No need to make this up. The same big name "bhavv" banging about vram there, saying "lol I was like a ghost chasing you" or something. Or you are so famous that someone would like to pretend to be you?
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
[In]formal proof of the limitation of 1.5GB VRAM at 1920x1200 resolution
- Thread starter harmony
- Start date
- Status
No need to make this up. The same big name "bhavv" banging about vram there, saying "lol I was like a ghost chasing you" or something. Or you are so famous that someone would like to pretend to be you?
NAME THE FORUM.
Obviously someone pretending to me if they had an IP from Saudi Arabia you dumb **** and I want to look up where this bloody happened.
Or its simply more likely that it was someone with the same name as me, you are aware how common my name actually is arent you?
... Or as I suspect, you are completely making it up.
Last edited:
Soldato
- Joined
- 6 Feb 2010
- Posts
- 14,583
^This.
While SLI and Crossfire has vastly improved over the years, it is still NOWHERE near perfect. bit-tech even go as far as not recommending SLI and Crossfire due to inconsistence reliability and performance.
For example, two users with same pair of graphic cards and using the same driver...one can have the SLI/Crossfire running perfectly fine, the other can run into problems getting it work or have performance issue. Trying to compare how VRAM impact on performance in such an inconsistent environment simplily isn't going to deliver reliable results for comparison.
Then this has to be done with the 28nm flagship cards, if there is going to be a single GPU capable of delivering the GPU power of GTX580 SLI, and if there are going to be such 1GB, 1.5GB models avaiable.
Currently the most powerful 40nm GPU is the GF110, which is still not able to sustain high framerates (e.g. 60fps) to contrast against lag spikes.
BTW, a single GF110 is capable of WoW dual instance, however I don't know how to repeat test scenes in a precise way.
Last edited:
HARMONY Name the other forum that you use that you have stated that I was apparently a banned member of.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
You either:
a) still don't agree that lower average fps is NOT the only symptom of vram shortage. still don't believe in the tale of "lag spikes / stuttering".
b) partially agree with the tale of "lag spikes / stuttering" but practically take performance / price ratio into account, not considering vram seriously, because GPU power is of first priority.
c) simply making fun of me, just like bhavv
I'm not really expecting a serious response from you but from the other posters. Has Harmony provided any proof of this increase in lag spikes/stuttering in the past 5-6 pages that I haven't bothered reading, or is this proof of lag spikes, the whole one maybe two extra spikes in 3000 frames in the first page of graphs that were posted?
Because I'll point out, 2 gpu's on one slot are almost certainly going to spike more than 2 separate cards with essentially double the pci-e bandwidth, secondly, a couple of spikes is frankly within the margin of error.
Back to asking for a response directly from Harmony, do you have any proof anywhere that ANY of the games you've posted with a higher APPARENT memory usage actually REQUIRE that amount of memory?
What do almost all cpu's, gpu's, os's and many many programs do, what's incredibly common and has been for decades of computing.... the answer you're looking for is cache.
If you've loaded a texture onto a GPU what is the point in deleting it, only to have to reload it again, if you don't have to?
You could quite easily have Nvidia make you a 6gb version of a card, and potentially fill the full 6gb's up with data, slowly as the game simply keeps things it may use in the future in memory.
Now there is a potential performance upside there, and even a power saving potential(minor in both cases), but you've failed at any time to demonstrate the clear performance loss you've claimed over and over again, and when provided with graphs/benchmarks that actually show losing performance due to lack of memory, you completely ignore it.
YOu also seem to be campaigning like you were the first one to notice this entirely new phenomena of lack of memory. Lack of system memory, video memory, cache, page file even, they all share a VERY common and VERY obvious trait that is when they run out, and you have to drop down to the next level of data storage, you get significantly less performance, not marginal to the point where you're scraping the barrel trying to prove that 3 frames are slower out of 3000, on a different system, that are within the margin of error/variability anyway.
Each and every thread, is not new, you fail to answer simple questions, you fail to provide any proof and with each thread, your arguments seem to get more ridiculous.
Like I said, you feel SO strongly about this, e-mail AMD, and Nvidia, and Anandtech, and [H]ocp and a bunch of other review sites and ask them to look into it, ask AMD/Nvidia direct questions about memory and lag spikes and stuttering.
Lastly, realise people are more pee'd off with you and the way you post, than the actual subject matter.
Everyone, bar none, here would buy a card with more memory IF it was good value to do so, £10 more for a 2gb version of a 6950, no brainer, £50 more for a 2gb version of a 560ti, also a no brainer, no one should buy the 2gb version of that card as its horrific value.
£150, you can't get a performance 2gb card, from £200-300, AMD have value at 6950, and the 6970 is almost always competing with the 580gtx these days, quite often beating it now. Basically, which 1gb cards can you buy where the 2gb card is the better option.... none. 2.5gb 570gtx, lol, 3gb 580gtx's, awful value, 6950/6970, 2gb anyway, below 6950 in price, everything is 1gb anyway.
Just which cards does your multiple thread requiring argument effect......... I can't think of a single card.
£200 for a 560ti with 2gb, well a 6950 is MUCH faster than it anyway, 570gtx 2.5gb, 6970 is faster and cheaper than the 1.25gb version, 3gb 580gtx.......6970's have the memory, sometimes faster.
