Originally posted by: apoppin
That must have been "some dinner" 
... you are a lot more mellow and ... strangely, we're *agreed*
... 32-bit is starting to show its limitations in gaming. i also think such a rig as you mention *should* be tested with both 64-bit and 32-bit - at least to lay this ghost to rest. And in the future, they should test on 64-bit so there is no "question" about any limitation whatsoever.
Well that was my point with this review and agreeing with the OP. That they said they would be doing a final 32-bit vs. 64-bit comparison after the 8800GT review and using 64-bit going forward, then turn around with Tri-SLI Ultras with QX9850 or whatever, WITH 4GB and are still using a 32-bit OS while running games with EVERYTHING maxed out at 2560 w/ AA.....
And while I would tend to agree saying 32-bit is "no longer valid" is a bit harsh, its clearly no longer "high end." It also does bring some of their results and conclusions into question, as the German site that did the first circulated Tri-SLI review showed much better scaling with the 3rd card. AT's review shows very little scaling with the 3rd card. Whether or not that's due to 64-bit is inconclusive and not something I'm going to go back and pore over, but again, why leave that to question on a free OS upgrade when testing $5000+ worth of hardware?
... i *know* the difference between 1 GB and 2GB And if you benchmark a 1GB rig with an otherwise identical 2GB rig, you will get less performance - it is very noticeable. PREpatch, i *knew* when Hg"L ran out of 2GB memory - it was beyond obvious [chug ... chug .. stall .. watch the HD light light up -- chug ... chug ... maybe return to normal ... maybe crash]. The "problem games" - CoH/Stalker/SC/Hg:L - were not programmed right - the devs didn't plan for 512MB video cards and "only" 2GB of addressable space. Fortunately the hotfix is a decent workaround for these games.
Again, that's not necessarily true if you use FPS as a gauge of performance. I can honestly say its about a coin-flip whether thrashing drops FPS or not. Many games will basically render the last full frame at the same FPS and then "lurch forward" or "thrash" once the new textures are loaded and rendered and although FPS does not drop at all, its clearly worst in terms of performance. A very good example of this is demonstrated in Fly-By demos, time demos or static travel routes and is often cited by review sites as they'll usually toss out the first result for texture loading. Thing is, even though the run can be noticeably jerkier, the difference in FPS is often negligible. Crysis demo is a good recent example of this.
Even if you used FPS as a gauge and it resulted in a drop in FPS, averaged in with the rest of the test would result in very little impact although what you're visually seeing would significantly impact your gaming experience.
In other games its less obvious unless you measure in load times, particularly in FPS titles since they typically manage to fit all the textures needed for a particular level in a small footprint. The difference though is when you rotate back to a previous map, the load times are significantly faster. The BF-series games have always been a good example of this.
Open area games or games where you zone a lot are another good example. Re-loads are much faster with more RAM, but in open area/contiguous games the difference is also quite obvious. Its similar to adjusting rendering distance, with more RAM you're less likely to see "texture/object popping" as you move and suddenly stuff starts popping up on your screen as the textures are finally loaded from the HDD. With more RAM, either the game engine loads it up with the rest of the zone textures, or they're not flushed as you move away.
i think *you're* starting to finally get it
Of course it is SC's [devs] fault ... they didn't plan far ahead enough; AT found their own workaround - the SC devs had RCs of Vista to work with - at any rate, the hotfix *works*
There's no way its the devs fault...again, AT ran into this problem with every game they tested as Ryan Smith posted in a comment to the article. The increase in memory use was a change that MS forced with their new WDDM driver model in an attempt to increase OS stability and remove various "problematic" subsystem stacks from the OS kernel. Its definitely worked to some degree but their initial implementation had the nasty side-effect of reducing virtual memory to unacceptable levels on a 32-bit OS.
Now, considering Vista is touted to run existing 32-bit applications seamlessly and devs are coding for XP and 32-bit coding practices (2GB, up to 3GB with /laa), a superficial 256-768MB reduction in virtual address space is significant. If MS didn't hot fix a problem they spawned, the other solution would be for Devs to change their code and games to recognize 32-bit Vista which would be a poor solution that would ultimately result in worst performance, slow adoption, and continued compatibility problems.
Again ... the future. It is a bit of a Catch-22: No one will upgrade until games are specifically written for or ported to 64-bit. And no dev wants to port games to 64-bit or code for them without a large HW base. But don't lay the blame at my feet. Blame MS ... or intel ... for dragging their feet.
Sure its a catch-22 but most parts are in place already. Chances are anyone who built or bought a new PC in the last year is running a 64-bit CPU and a 64-bit capable system. MS has done a few things to force people to switch and have been blasted for it as a result (making DX10 exclusive for Vista etc.). The only thing they maybe could've done better would be to include both 32 and 64 bit DVD versions in ALL Vista packages, but at least they aren't charging for 2 different OSes like they did with XP and XP64. Devs are clearly more supportive of Vista and x64 with DX10 features and have begun flagging /largeaddressaware for their games (for 64-bit only). The only thing left is for people to actually switch to the OS, and as I've acknowledged, that was certainly something to be wary of at launch but something I'm 100% comfortable recommending TODAY with tangible benefits TODAY.
i still will agree to disagree - until we *see* some PROOF in [perhaps] a follow-up article by AT - i don't think 64-bit gives any *practical* advantage [today] over 32-bit in gaming. That all games are currently developed with the 2GB barrier in mind.[/b]
Well again, AT is supposed to follow-up with a 32 vs 64 bit update, hopefully they've been monitoring this thread. If they just run a bunch of FPS graphs at different resolutions that'd be a poor review. Testing things such as load times, different settings, HDD use/hard faults and even FRAPs'ing gameplay would be much more relevant.
And not all games are being developed now with ONLY the 2GB barrier in mind as many titles since the AT review have released /largeaddressaware patches or have the option from launch and in those games there's clearly a benefit from 64-bit and 4GB+. Whether or not you think its tangible or practical isn't my concern.
But anyways, looks like you'll be able to do some testing yourself soon. Should've gone with 2x2GB imo leaving you room for 4x2GB or 2x2GB + 2x1GB.
Honestly I don't think you'll see much difference in a 32-bit OS since that extra physical RAM will just make your OS a bit snappier, which is what I saw with XP. Next upgrade will be to 64-bit which is free for you anyways.
Facebook
Twitter