knopflerbruce
-
Posts
4290 -
Joined
-
Last visited
Content Type
Profiles
Forums
Events
Blogs
Posts posted by knopflerbruce
-
-
Sigh - do you know what the visible difference between 9800GTX and GTX+ is - the clocks only- do you know what the visible difference between a 6800gt and 6800 Ultra is - the clocks only...
Both examples have same device main IDs and only differ on the clock rates, a 6800 gt and ultra for example are recognized as 6800 series both, there is no ultra or gt shown, the 9800 models are recognized at 9800GTX/GTX+. And the 9800GTX and 9800GTX+ also was made by nividia, official cards, this was not a board partners choice and all did it... I really have no idea what you want to tell me, but it is wrong anyway
websmile`s Aquamark score: 105691 marks with a GeForce 6800 GT
websmile`s Aquamark score: 116479 marks with a GeForce 6800 Ultra
To top it, here is number 1 at gt category
T0lsty`s Aquamark score: 142353 marks with a GeForce 6800 GT
Here is number one at Ultra category
K404`s Aquamark score: 132104 marks with a GeForce 6800 Ultra
even same core clocks because one is oc model
Making a correct guideline for a possible regroup will stir upa lot of sh.. and these examples only show that a lot was handled differently at hwbot, if you like to admit it or not. I would prefer to not discuss this anymore unless you really get into the topic. You know a lot more most likely about AMD cpus than me, but this vga discussion makes no sense when you do not check the facts
What facts? 9800GTX were split at first, but as it was not possible to verify that any given sub card was done with a (+) or regular model the categories were merged. You can to this day easily separate a 6800 GT from an Ultra, and therefore both categories are still there. And by the same logic, Mars and 295 should be split, they're more different than 6800GT vs Ultra - and from what I know there never was a discussion regarding these Mars GPUs. The whole thing went below the radar - as far as I can tell. Hence it's not an older decision that's changed, this is the first time this problem has been discussed.
I don't get your reasoning at all - surely we must agree that there should be a new category here when you provide proof that HWBot split much more similar cards than Mars vs 295 several times already?
-
Because ASUS chose to use the device ID of the GTX 295 for convenience does not make it a GTX 295.
Exactly. They could've made it look like a 6200 Turbocache if they wanted to, I guess. They chose the closest match, which is fair, but it doesn't make it 100% right.
I wonder why people started subbing this as a 295 in the first place when it was clear the specs were too different?
-
MARS was created to beat GTX295, and it was used for this reason. It was bought and benched at hwbot as GTX295. If Nvidia had make GTX299 with full specs (240sp and 512bit), MARS would be there even with GTX295 Device ID. But Nvidia didn't make it. MARS is special limited card that not 'own' rankings and it doesn't have a reference analog, is it really need to make it even more special and useless at hwbot?
I think you agree with the fact that difference in overclocked performance between MARS and GTX295 is the same as between GTX295 and other GTX295. Let's it stay as is and no one will be hurt.
Difference between cards always existed in every model with the same GPU specs. I hope that we will not have tons of new categories just for specs difference.
The question is what differences. In this case it seems obvious: the added bus width and ROPs make the cards substantially different from any regular GTX 295. Not even ASUS recognizes it as a 295 on their website, but as a 2x GTX 285 card.
-
Just to reply on your post, regardless of Mars topic.
I wish you were right, but that´s not the way hwbot database works. I took examples 7900 and 6800 plus 9800GTX(+) because these all were handled differently. It does not prove your point but denies it. 6800 U and gt are at same same core, device ID but have different clocks, seperated, 7900GTX and GTO have same core ad core clock, different device ID but mem speed differs, different category, 9800GTX and 9800GTX+ have same core, same device ID and different clocks but are merged at same category... Same goes for dozens of other categories, plus we have some divided by mem size, some not, we have unlocked 6950 with 6970 specs at 6950 category because of device id and so on...