Seriously, which cards does this actually effect........ all the pretty obvious choices at every price bracket either have no 2gb version, or come with 2gb+ anyway.
I'd give you + rep for this if I could, but I cant, so I will just have to say:
^^ +1.
Something that Harmony cant and will never understand - I bought my MSI Twinfrozr 1 Gb GTX 560 Tis on launch at £199 each:
1) AMD 6950 2 Gb cost £240-£260 at this point
2) AMD 6950s were not available with cusom coolers
3) I wanted an MSI Twinfrozr II card and nothing else
4) My normal GPU budget is £150 per card. With the GTX 460 / 560 I went over that budget because my 5770s were not powerful enough for me.
5) today, 2 Gb GTX 560 Tis still cost around £50 than a 1 Gb version.
Last edited:
Soldato
- Joined
- 23 Nov 2009
- Posts
- 6,038
- Location
- North Leicestershire
Will someone send me a 6950 1Gb card for a few days. I will quash this thread once and for all in favour of either party @1920x1080
i had stalker cop,gta4,crysis2 all showing over 1gb of vram usage crysis 2 being the most at 1890Mb at one point.
Would be interesting to stick a 1gb card in my rig and see what happens to the frame rates.
i'm not siding with either party until i see some hard evidence and not all these conflicting benchmarks that keep being posted.
i had stalker cop,gta4,crysis2 all showing over 1gb of vram usage crysis 2 being the most at 1890Mb at one point.
Would be interesting to stick a 1gb card in my rig and see what happens to the frame rates.
i'm not siding with either party until i see some hard evidence and not all these conflicting benchmarks that keep being posted.
You are obviously not good at memory management. Read more stuff about pin page, page fault, thrashing. If you REALLY understand these terms in the memory hierarchy, then you should understand that the performance may not always be affected all the time (aka average fps), but instead, may only be affected during short periods of times (aka lag spikes). It is within this margin that makes this topic controversial over decades. Some people can notice this, some cannot. Two groups cannot persuade each other.
Last edited:
When you exceed your Vram limit, the extra data that needs to go to the Vram goes into the shared video ram instead. At the most on a properly coded and optimized PC game, this reduces FPS by a whopping 2-3 FPS at the most.
The exception to this rule is lousy console ports such as GTA4 and DA2 which were purely developed for consoles and not optimized for PC hardware. Also another factor to consider is that review sites that do any tests regarding Vram comparisons are also completely ignorant to the issue of shared ram, and they do the comparisons on PCs with only 4-6 Gb system ram.
Edit - Point proven:
http://www.hardocp.com/article/2011/10/17/msi_r6950_1gb_twin_frozr_iii_power_edition_review/2
Only 6 Gb system ram, not enough for a 4 Gb shared ram size + enough ram for your game.
Last edited:
Harmony, given that you have SLI you probably suffer from microstuttering more than the rest of forum do lag spikes from a lack of VRAM. Give up the crusade already.
Unfortunately no. The microstuttering is already shown in my plots, with lag spikes a lot more obvious than the microstuttering.
I don't think you can. Either result you get would attract questions and attacks from the party that the result is against.
A 6950 is not much faster at all than a 560ti, otherwise some valid points in a well constructed post, DM.
I would say the OP's argument focuses on extreme set-ups with users that are sensitive to subtle changes and expect a minimum of 60fps all the time. Otherwise for a majority of people playing at (in my case 1920x1200), a couple of 1+gig cards should be fine. If it's worth not too much more cost for a bit more memory then fair enough.
I would say the OP's argument focuses on extreme set-ups with users that are sensitive to subtle changes and expect a minimum of 60fps all the time. Otherwise for a majority of people playing at (in my case 1920x1200), a couple of 1+gig cards should be fine. If it's worth not too much more cost for a bit more memory then fair enough.
No it isnt.
And can I ask you again to please provide the name and a link to the other forum that you use for this discussion that you keep on accusing me of having been account banned from? I really need to check into whether or not someone was pretending to be me thanks.
You cannot observe the fact that the curves are more thick in the vertical direction than in the horizontal direction? That is obviously microstuttering.
No I dont see any proof of microstuttering there at all.
All I see is that a GTX 590 is clearly better than a pair of 3 Gb GTX 580s, if the vertical axis = frame rates.
Also, you still arent naming the other forum you are posting this on, why do you continue to ignore my request to check what seem to be more made up unreal claims you have made about me in this thread?
All I see is that a GTX 590 is clearly better than a pair of 3 Gb GTX 580s, if the vertical axis = frame rates.
Also, you still arent naming the other forum you are posting this on, why do you continue to ignore my request to check what seem to be more made up unreal claims you have made about me in this thread?
Different games suffer from different degree of microstuttering. With slim curves:
I forgot to lable the axis. The x-axis is the frame number, and the y-axis is the time spent to render each frame. The GTX590 curve is worse. These plots are in the same manner as those used by Techreport.
I forgot to lable the axis. The x-axis is the frame number, and the y-axis is the time spent to render each frame. The GTX590 curve is worse. These plots are in the same manner as those used by Techreport.
- Status