Oct 19th AMD (ATi) and Nvidia news

[edplayer]

You haven't been paying attention to Anand's reviews, have you?


But then .. i don't care to convince you or anyone .. you will soon see for yourself.

Oh yeah, your lack of detail was totally convincing. But you are right about waiting. We will all know in time.

 

[CitanUzuki]

i am saying that with Nvidia's more mature process they can now overclock the GF100 core to over 800MHz and use a three slot cooler - something they could not dream of 6 months ago in a sub $500 video card that is not watercooled

With their more mature process they may very well be able to increase the clocks at lower volts with better thermal characteristics. But how does the card you reviewed give you the impression that the above has anything to do with how cool that particular card runs? Or that that card is somehow proof of the above?

I can hit 800Mhz with my launch 480 at stock volts, put a massive cooler and a back plate I could probably go higher all while having much lower temps than with the stock cooler.

Take that card and slap a stock cooler on it and then maybe that would say something.

 

[apoppin]

With their more mature process they may very well be able to increase the clocks at lower volts with better thermal characteristics. But how does the card you reviewed give you the impression that the above has anything to do with how cool that particular card runs? Or that that card is somehow proof of the above?

I can hit 800Mhz with my launch 480 at stock volts, put a massive cooler and a back plate I could probably go higher all while having much lower temps than with the stock cooler.

Take that card and slap a stock cooler on it and then maybe that would say something.

Well, good luck to you. Maybe you should try it. It is a world of difference between my Nvidia reference GTX 480 and my Galaxy GTX 480 SOC.

The Fermi process has matured just as Cypress did back in April when AMD launched their Overclocked HD 5870s.

Oh yeah, your lack of detail was totally convincing. But you are right about waiting. We will all know in time.

Seriously. Go back and check each one of Anand's Fermi product reviews. In each one is a little gem describing how that particular product is "missing" certain areas that can be enabled ... well, anytime

 

[extra]

No need. The 260 release candidates already gave a nice performance boost to Fermi owners.

Haven't you been paying attention to (especially Anand) reviews that talk about disabled parts. We haven't seen GTX 485/475/GTS 455 nor GT 435. All of them have room for solid clock bumps as the process have matured.

You have to realize that Nvidia planned for HD 6000 with Fermi variants. it is their plan. Just because they were late, AMD was late .. and neither company is sleeping.

Remember what i say when i predict it is going to be a very interesting upcoming six months - much more so than the last year.

The only people who will be unhappy will be the ones who want either company killed off by the other (product-wise)

FYI there is no possible "435" made from the 430, it's already totally enabled.

 

[tviceman]

NV cannot release jacksh!t for at least another 6-12 months, at least nothing competitive in the top bracket.

Fanboy post of the day!


Calling out your fellow forum members is not acceptable. No personal attacks.

Moderator Idontcare

 

[Madcatatlas]

Fanboy post of the day!

Is this a member callout? or just treadcrapping?



Apoppin, are there any gems to be found in your own reviews? or do you just like the cards you get? Are they yours for "life", after the review? Or do you send them off to some other reviews site?

None of the Anandtech Fermi card reviews have mentioned a GTX 455, 475 or a GTX 495. GTX 485 maybe, but youll have to link it or redact it, seeing as your putting words into a fellow reviewers mouth. Do you want us adressing Ryan Smith on things that are not in his reviews?

Ill belive in a good Fermi product when i see one. So far its only been the GTX460 and guess what, i got one.


On the whole 40nm process, i belive neither AMD nor Nvidia will make it a truly second gen 40 product, but rather a 1,5 product, meaning the increases in performance wont be that high.

Unlike Nvidia however, AMD will be able to refine the process even further without adding much cost on production. The result will be a better 40nm product which passes their current lineup by enough of a margin to be considered sound in a business perspective, and it SHOULD also beat the competitions CURRENT products (as one really cant predict just what the competition will do 7 months from now.)
All of this, SHOULD be priced so the customer actually values a 6xxx card over a 5xxx card. And i have no doubt AMD will make this happen.

Hell, the reported tessalation improvements in the HD6xxx series alone should prove better than anything Nvidia can do to Fermi in the next 7 months.

Yes, i wouldnt expect another Fermi product before april next year.

