Intel Core i7/X58 gaming performance

[Sharrow 0]

Hi allThe Core i7/x58 NDA was lifted today and the web is awash with reviews. As expected there are a lot of opinions out there (some contradictory) but the general gist seems to be that Core i7/x58 is going to bring limited performance increases in games.Anandtech.com has one of the most informative articles on Core i7/x58 I have read. Available here: http://www.anandtech.com/cpuchipsets/intel...doc.aspx?i=3448It's a big one so here are some pertinent quotes:"Looking at Nehalem's microarchitecture one thing becomes very clear: this is a CPU designed to address Intel's shortcomings in the server space.""Intel opted for a very Opteron-like cache hierarchy with Nehalem, each core gets a small L2 cache and they all sit behind one large, shared L3 cache. This sort of a setup benefits large codebase applications that are also well threaded, for example the type of things you'd encounter in a database server. The problem is that the CPU launching today, the Core i7, is designed to be used in a desktop.""Gaming performance is actually better than expected for Nehalem, there were enough cases where the new architecture pulled ahead despite its very small L2 cache that I wouldn't mind recommending it for gamers. In most GPU limited situations however you won't see any performance improvement, at least with today's GPUs, over Penryn.""It seems odd debating over the usefulness of a processor that can easily offer a 20 - 40% increase in performance, the issue is that the advantages are very specific in their nature. While Conroe reset the entire board, Nehalem is very targeted in where it improves performance the most. That is one benefit of the tick-tock model however, if Intel was too aggressive (or conservative?) with this design then it only needs to last two years before it's replaced with something else. I am guessing that once Intel moves to 32nm however, L2 cache sizes will increase once more and perhaps bring greater performance to all applications."-----There are of course no benchmarks for FSX anywhere though given how CPU bound FSX can be it would be very interesting to see some performance comparisons. Looks like those running a Q6600 or better will see very little performance increase by moving to Core i7. Given the excessive cost of X58 motherboards (and moving to DDR3 for those on DDR2), it makes very little sense from a performance for your buck perspective as well.If anybody comes across anything Core i7 and FSX related in the next few weeks please shout!Konrad

[Guest D17S]

From an earlier post: "Early benchies. HardOCP ran some FarCry2 benchmark runs today. I have the game and played along. This is a canned benchmark run provided with FC2. The 1st pic is Nehalem at 3.2/19x14. The second is my Q6600 at 3.6/19x10. Clock for clock, the Nehalem runs a ~ 15% advantage. That's tracking pretty well with what we're hearing.And look at those settings. 50+ (silkysmooth) FPS at 19x10 with a 9800. DX10, Bloom, Physics, Shadows, AG full blast, traffic (i.e., bad guys everywhere). I hope aces is watching all this.http://forums.avsim.net/user_files/194011.jpghttp://forums.avsim.net/user_files/194012.jpgThese little 10% bumps are sooo irrelevant. Consider too, that Intel has installed thermal throttling to prevent us from O/Cing our way from a $175 Q6600 to (even beyond) a $1400 QXYZ. They're on to us. The next big deal will be FS11 and the Sandybridge shrink to 40 cores. That'll our the next 10X opportunity (Remember? That $275 P4 @ 2.8 to a $275 Core2Quad @ 3.6). From here on out, it'll be about number of cores. Come-on software.

[Tabs 2,790]

"In most GPU limited situations however you won't see any performance improvement, at least with today's GPUs, over Penryn."This is what I'm curious about though - FSX is not a "GPU limited situation", it's heavily CPU dependent. We'll have to wait for someone to take the plunge but this CPU might have the potential for significant gains in FSX...

[Sharrow 0]

Ryan, that's the hope but I suspect that expecting significant gains in FSX over C2Q may be optimistic. I very much get the sense that only heavily multi-thread optimised apps which utilize HTT are going to benefit in any significant way from i7. The castrated cache of the i7 compared to high end C2Q seems to be a big factor in all this.My one other point is that i7 is looking like it will still not be able to handle FSX like a lot of folks would want it to. Even if we end up with a 15% overall increase over C2Q what will that translate to in real world FPS - 5 or perhaps 10? Aces always suggested that FSX was really developed for next gen hardware. Well, here we are and the situation seems to have changed little...I would not like to be in the position of having to buy a new desktop for FSX in the next 2 months - a lot more tough choices today than there were only 2 days ago.Konrad

[Tabs 2,790]

Which is why I seriously hope they're developing FS11 for current hardware this time like basically every other game in existence...

