't seem to be able to comprehend what I'm saying.Quote:
Originally Posted by informal
Since it is in a foreign language, I have no idea how they tested their platform. (yes I see your pretty little charts but I have no clue their testing methodology). Do you get it now?
No, I am a programmer and it is possible to check how the processor performs doing test apps. I am also developing applications that need performance.Quote:
Originally Posted by Xoulz
There is a "System setup" page in every review so those have it to.Be creative and click (for example) pages 3,4 or 5 and you'll see it don't worry. BTW,there is a thing called google translate and it does the job great for the quick'n'dirty translations if you really need to know the details of what and how they tested.Quote:
Originally Posted by Clairvoyant129
Yes, thats what i worry about as well, 5870 x 2 or 5870x2, which is what i wanted to build out...but with 8x pci-e 2.0 have enough bandwidth for maxed out 1920x1080 resolutions 16af 8aa etc .. Which is what i d like to play the new games at, and with eyevision, maybe grab another monitor as well.. so.. its pretty pertinent. :) I like the lynnfield because the oc results seem pretty easy to hit 4.0ghz 24/7 on air, with the right airflow.. versus the 920 where there's a bit of lucky chip factor involved still. That and the mobo's for x58 are about 70$ moreQuote:
Originally Posted by golden_boy
But even though you get x16 if you use a single slot, if you put a raid card or sound card or dvr card, whatever, on the second x16 slot it still knocks you down to x8 on that SINGLE gpu..i would not like being restricted like that.
You could just use a regular pci slot ^^ , i guess though a raid card is usually pci-e. Yeah for workstation type uses , not the ideal platform, but thats what x58 is for :PQuote:
Originally Posted by Computurd
Quote:
Originally Posted by Computurd
Or just use the pci-e slot wich dont split down to x8.. :rofl: there is boards that just use x16 / x4, or x8 / x8 and 8x / x8 / x4.Quote:
Originally Posted by villa1n
Most of the 1156 MB comming with x16 (x8x8)/ x1/ x1. Right now PCIe 2.0 x1 is enough for most periphery devices (including one SATA 6GB device).Quote:
Originally Posted by Migi06
These new Core i7 do appeal to me. The i7 860 is slightly faster than the old i7 920 and consumes less power which is another bonus and is priced closely. I may considering getting this with the P55 board. Does anybody know does the new Core i7 generates less heat than the older i7's?
yes the new i7's 860/870 do generate less heat and eat less also but they are not the ideal choice if you have the cash get a i7 920 instead.
Well what I can see the 1156 cpus reminds me more of the old stepping Bloomfields. They require more volt but also run a bit cooler. I run my rather crappy sample i7-860 at 4.0GHz 1.41v (bios) / 1.376v CPU-Z and I get 33~37C idle and 68~71C load with HT disabled and 73~75C with HT enabled in LinX with a CoolerMaster Hyper 212 Plus & Arctic Cooling MX-3 with stock fan @ ~1700rpm in 20~21C room temp. When gaming ~55C.
I could have afford X58 but I preferred choosing P55.
Q9550/9450 is quite old ... But I don't think you're serious :yepp:Quote:
Originally Posted by gosh
Intel Core i7-870 Review
http://www.nordichardware.com/Reviews/?skrivelse=559
No the guy is serious. There is actually a handful of these people over at amdzone who actually believe this. :shakes: Once in awhile they forget where they're at and start splewging out nonsense like he did and think that normal people will accept it like they do over at bizarrozone.Quote:
Originally Posted by Olivon
maybe you should read this document: http://software.intel.com/en-us/blog...t-prefetching/Quote:
Originally Posted by qurious63ss
Intel do have big focus on prefetchers and that makes the cpu fast in simpler loads, they don't work in heavier loads
Can you show some examples?Quote:
Originally Posted by gosh
do you not understand the document?Quote:
Originally Posted by gallag
Quote:
Originally Posted by gosh
No where does it say it can't handle heavy work loads.Quote:
it doesn’t help but likely won’t hurt performance either.
But just for fun, let me rub it in your face so you can have trouble sleeping at night.
Nehalem is better than Deneb, light or heavy workload. If Deneb is so much superior under 100% CPU usage, why is it that professionals and workstation users prefer i7?
Remember that it is a document from Intel. You just have to read about how prefetching works and how it affects the cpu and then it is simple to draw conclusions.Quote:
Originally Posted by Clairvoyant129
Workstations that do computational work should gain a lot from prefetchers. Even if there is very advanced algorithms that type of load is very simple to optimize and it needs very little code to execute.
Compare this when one advanced game need to calculate all information from a very complicated game scene. A lot of enemies (code for how they behave). AI for the enemies. Physics and code for that, explosions, wind, fire etc. All this needs very much code to be calculated.
If you just want to do a face which is looks very real with muscle movements etc. it is VERY complex.
Check 79 minutes in to this presentation from John Carmack when he describes what is complicated to do in games.
QuakeCon 09: John Carmack Keynote
Just calculating triangles is a very simple task, so simple that now with DX11 the gpu can do it (tesselation). In simple scenarios in games where streamlined computation doing triangle caluculations, prefetchers on i5/i7 would work very well for better performance. I also think that is the reason why i7 in some performance test is so fast and in others it is loosing, the L1 and L2 cache isn't that big, it is designed to be fast and work well with prefetchers but when it doesn't work it isn't that fast anymore. There AMD is faster.