 

[apoppin]

Apoppin, are there any gems to be found in your own reviews? or do you just like the cards you get? Are they yours for "life", after the review? Or do you send them off to some other reviews site?

None of the Anandtech Fermi card reviews have mentioned a GTX 455, 475 or a GTX 495. GTX 485 maybe, but youll have to link it or redact it, seeing as your putting words into a fellow reviewers mouth. Do you want us adressing Ryan Smith on things that are not in his reviews?

Ill belive in a good Fermi product when i see one. So far its only been the GTX460 and guess what, i got one.


On the whole 40nm process, i belive neither AMD nor Nvidia will make it a truly second gen 40 product, but rather a 1,5 product, meaning the increases in performance wont be that high.

Unlike Nvidia however, AMD will be able to refine the process even further without adding much cost on production. The result will be a better 40nm product which passes their current lineup by enough of a margin to be considered sound in a business perspective, and it SHOULD also beat the competitions CURRENT products (as one really cant predict just what the competition will do 7 months from now.)
All of this, SHOULD be priced so the customer actually values a 6xxx card over a 5xxx card. And i have no doubt AMD will make this happen.

Hell, the reported tessalation improvements in the HD6xxx series alone should prove better than anything Nvidia can do to Fermi in the next 7 months.

Yes, i wouldnt expect another Fermi product before april next year.

Yes, of course there are hidden gems in my own reviews. My tech site is a media partner with both Nvidia and AMD and we get video cards and hardware to review. Unless there are special arrangements, we keep the cards that we review. In my own case, i keep them until they are legacy or not needed for future reviews. And my opinions are my own. i am putting no words in anyone's mouth.

First of all, Anand said this in his GT 430 review:

unlike all of NVIDIA’s other desktop launches which had GPUs with disabled functional units, the GT 430 uses a fully enabled GF108 GPU.

So now we know that every other GPU has disabled functional units. And we know now that Fermi's yield has improved and the process has matured. So here is a gem from Anand:

For the memory NVIDIA is using DDR3, ... and NVIDIA is already talking about working with partners on a GDDR5 version of the card in the future.... In practice we usually find that 512MB of GDDR5 is better than 1GB of DDR3 in most cases.
..
For GF108 there are 4 ROPs in a single block, and that block is attached to the two 64bit memory controllers that make up the GPU’s 128bit memory bus. This is a surprising change since it would have required additional work on their part to make the necessary changes to their ROPs – sticking to the standard ratio would have been much easier

Do you see any room for improvement for GT 435?

Historically, when the process matures, there is a "b" .. just like with GT200>GT200b
.. and also, *historically* with Nvidia there is a xx5 and a "+" quite often
. . GTX 260 > GTX 260+ > GTX 275 ... etc.

Unless you *really* think Nvidia is going to play dead and roll over?

 

[ugaboga232]

Even with a 512 Shader product with a heavy overclock (something that was quite difficult on the 40 nm size before but I believe nvidia could produce if they want to) would still probably lose to the top end Cayman or maybe tie it at a higher pricepoint. I don't think adding 32 shaders or heavy overclocks will allow nvidia to compete with ati's coming top end as while there are many disabled parts, they aren't that numerous or performance enhancing. I kind of agree that nvidia needs a true refresh and not just some tweaks. I doubt a full gf104 will compete either but possibly a bigger chip based on the gf104 could compete. Also AMD supposedly has much better dx11 performance which will be interesting to see. I see the ball in nvidia's court and since their chip is pretty much as big as it can get on 40nm, it will take a good refresh which is at least a couple months to achieve performance and price parity with AMD.

 

[Kenmitch]

Do you see any room for improvement for GT 435?

Historically, when the process matures, there is a "b" .. just like with GT200>GT200b
.. and also, *historically* with Nvidia there is a xx5 and a "+" quite often
. . GTX 260 > GTX 260+ > GTX 275 ... etc.

Unless you *really* think Nvidia is going to play dead and roll over?

So your kinda hypotheticaly speaking like somebody whom may or may not be under an NDA more or less.

 

[tweakboy]

I DONT THINK SO. nVidia is not releasing their new generation of cards 5xx until late late next year.

This is a announcment for the 430GT and price drops.