I think I've already mentioned this, but 9800GTX was combined as it was impossble to tell the difference by looking at a screenshot. If this was not the case there would've been two categories to this day. It wasn't handled differently at all. As for 6800 Ultra and 7900 GTO nvidia made those names, it would make no sense not to have separate entries. Luckily, Mars cards can easily be identified by looking at the ROP count and bus width, so there is no need to merge those with 295's.
-
This for example is simply wrong, mem timings matter, I thought you also benched vgas? I really doubt this. And the device ID is no software error, maybe you should google the basics before we keep on discussing at least on this. The Mars has the identical ID like the generic 295GTX
On your logic, you don´t get the point, maybe you should read again what I wrote. What is reference speed? The 7900GTO is a GTX with downclocked mem speed for example, the 6800 Ultra is simply an overclocked 6800GT if you see it this way (they share even same device id iirc), no difference at all apart from this on shaders and rops etc, we has the decision to keep 8800GTS 640 112 at same category like 96 model because it was said another tiny category makes no sense. the 2700k is unlocked but has default 100 mhz more than identical 2600k, therefore it is technically an overclocked 2600k if you want to see it this way, and there are a lot of similar examples like 9800GTX(+) on which was decided to merge or divide despite your holy reference speed being same or different, there is no pattern in these decisions. So why do we divide these and do not on 8800gt amp and card with nv spec clocks? And on others we do different? It was done because of microcode, device id, marketing, sometimes it was done on manageability etc, and a decision made, even by consent, should be reliable if the facts don´t change. We discussed unlocked 6950 to 6970 a few months ago, with the way things are handled here we will discuss this again in half a year and staff can make a task for stuff like this to be overthrown randomly regularly. Meanwhile people who relied on database rankings for years get annoyed and do what alot of people already did, they stop benching.
On the time frame, many people asked over the years to remove or merge a lot of the categories where cpus that never were sold publicly are listed, how irrelevant would it be for you if hwbot decides to do this? There are enough reasons for this, you know the discussions most likely yourself. You beg for es at amd, hunt on ebay for obscure cpus and pay big cash and someone says april´s fools, all cups for these gone after 10 years...
Well, how biig is the difference between the stock performance of a GTX 1080 with memory from manufacturer X vs Y? 5%? 10%? Sure, I agree with you in principle, but the difference must be non-negligible (which is the case if we're talking about 1-2% performance difference at best). Besides, when you look at specs for GPUs timings are never listed. Timings don't define what a GPU is, and it has never been the case.
The device ID might be the same, but it's still not a GTX 295, as per specs. It's just as little of a 295 as a 285 is not a 275. A coded string doesn't change that.
Reference speed is what nvidia put as the default. In the case of 7900GTO it would be 650/660, or something like that. 6800 Ultra is also clearly defined. But, if you accept that Mars is a 295 there is no reason to separate neither, they're much more comparable to each other than a Mars and a stock 295. And, as stated before, by the same logic a GTX 1070 is equivalent enough to be matched agains 1080ti/Titan X Pascal. If this is essentially what some people have requested in terms of merging CPU categories (released or not is irrelevant) I understand why the requests have been turned down.
I don't see nobody replying to that part of my argument, which I can understand. Device ID is not what defines a GPU, that's already agreed on (6800 Ultra vs GT), so why can't all GPUs with a certain core just be merged?
Regarding marketing, compare:
https://www.asus.com/Graphics-Cards/MARS2DI4GD3/ - says "dual GTX 285" on the right. Also no 295 in the part number.
https://www.asus.com/Graphics-Cards/ENGTX2952DI1792MD3/ - says "GTX 295" on the right, also has 295 as a part of the part number.
The only reason GTX 295 is mentioned is that it was the card to beat at the time.
I was already screwed by the UCBench decision, as well as PCMark05 globals (not to mention changes to how globals work). It's not something new to me. But both had to be done, although I hated it at the time and fought against it. I probably spent more tme running now pointless sessions with UCBench than most people with a Mars spent running it (not to mention: lost a ton more points).
