Next NV High-end
- Thread starter Geo
- Start date
By the way, Ailuros, I don't think only downclocking the memory would be a good idea to really measure the performance.
The proposed performance issue with the downclocked X1800 was that the memory latency for its memory was higher than that of the 7800's memory. Since the chip's caches and drivers are likely built with the latency of the memory in mind, it would make more sense to downclock both the core and the memory of the X1800 part until the nVidia part could be brought to a level where it could compete.
The nVidia part would still have lower latency in this scenario, but it is doubtful that it would be of significant help, because the ATI part would still be within the latency it was designed to handle.
The proposed performance issue with the downclocked X1800 was that the memory latency for its memory was higher than that of the 7800's memory. Since the chip's caches and drivers are likely built with the latency of the memory in mind, it would make more sense to downclock both the core and the memory of the X1800 part until the nVidia part could be brought to a level where it could compete.
The nVidia part would still have lower latency in this scenario, but it is doubtful that it would be of significant help, because the ATI part would still be within the latency it was designed to handle.
I'm not talking about comparisons between products; I'm talking about a simple experiment on only a X1800XT.
Granted latency is a valid point in such an experiment, but I'm just curious - despite the latency - what difference in performance reducing it's maximum bandwidth makes. I'm curious to see if a let's say 25% lower memory frequency also leads to signficant drop in overall performance.
The reason I'm asking should be obvious considering my former posts; increasing by 15% memory frequency on the G70 increases performance about 5% or so in most cases. Albeit an obvious flawed experiment, I'd like to see how a XT reacts when you take off an X-percentage of it's bandwidth.
Granted latency is a valid point in such an experiment, but I'm just curious - despite the latency - what difference in performance reducing it's maximum bandwidth makes. I'm curious to see if a let's say 25% lower memory frequency also leads to signficant drop in overall performance.
The reason I'm asking should be obvious considering my former posts; increasing by 15% memory frequency on the G70 increases performance about 5% or so in most cases. Albeit an obvious flawed experiment, I'd like to see how a XT reacts when you take off an X-percentage of it's bandwidth.
we didn't have cards for 7800 GT until a few days before (hence why everyone seemed to miss the 'shimmer' problems) - I think this is going to be a little better organised than that, but not quite as organised as GTX launch. I would be surprised if there wasn't volume at launch.
Don't forget, X1k launch, the euro hacks didn't have cards til the weekend before a wednesday launch.
EDIT: btw, I don't believe the date is correct but it could be somewhere near that date. I think that OcUK might have stock arrival dates mixed up with launch dates.
Don't forget, X1k launch, the euro hacks didn't have cards til the weekend before a wednesday launch.
EDIT: btw, I don't believe the date is correct but it could be somewhere near that date. I think that OcUK might have stock arrival dates mixed up with launch dates.
Last edited by a moderator:
BrynS said:
wish people would stop cross quoting this guy, he has no idea what hes talking about. All he does is hype Nvidia products and currently his score for fact vs false is worse then the inq. Guess we can add this to the "far inferior X1800 so buy Nvidia (which btw are on pre-oder from their site for about $160 over MSRP)" and the "GTX SLI 22000 3dmark05".
Seeing his comments, especially at this board, tend to annoy me a bit.
bigz said:
Yes, but X1k launch was not a "volume" launch of 3/4 of the parts involved, including the top of the line that everyone was intensely interested in seeing benched. So it still seems a little curious to me that they'd rush you guys on reviews when they must have retail packages in goodly quantity for the fabled "hard launch".
Totally, I can't say I was best pleased with the GT launch - I didn't sleep more than 6 hours in 72 - takes me ~8-10hrs to get through my tests with each vid card configuration and I had two different vendor cards clocked at different speeds with both SLI and single card configs to run through. I don't fancy doing that again any time soon. I was out of action from the Thursday til the Monday catching up on sleep. :|
Jawed
Legend
It doesn't seem that difficult to crush X1600XT 
It's a low mainstream card masquerading as a top mainstream card. BF2 is about the only exception.
The X1800XL should be MSRP'd at X1600XT's MSRP.
Looks like ATI is set for another round of X700XT->X800GT->X800GTO humble-pie. Oh dear. The only escape is if there's about 50% extra performance lying waiting in X1600XT for the drivers to release it.
Jawed
It's a low mainstream card masquerading as a top mainstream card. BF2 is about the only exception.The X1800XL should be MSRP'd at X1600XT's MSRP.
Looks like ATI is set for another round of X700XT->X800GT->X800GTO humble-pie. Oh dear. The only escape is if there's about 50% extra performance lying waiting in X1600XT for the drivers to release it.
Jawed
Given the featureset of the X1k series, you should expect a lower performance/price ratio than the NV4x series anyway. Otherwise you wouldn't have the fun of conversations like:
I prefer image quality, therefore I'm going with this.
But higher resolution is image quality, and I get the benefit of the choice of better framerates too!
But my card is more future-proof.
No, because the higher performance of my card means that new games will still run better.
No way in hell, man, the new features of my card will make it way faster in new games!
But mine's better in Linux!
I prefer image quality, therefore I'm going with this.
But higher resolution is image quality, and I get the benefit of the choice of better framerates too!
But my card is more future-proof.
No, because the higher performance of my card means that new games will still run better.
No way in hell, man, the new features of my card will make it way faster in new games!
But mine's better in Linux!
As a salesguy he knows these things earlier than we. We shouldn't we trust him? Because he was complainig in the past about ATi's availability problems? He was right about that. Please elaborate about his "bad record".serenity said:Yeah, I'm disappointed the Tim gave him as a "reliable source" in their article. Nvidia might have tightened up things a bit around bit-tech's insider but they shouldnt scoop up forum posts by a salesguy. (especially a notorious one with a bad record
HaLDoL said:
If you hang around there during the release of g70/r520 posts from gibbo are structured like the following:
"omfg! g70sli 22000 3dmarks! I heard it from a reliable source! here is the link for the pre-order."
"no way is r520 gonna score anything over 9000 3dmarks! I have a reliable source! here is the link to order the 7800 GT(x)"
and the most LAUGHABLE statement (straight from their front page!)
I'm just missing the OMFG WE'RE SO GOOD part there..
My article was reasonable speculation at the time, but I've found out that it is out by a larger margin than I initially thought (well after my initial thoughts that Gibbo's statement wasn't 100% correct). I don't think OcUK liked being 'quoted' as giving out release dates - that'd explain why they removed the thread. There is something coming on the 7th though.
I can't tell you anything about clock speeds on the 512MB card at the moment, unfortunately.
But I will say that November should be an interesting month for consumers and I can't wait to see how the two flagship cards stack up (and compete) on both price and performance. I also have to wonder whether we'll see another Platinum Edition from ATI, as I don't think this 'war' is over just yet. I still believe both companies are holding their cards very close to their chest and there's more to come from both sides.
I can't tell you anything about clock speeds on the 512MB card at the moment, unfortunately.
But I will say that November should be an interesting month for consumers and I can't wait to see how the two flagship cards stack up (and compete) on both price and performance. I also have to wonder whether we'll see another Platinum Edition from ATI, as I don't think this 'war' is over just yet. I still believe both companies are holding their cards very close to their chest and there's more to come from both sides.