[Sharrow 0]

You gotta admit that the life span of a normal game when compared to FS is not the same thing. Look at FS9, what's that - 5 and half years and still going strong?! You know anyone playing Far Cry or Doom3? Aces need to have a much longer term view with FS than your average game developer.I don't know with FS11 - with DX11 and Windows 7 around the corner, the unpopularity of Vista/DX10, another "tick" from Intel by the end of 2009... if you were Aces what on earth would you develop for? Tough one that I reckon.Konrad

[tfm 169]

All this stuff about "programming for the next generation of CPUs" doesn't wash with me. I think Aces just messed up with FSX, probably under pressure from MS managers to get everything ready in time to cross-sell Vista.When they planned FSX, I think they ASSUMED that by the time of release the old Pentiums would be running at 6+GHz - which was indeed a common expectation about 4 years ago - with still more Hertz under the bonnet to come. What they planned for was a game that WOULD work with "current" generation CPUs: what they got wrong, was what "current" generation CPUs would be able to handle. By the time FSX was due to come to the market, they realised their mistake: basically, CPUs had developed more cores instead of more Hertz. But by now, of course, it was too late to re-code FSX to take advantage of the extra cores. Some excuse was evidently required, so the mantra was born that "we programmed for the technology of tomorrow". It is just more marketing spin, IMHO.It is bad enough that FSX is sub-optimal for multi-core CPUs. But what makes it worse, IMHO, is that it now looks as if is also sub-optimally programmed even for single-core operations. Yes, FSX has to work harder than most 3D applications. But Phil Taylor himself has given what IMHO is tantamount to an admission of inherent coding defects, more than once (for example, I recall reading one of his posts as pretty much an acknowledgement that the application makes memory calls inefficiently). Another way to test this is by noting that Aces conventionally blame much of FSX's sluggish performance on the need to maintain backwards compatibility with older add-ons: but what compatibility is that, exactly? So far as I can tell, many of the worthwhile FS9 add-ons are broken under FSX. To my mind, this is further evidence of the unseemly rush to get this product to market so as to cross-sell Vista. I think they just hoped - and assumed - that the raw power of the Pentiums they expected to be running at 6+GHz would cover the tracks of sub-optimal programming.Aces or, I suspect, MS management have a lot to answer for. IMHO, the main target of FSX was not the 50 dollars or so for buying the game, but the hope that we'd fork out the 300 dollars or so for a nice new Vista O/S to run it on and - who knows - perhaps some other nice bells and whistles from the MS stable while we were in the mood for "upgrading". IMHO - though I have come to love it as much as loathe it - FSX was in some respects a misconceived product, prematurely released and marketed with breathtaking cynicism as well as recklessness for the truth (as I have mentioned before, my FSX box has on it a sticker saying something like "Works best with Vista" - which IMHO was an unsupportable assertion at time of launch).But on the other hand, whenever I have a rant like this, I need to take a deep breath and remember that what they DID manage to produce is still pretty impressive. Using the default aircraft with my current PC, I now seldom get <40fps even in dense scenery (though with the decent add-ons, it's a different story of course ...) Overall, I think that FSX has raised the bar: it's just a pity that it's inflicted such IMHO unnecessary agony on us in the process.Tim

[Guest XM DUDE]

I don't see myself buying FS11 as I might need a new computer for that too. I would be happy running FSX. I see a trend here, new OS+ new flight sim =needing new computer. The trend just keeps repeating itself.Michael P.

[michal 44]

>Aces conventionally blame much of FSX's sluggish performance>on the need to maintain backwards compatibility with older>add-onsThat is simply not true. I read all their blogs and they never said things like that. I know it helps with rant when you can invent things on the way ... This "backward compatibility" versus "performance" got to be the most notorious urban legend in the flight simulation.Michael J.http://img142.imageshack.us/img142/9320/apollo17vf7.jpg

[Guest D17S]

It is interesting to note though, that HardOCP's benchmarks were run with a BFG GTX280/OC. My runs were with a 9800GTX driving 2 x 40" Samungs at 19x10 each. There was virtually no difference. This could suggest that the FPS metric was being stopped-out by some other component, other than the Vcard. In these cases, it appears these runs are CPU limited. I had been playing the game, but had never had a frame counter up. It just worked. Every thing was at Very High, looked great and was running smooth. When I initially saw the frames counter in the benchie, I was amazed. "Where are all these frames coming from?" I pondered. Being used to FSX, scenery like that ought to be a slide show.I then pulled up TM and watched CPU loading. All 4 cores were evenly engaged at 25% - 50% each. That's where the frames are coming from. The software guys are - Finally - figuring our how to use multicore CPUs. These slowpoke quads are starting to show their potential. A fast(er) dual is a + 20%. A slow(er) quad is worth + 2X.