AMD on the other hand is releasing new technology in Nov Dec, the 6xxx series.

nVidia is 1 year behind AMD, who would have thunk it.

 

[apoppin]

So your kinda hypotheticaly speaking like somebody whom may or may not be under an NDA more or less.

Well, in this particular case i am speaking from a historical perspective and from my own observations. The NDA that i currently am under is not with Nvidia.

Look at the current situation and tell me how it would have been different if Nvidia had launched GF100 and GTX 480 when AMD launched Cypress.

What what have happened if Nvidia was not "late" with Fermi? AMD would have launched Cypress and both would have brought out their full product line at the same time.

Only it didn't happen in a timely fashion for Nvidia? Do you think they have completely "lost" six months? What were they planning if they were on time (six months ago with Fermi) when AMD would have released HD 6000 series anyway?

To answer those questions you have to realize that Nvidia does very likely have an answer and an upgrade path with Fermi on 40 nm. Their *usual* bump and refresh at the very least. Will it be enough? Well, we shall all know very soon.

 

[Sickamore]

Nvidia has to hit their heads hard so they can come hard. I have ati cards something about me tells me just believe in nvidia. What i also notice is that the imac and mac pro only have ati cards. The laptops have nvidia chips. I see nvidia becoming more of a mobile gpu company. Nvidia knows where their money is coming from. I believe in them.

 

[Will Robinson]

Cheerleading about totally unannounced,non existent hardware just to give the impression that NVDA has a competitive answer to the new AMD 6 series is lame.
What's next?
An overclocked GTX480 "comparison test" against AMD's new sub $200 part?
Maybe they fall all over that at AlienBabyTech....


Your persistent thread-crapping and hurling of insults are not acceptable.

Moderator Idontcare

 

[solofly]

Looks like I won't be able to sleep until 19th...

 

[CitanUzuki]

Well, good luck to you. Maybe you should try it. It is a world of difference between my Nvidia reference GTX 480 and my Galaxy GTX 480 SOC.

The Fermi process has matured just as Cypress did back in April when AMD launched their Overclocked HD 5870s.

In what way is it a world of difference between your reference card and the Galaxy? The vid is lower, it consumes less power, what? It is painfully obvious that the card runs much cooler with much better cooling. I had 4890 that was hot as hell, and when I put a t-rad2gtx on it, it was like night and day. So what?

Im certainly not saying the process has not matured, but how you can come to that conclusion based on the how cool the galaxy runs makes no sense to me.

 

[apoppin]

Cheerleading about totally unannounced,non existent hardware just to give the impression that NVDA has a competitive answer to the new AMD 6 series is lame.
What's next?
An overclocked GTX480 "comparison test" against AMD's new sub $200 part?
Maybe they fall all over that at AlienBabyTech....

Saying that Nvidia has no answer is beyond lame

What IS next?

i already did the overclocked GTX 480 "comparison test" against AMD's overclocked HD 5870 .. maybe 5 times already, including the last one a month ago with Galaxy GTX 480 SOC vs PowerColor HD 5870 PCS+ and each time the GeForce gets further ahead of the Radeon
-so, you're right .. it's time for a new comparison

 

[TemjinGold]

apoppin: Yes, I really DO think nVidia has no answer. Yes they are a big company and all that but that doesn't mean it isn't possible for them to be caught without an answer (remember that time when Cypress was out and nVidia's answer was to EOL 2xx?) What you are posting sounds more like desperate hope grasping at straws than anything else. If they end up redeeming themselves, fine, good on them. But until then, it is MORE logical to believe that they are screwed. You can theorycraft and speculate but that doesn't do diddly. Historically, plenty of "things" have been hinted at and whatnot, with a large chunk of those things being false. I mean heck, Y2K and 2012 have more "proof" than nVidia...

 

[Spike]

All this nVidia hate is kinda strange. Yes, I prefer AMD/ATI cards but I'll buy whatever is the best price/performance and for the last few years nVidia has almost always come through with at least one winner. Look at the 460, I just helped a friend build a gaming PC and put a 768mb 460 in there since nothing AMD has can touch it for that price.

While I eagerly await the 6xxx series as I'm building a new pc soon I still expect something fast from nVidia. At the very least I not only exepct I WANT it so AMD keeps their prices in check. It should be quite a fun few months in the vid card arena.

