Bigger hardware flop

[lost]

So in your opinions, what was the bigger flop, the NV3x architecture or the new prescott core. Granted, both have their benefits over the previous architectures, but neither lived up to the marketing and both were delayed.

In my opinion, the nv3x architecture dissapoints more than the prescott. At least the prescott offers equal performance to the existing p4 and is competitive with amd; however, the nv3x offers superior dx8 speed while sacrificing dx9 speed to ATI. I also feel like the nv3x was over marketed whereas Intel just likes giving numbers that do not show real life performance figures.

So what do you think? Maybe the volari deserves to be in the list

 

[Tim Murray]

my god, people are painting NV3x as a bigger failure than Volari? it's louder than the NV3x, no AF, only SSAA, incredibly slow, incredibly loud, incredibly hot, insane power requirements, expensive if you can even find it. oh yeah, it also came out a year later.

NV30 failed because A. it was hyped so ridiculously much and B. the R300 spanked it in every way. the Volari can't even make a half-assed excuse like that--it just sucked.

 

[Clashman]

The difference here being that most of us expected it to suck, or at least were VERY guarded and cautiously optimistic about it, being new technology resultant from the merger of two companies whose past history in the 3D world was shakey at best. NV30, on the other hand, was expected to succeed by large numbers of the enthusiast crowd, and so thus represents the larger failure in my book.

 

[Tim Murray]

perspective, I guess. I expect all new cards to suck equally so I'm happy when they don't...

 

[cthellis42]

Volari's definitely the bigger flop, which which chip was the biggest in proportion to its hype and expectations?

 

[Guden Oden]

Something's surely WRONG with people today when they talk about a product "failing" when it's not even on sale yet.

Prescott's ONLY JUST been released, it's not available in channels yet, and what hype? Haven't heard a peep from Intel, other than them saying the core's designed to scale up towards the 5GHz region, which it'll undoubtedly do, considering they managed to take the P6 core from 120MHz up to 1333...

Hype? What hype?

And NV3x a failure? I don't THINK so. Sure, performance's less than what just about everybody expected of Nvidia, but failure? Man, that's a strong word for a product that's selling by the truckload.

 

[vnet]

Volari's definitely the bigger flop, which which chip was the biggest in proportion to its hype and expectations?

Hype? What hype? I don't see it on sale anywhere in North America so ho can there be hype. Do you see any website hyping it ? Most websites don't EVEN mention it as they don't even know it exists. You call that hype ?

NV30 had a much bigger hype : everyone announced it as revolutionary and it was expected to be another success. The engineers had no excuse as they had a solid experience / track record (Geforce series) and solid drivers in the past. It also was the result of the merger of two of the greatest video companies IMO (3dfx and nVidia) and so thus represents the largest failure in my book.

Volari on the other hand - XGI has very little experience compared to ATI and NV, neither SIS nor Trident had anything groundbreaking before, nobody said it would be a R300 killer but somehow everyone had high expectance of it and decided that if it couldn't kill the 9800XT card (which costs over 140 euros more than Volari over in Europe but never mind), then it had to be a flop.
Rome was not build in one day, and neither NV nor ATI reached their current position in one single product generation so it's a bit premature to bash XGI.

 

[Joe DeFuria]

Pretty odd poll...I voted for Volari.

Now, if you had said NV30, (FX 5800) instead of the NV3x, then I may have gone with NV30. But while NV3x cores in general are not good enough" to hold on to market share, it's certainly not a flop.

As for Intel...I think it's pretty damn remarkable that that they extended the pipeline to the extent they did, and managed to basically retain the same clock-for-clock performance. Intel should have a fairly "easy" time scaling this chip.

As for me, I was waiting for the Prescott to be released before deciding which P4 to buy...and it looks like I'll be getting a 3.2 Nothwood. (Now for about $260 street.) Mostly because even though clock-for-clock the Pescott is fine WRT performance, Northwood power consumption is considerably less than Prescott. Less heat=quieter system, and I'm shooting for as quiet a system as I can get.

 

[Guden Oden]

Joe,

Surely Prescott *can't* consume more power at the same clockspeed, can it? That would sound totally nuts!

 

[Joe DeFuria]

Joe,

Surely Prescott *can't* consume more power at the same clockspeed, can it? That would sound totally nuts!

Not sure what you mean. (Are you being sarcastic or not?) It wouldn't be totally nuts either way...if Prescott consumed more or less power than Northwood.