-
You can easily compare the time frame, and how it's irrelevant how it affects some (or many) users.
By the same reasoning you shouldn't even differ between like a GTX1070 or a Titan X Pascal, or any Pascal GPU. They're all pascals, thus equal. At least just as equal as a Mars vs a 295. If # of ROPs and bus width is not enough, why are these cards separated? Besides, the Mars isn't marketed as a 295, even: https://www.asus.com/Graphics-Cards/MARS2DI4GD3/ - it's it's own thing. Maybe some software, by error, reads it as a 295 doesn't make it a GTX 295.
The line has been drawn at similar performance at reference speeds. Memory IC, PCB, production process, clocks etc don't change that, but bus width and # of ROPs usually do. This is one of the cases that go against the common practise above. And if this argument doesn't hold, then GTX 275 and 285 should also be merged as the difference between the two is the same as the difference between Mars and 295. Agree?
-
Nah, let me just make my own brand and release some unique cards.
BestVGA GTX 260 Unicorn RGB, 32 % faster than generic GTX 260. Limited edition, only 10 "manufactured".
You mean RGBGA GTX 260 Seizure edition:D
-
So what? The fact you can cheat proves the normal recognition to be wrong? I can do this on practically each x1 ati card and noone bothers. It only shows you run out of arguments
P.S. Yes, Knut, it was made when the 216 was released, and not 10 years later after people bought Mars relying on the ranking at 295 - good you get my point, there is hope
I certainly get your point. There was once a decision regarding UCBench as well, that setting a higher thread count was OK. Years later, BOOM, not OK. This affected many more users than the relatively few who purchased Mars for benching purposes - the fact that it shouldn't have been allowed in the first place didn't matter. That was actually even worse, as allowing the thread count to be altered was a decision based on a discussion in the crew section - I do not think anyone took notice of this problem with the Mars back then.
I must admit I thought the shader count was different, which is not the case, but having different bus widths has usually been enough to add different categories (many low end GPUs are split that way, and I guess some of us remember the 192bit and 256bit versions of GTX460, which both have the same # of shaders and are split into different categories here)
-
I can't see why tap water is not OK. Everyone has it (at least in the overclocking scene), and it's even easier to use than regular water cooling setups, which are allowed.
Mother nature isn't a "chiller".
-
The same thing goes for the 260's, but because they really aren't the same models despite having the same name, a new category was made when the 216sp edition was released.
-
So, it seems to me that the community have decided to keep these cards in the original category (if by a small margin).
In my opinion the "correct" category would be in a category of its own, BUT ONLY IF the card was released today. Since that is not the case and no one has voiced any concern for ten years, it should remain where it has always been. You can't change history!
As of right now it's a draw:p
-
Lol if you ever benched a dual GPU card, especially on LN2, you would know that it's not the same as 2 seperate cards. Especially when it comes to cooling.... Surely 1 PCIe Slot vs. 2 also makes a difference.
I'm amazed in which direction hwbot is heading here. I won't bother to bench for the rankings anymore. If we can't even have stability over these years - where is the point
You completely missed my point. I was referring to specs, only (apart from clock speeds), and did that to show that putting it with the 285 SLI cards makes no sense (as little sense as tossing it into the 295 category does, as shown per polcy of the GTX 260 192 SP vs 216 SP - if anything this Mars card is a 480sp 295).
-
My 2 cents on this as of tonight: the point in having the enthusiast league is to keep the cost down for the OC'er. At least that's what iit seems from the explanations here: Overclocking, overclocking, and much more! Like overclocking.
Perhaps one wants to remove the skill barrier that comes when you need to start insulating as well, that seems a part of it based on the Enthusiast league description. In the end it doesn't seem to be about operating temperature itself, but the ambient temperature. This is the only line in the rules about this: "Enthusiast: ambient cooling, registered >1y".