 

[Kenmitch]

Well, in this particular case i am speaking from a historical perspective and from my own observations. The NDA that i currently am under is not with Nvidia.

I was thinking your under AMD's by an earlier post a week or two ago. Just wasn't shure about nvidia.

Only it didn't happen in a timely fashion for Nvidia? Do you think they have completely "lost" six months? What were they planning if they were on time (six months ago with Fermi) when AMD would have released HD 6000 series anyway?

Not sure what to think. Seems kinda strange to me that nvidia didn't even try to trump the 5970.

 

[notty22]

All this nVidia hate is kinda strange. Yes, I prefer AMD/ATI cards but I'll buy whatever is the best price/performance and for the last few years nVidia has almost always come through with at least one winner. Look at the 460, I just helped a friend build a gaming PC and put a 768mb 460 in there since nothing AMD has can touch it for that price.

While I eagerly await the 6xxx series as I'm building a new pc soon I still expect something fast from nVidia. At the very least I not only exepct I WANT it so AMD keeps their prices in check. It should be quite a fun few months in the vid card arena.

I agree, I also think Nvidia has more to bring to the landscape. I think there will be dual gpu cards as well. Not sure about the hostility from suggesting Nvidia is not dead against a product from AMD we don't even have the price/performance numbers from .

 

[HurleyBird]

Apoppin: I hope you're right. For the sake of the consumer I really do. But I think you're waaayyy out to lunch on this one. Sure, Jensen talked about a "mid-life-kicker", but JHH is about as trustworthy as a used cars salesman. Even taking JHH at face value, the possibility that makes the most sense is that the "mid-life-kicker" is simply higher clocked fully enabled versions of existing products. With GT430 we finally saw how a fully enabled Fermi derivative stacked up a Radeon with a similarly sized die (actually the die in HD 5670 is about 10% smaller), and it got completely obliterated. In terms of architectural performance, AMD is *already* at least a half generation ahead of Nvidia, and now they are going to extend that lead in a few weeks. I don't think Nvidia is going to be able to make a die big enough to hide their architectural deficiencies this time. I think the best that we can hope for is for them to not simply withdrawl their uncompetitive products like they did with GT200 vs. Cypress, and instead focus on dropping prices through the floor to compete. Basically same strategy AMD is using with their CPU division. That could end up being very good for the consumer, but if they do withdraw there will be no end to the price gouging.

 

[SlowSpyder]

Well, in this particular case i am speaking from a historical perspective and from my own observations. The NDA that i currently am under is not with Nvidia.

Look at the current situation and tell me how it would have been different if Nvidia had launched GF100 and GTX 480 when AMD launched Cypress.

What what have happened if Nvidia was not "late" with Fermi? AMD would have launched Cypress and both would have brought out their full product line at the same time.

Only it didn't happen in a timely fashion for Nvidia? Do you think they have completely "lost" six months? What were they planning if they were on time (six months ago with Fermi) when AMD would have released HD 6000 series anyway?

To answer those questions you have to realize that Nvidia does very likely have an answer and an upgrade path with Fermi on 40 nm. Their *usual* bump and refresh at the very least. Will it be enough? Well, we shall all know very soon.

Here is why I think some of us have some doubts regarding Nvidia doing much until 28nm.


When was the last time Nvidia did a 'midlife kicker' on the high end on the same process? Maybe the 5800 -> 5900? (I think both of those were 130nm?). Nvidia has brought out a few parts on the same process that could be considered mid life kickers, like the 8800GT, the 7950, but they were not at the highest end of their single GPU parts.

We are always told that AMD and Nvidia have people working on multiple projects, so Fermi's delays shouldn't necessarily effect Fermi 2. AMD's 5xxx shouldn't necessarily affect the 6xxx. But we heard that AMD's 6xxx taped out months ago, and now it's almost to retail. We haven't heard anything about Fermi 1.5 / 2 taping out.

I have to wonder if the 32nm cancellation hurt Nvidia more than AMD. Nvidia's parts are already pretty big and power hungry. AMD had room to build a bigger GPU if they wanted. They also had room to use more power. Nvidia doesn't have those luxuries.