[michal 44]

>The software guys are - Finally - figuring our>how to use multicore CPUs. I hope flight simulation should not be far behind in this respect, there is a fair amount of parallel processing in a typical simulator - painting different parts of avionics/cockpit, weather, terrain, AI, etc. all could take advantage of parallel processing.Michael J.http://img142.imageshack.us/img142/9320/apollo17vf7.jpg

[Mace 1,902]

Also tim don't forget software rendering engine.He said that the basic FS8/9/10 architecture was originally designed for software rendering. In other words, for the time before 3D graphic cards were prevalent. Back in those days, software rendering was the way around lacking the hardware--the drawback was, that the software renderer was much slower than the hardware renderer could be.That's mainly why we're so CPU dependent relative to most other entertainment titles.RhettFS box: E8500 (@ 3.80 ghz), AC Freezer 7 Pro, ASUS P5E3 Premium, BFG 8800GTX 756 (nVidia 169 WHQL), 4gb DDR3 1600 Patriot Cas7 7-7-7-20 (2T), PC Power 750, WD 150gb 10000rpm Raptor, Seagate 500gb, Silverstone TJ09 case, Vista Ultimate 64ASX Client: AMD 3700+ (@ 2.6 ghz), 7800GT

[tfm 169]

Hmmm. I think if you go back and read Phil Taylor's blog your memory may be jogged. Sadly the search function seems to have been taken down, but I'm sure you'll find at least one reference if you look. Tim

[teropa 4]

Now that someone touched the question of "next gen HW", and we have finally arrived at that station, I think it's also possible to discuss FSX and the future of FS.I think for FSnext expectations need to come down a lot.For one there is still the utopia of wanting to see almost life-like environment (as in some much simpler environments in other "games"). Given the fact that Nehalem is not going to be the savior of FSX, in a way that PT and ACES in their marketing talk suggested (next gen HW), I can make an educated guess that folks need to delete those DX10 fantasy images from their hard drives, for good. Sure, there will be developments on the GPU front, but for this to benefit FS, the core needs to be rewritten not to crash the performance for FSnext (if anything similar to those DX10 photos were to be introduced in FSnext).What we are faced with are the brutal realities of past, present and the future. If FSnext is coded similarly as previous titles, you can not expect both big graphical improvements and increased performance. Add to that the dismal usage of multi-core CPUs in FSX, and you have a realistic base for "past", and ... yikes, "present" as well. As with some large airplanes, you can't slow down and come down at the same time. For FSnext to evolve graphically and in terms of performance the equation seems to be nearly impossible. Unless, ACES takes the same road again, and designs FSnext for HWnext! (you'd think they had learned something with FSX, wouldn't you) OR if they decide to do it all from the scratch, which I doubt.The earlier comparison between Hz and Corecount was bang on. And also the reasoning behind the fact that FS is CPU-bound.What is then realistic for next FS, and what HW is there to run it? To be honest, I don't expect much. Nehalem is here, and we already have some idea about how it is running. Not the full picture by any means, but a glance. And it's a sufficient glance.I suspect they'd need to code an entirely new FS engine, that takes full advantage of CURRENT HW, multi-core processors (not all customers can afford QEXTREMEs or i7s to run a "game title"!). They will definitely be looking at how to improve performance. But the question is, compared to what?. Most likely there will not be any life-like graphics, there will not be sloped runways, realistic ground friction etc. etc. They will probably re-do some of the core code (maybe all of it), and introduce bug-fixes for the Weather System, throw in a couple of goodies from the GPU side, and the new sim will still look almost exactly like FSX. Maybe it runs a little better, and it might even have a new UI.But looking at the FSX story in full view, and comparing FS9 to FSX still gives me shudders. This "upgrade" (to end user) was almost something like we see with the EA Sports NHL Hockey titles. No one except HC fans will be able to tell any difference. Wow, you can fly using a game-pad! Who cares. If MS/ACES is trying to recover from this disaster that FSX ultimately is, they need to do a lot of work. Given the current financial situation, the aim is to generate CASH. And cash is generated by selling and keeping the product cost low. And btw, I also remember the postings about blaming backward compatibility for bad performance. The wording is not exact, but it went something along these lines: to ensure backward compatibility, which is very important for our customers, we could not re-write the FS Engine (that would have increased performance). This would have broken all backward compatibility, and we don't want that to happen. So instead, you have good backward compatibility, and bad performance. Maybe we will reconsider this stance for the next FS. (a little sarcasm added there)My expectations are very low. If they happen to surprise me, I will eat all my words, and drink to them! :)Tero

[Sharrow 0]

Surely Aces still have more SP's for FSX in the pipeline before FSNext is released? Over and above sorting out the DX10 issue (which is meaningless for me on XP anyway) the other point of focus for Aces must be improving overall efficiency and multi-core cpu useage. Aces may have missed the mark upon release by coding FSX to take a brute force approach (which made sense at the time with folks expecting 6+GHz clock speeds), but surely now they are well aware of the muti-core trend going forward. I am not suggesting they re-code FSX from scratch as that is obviously not possible but I am sure there is a lot they can, and will, still do to improve overall performance in FSX.With all the money and effort currently going into developing FSX SDK add-ons I cannot see how MS/Aces would not be under pressure to sort out as much as they can with FSX and thereby increase its life span to something approximating that of FS9.That is my hope, at least!Konrad