"Ambient" simply means the temperature of the surroundings, so if you put your rig outside in Sibiria at winter time, ambient will be awesome - comparable to something between single stage and dry ice. That's just how it is. You can't have a set of rules where you can run your AC unit full force in the summertime in Australia, but someone living in Eastern Europe can't even open the wndow at night (skill level and cost = 0, AC units are expensive). With this in mind I'd be surprised if hooking your water block up to the water supply or putting some ice collected outsde on top of your radiator are things that are not meant for a league like this, either - whether or not it can be called "ambient" (answer: no). Both cooling methods are almost free and require very little skill to set up, so it makes no sense to compare good use of ones surroundings to phase change units and such.
-
Go ahead, photochop them all if you like
Pretty straight-on with good lightning (thanks to my source, I don't know how to handle a camera). -
I didn't know this one wasn't listed anywhere at all, though... not a single hit on the part number on google. Funny, as I have a few somewhere in my trays
all from the same source. -
Your wish is my command. It looks very normal. Not the CA ending on this and the 165 chip
-
Rad in the window is the oldest trick in the book, not necessarily just to cool the system, but to dump the heat outside. Classifying it as chilled when all you did was dumpng the heat out of your room is a bit odd. AC is borderlne, but strictly speaking it's chilled. Not that I think it messes with the rankings if you run your AC at 16C and post your results, but per definition t has to be that way.
-
Nvidia GeForce GTX 260 192SP @ HWBOT
Nvidia GeForce GTX 260 216SP @ HWBOT
This was done for a reason, the same reason that requires the Dual 285 card to have it's own category. Calling it "GTX 285 SLI" is also not correct, as there is no merging done to GTX 295 and GTX 275 SLI, which is the same situation.
I've had 10 year old results taken down recently. It's only fair to fix mistakes when they appear to be mistakes, like calling a dual 285 card a 295.
-
9800GX2 = 9800 GTX SLI (apart from clock speed differences), GTX 295 = GTX 275 SLI, GTX 590 = GTX 580 SLI, GTX 690 = GTX 680 SLI (and so on). yet none of these have been merged. Mars dual 285 should have it's own category, by the same argument that allows splitting of the ones above. Memory size does not matter, however.
For the record: I own none of the Ares/Mars cards.
-
Overclocking, overclocking, and much more! Like overclocking.
1: this is for socket 940, despite whatever CPUZ says.
2: these are retails. No ES markings, just unreleased stuff.
-
I've been scammed for $500 items twice. From my experience your protection as a buyer isn't worth jack shit. Seller has shitty feedback with similar complaints about empty boxes arriving? Noooo, must be the guy with 1500 feedback and no negatives who is the bad guy...
#ebaylogic
-
If it's unclear, it doesn't hurt to have a list somewhere. I guess everyone prefers to avoid alot of b*tching later
-
Picture attached... this one was taken by the guy who sold it to me almost 10 years ago.
-
Wasn't this just a CPU that is misprogrammed and detected as FX-24?
If so, it doesn't really deserve its own category as a simple BIOS modification on those platforms can give you any desired output (but won't change CPU specs, obviously).
Impossible for me to know if it is misprogrammed or not. It clocks like shit, so might be something else than an FX57... Different specs => different category, if you ask me.
Personally I don't think Q3FE shold be mixed with 980's, either - different stock speeds
I guess that applies to Q3QP as well, to name another example.
Poll: ASUS Mars Dual GTX285, where should it be ranked?
in Live
Posted
In this case the decisions are pretty consistent. There was never a practise to separate cards by brand or factory OC, or memory timngs. However, bus width seems to always require new categories to be made. If you disagree, find me some exceptions. Same thing goes for # of ROPs. If anything, the trend is to separate more rather than less.
As you said yourself: you can flash the BIOS with different clocks, hence you can't tell if a card is a (+) or not when the + sign is nowhere to be found in GPUz. However, a 6800 GT is a 6800 GT in GPUz and an 6800 Ultra is a 6800 Ultra.