Some rumors say that AMD's parts are roughly the same size as their current parts, just with a tweaked/new architecture. But the difference here is, AMD's architecture is pretty long in the tooth now, the 5xxx is based off of the 4xxx which is based off of the 3xxx which is a tweaked 2xxx part. Fermi is a Nvidia's brand new whizbang architecture... do we really think they are going to be building a new architecure already? The 8800 was new, and became a rebadge in the 9800 and GTS250, and the GTX2xx are really built from those prior parts. So do we really think that Nvidia is going to change their architecture after one go around, that they planned for Fermi to just be a single generation?

So they can't make it bigger and more power hungry, and it would be very unlikely that they would change the architecture. At least that's how my not-in-the-industry-eyes see the situation. While I'm sure they aren't sitting still, I am just saying I wouldn't be shocked if they didn't have anything worthwhile to combat the 6xxx cards.

 

[apoppin]

In what way is it a world of difference between your reference card and the Galaxy? The vid is lower, it consumes less power, what? It is painfully obvious that the card runs much cooler with much better cooling. I had 4890 that was hot as hell, and when I put a t-rad2gtx on it, it was like night and day. So what?

Im certainly not saying the process has not matured, but how you can come to that conclusion based on the how cool the galaxy runs makes no sense to me.

i am not the only reviewer who has noticed this:
http://benchmarkreviews.com/index.php?option=com_content&task=view&id=607&Itemid=72

Benchmark Reviews tests the GF100 Fermi GPU six months after its launch, and is reminded that refined construction process and mature firmware can make a difference.

and

apoppin: Yes, I really DO think nVidia has no answer. Yes they are a big company and all that but that doesn't mean it isn't possible for them to be caught without an answer (remember that time when Cypress was out and nVidia's answer was to EOL 2xx?) What you are posting sounds more like desperate hope grasping at straws than anything else. If they end up redeeming themselves, fine, good on them. But until then, it is MORE logical to believe that they are screwed. You can theorycraft and speculate but that doesn't do diddly. Historically, plenty of "things" have been hinted at and whatnot, with a large chunk of those things being false. I mean heck, Y2K and 2012 have more "proof" than nVidia...

Nvidia's "answer" to Cypress has always been FERMI .. Fermi was delayed for 6 months.

If Fermi was ON TIME, what would Nvidia be planning when HD 6000 series launched as they knew it would after Cypress?
Why would the answer that they planned then be not valid now?

and

When was the last time Nvidia did a 'midlife kicker' on the high end on the same process?

it depends how you define midlife kicker and whether it needs to be on the same process (very technically you could call HD 6000 a midlife kicker - on the same process - and you could call HD 4890 a midlife kicker)

GT200>GT200b

GTX 260 and 280 to GTX 275 and 285 was a midlife kicker.

it also depends on "need" .. it depends on what we see next from AMD Graphics

it is "AMD's move"

 

[SlowSpyder]

GTX260 and 280 to GTX260b, GTX275, and GTX285 were not on the same process though. When was the last time they had a midlife kicker part on the same process? When was the last time they had one on their ultra highend?

The 4890 was definitely a mid life kicker type of part, it was just a slightly tweaked, though different, GPU. I consider the 6000 pretty much new with borrowed parts of the 5000. But it's not completely new, and it's not the same old thing. You can consider it a mid life kicker if you want. But it's not AMD I'm asking about. I want to know when Nvidia has done so on the same process technology.

I hope you don't throw out the rest of my post. The way I see it, Nvidia can't go bigger, AMD can. Nvidia can't go more power hungry, AMD can (on single GPU's). Nvidia is very unlikely to bring out a new architecture from their current parts seeing as Fermi IS their new architecture, the 5000 was just an evolutionary part with it's beginnings in the Radeon 2xxx parts.

 

[Kenmitch]

Nvidia's "answer" to Cypress has always been FERMI .. Fermi was delayed for 6 months.

If Fermi was ON TIME, what would Nvidia be planning when HD 6000 series launched as they knew it would after Cypress?
Why would the answer that they planned then be not valid now?

Fermi and 40nm don't seem to go together too good on the high end. Most likely what nvidia had planned as the answer to the 6xxx was also designed for the 32nm node if not the 28nm node. Seems logical to me at least. I think the 40nm crown is gonna go to AMD when it's all said and done as they got alot of wiggle room compared to nvidia.