On one hand, Prescott is 90nm ws. Northwood 130nm. So for power consumption, that works in favor of Prescott. On the other hand, Prescott has many more transistors (due largely to increased cache, but also to more stages, more instructions, etc.) and this works against Prescott.

In the end, Prescott consumes more power (generates more heat) at the same clock-speed as Northwood, and this is putting it mildly:

http://forums.silentpcreview.com/viewtopic.php?t=10021

 

[Guden Oden]

Not sure what you mean. (Are you being sarcastic or not?)

No, not sarcastic, just mindboggled.

On the other hand, Prescott has many more transistors (due largely to increased cache, but also to more stages, more instructions, etc.) and this works against Prescott.

Most of those transistors are SRAM, and those tend to not draw very much power (well, in comparison anyway, hehe). Prescott also runs at a lower voltage compared to northwood, which tends to have a major impact on power consumption...

In the end, Prescott consumes more power (generates more heat) at the same clock-speed as Northwood

Ugh! That was rather unexpected! Well, to me at least. I assumed the horror-stories of 100+ watts prescotts were for clock speeds greater than current top-of-the-line P4s.

 

[arjan de lumens]

Surely Prescott *can't* consume more power at the same clockspeed, can it? That would sound totally nuts!

2 reasons

• It has a LOT more transistors; according to Aceshardware, less that half of the transistor count increase can be explained by the larger caches (cache = about 6 transistors per bit).
• Leakage currents (per transistor!) tend to increase dramatically as transistors are made smaller. This effect was substantial already at 0.13u (IIRC, about 10W/cm2 at normal logic densities for Intel's processes) and even worse now at 0.09u (rumors indicate ~60W leakage for Prescott, haven't verified them yet).

 

[Joe DeFuria]

• It has a LOT more transistors; according to Aceshardware, less that half of the transistor count increase can be explained by the larger caches (cache = about 6 transistors per bit).
• Leakage currents (per transistor!) tend to increase dramatically as transistors are made smaller. This effect was substantial already at 0.13u (IIRC, about 10W/cm2 at normal logic densities for Intel's processes) and even worse now at 0.09u (rumors indicate ~60W leakage for Prescott, haven't verified them yet).

Wow...that's mucho leakage.

Hopefully, Intel can get it a bit more under control when they move to SOI, though that won't occur until they move to 65nm IIRC.

 

[pascal]

For me Prescott is the biggest flop ever. Using a more advanced .09 micron process but no real advantage at the same speed (mhz), no lower power at the same speed

I remenber how good were the tualatins compared to the coppermines. I had a P3-S 1.13Ghz (512kb cache), it was very cool, quiet, fast with all my applications, almost on par with Athlons MP 1.2Ghz

My daughter has a 1.2Ghz Celeron and it is really good.

I wish Pentium-M were available for desktops.

 

[cthellis42]

It's kinda what the P4's were like at launch, except to a lesser degree with the Prescott. It will have more advantages from it's ability to shoot higher, so when it DOES get higher--there ya go. Right now it's just not a useful option, and certainly no Second Coming or anything... The competition has been ramping up their speeds in the meanwhile as well, so they have ground to recover. But it's certainly not an enormous failure.

 

[mczak]

It's kinda what the P4's were like at launch, except to a lesser degree with the Prescott. It will have more advantages from it's ability to shoot higher, so when it DOES get higher--there ya go.

But intel really needs to do something about the power consumption, otherwise it's not going to scale well - some enthusiasts may not bother that the current stepping prescott needs something like 150W to reach 4Ghz, but I think the OEMs would complain a lot.
I am really surprised by the high voltage these 90nm chips need (intel says 1.25V-1.4V, not all cpus require the same voltage, but those I've seen in reviews all seemed to use 1.375V). 1.375V is really much more than I'd expected, it's a bit too close to the 130nm Northwood ~1.5V IMHO. Looks like something is not quite right with intel's 90nm process technology yet. It should get better a bit (intel promised future 3.2Ghz Prescotts will run on the now-not-Prescott-compatible motherboards). Looks like everyone is having problems with 90nm - AMD delayed it, intel's process needs too much voltage/current despite low-k/strained silicon, and the chip foundries like TSMC also don't seem to be quite ready. Remarkably though, IBM seems to got it working well (the 2Ghz PPC970 90nm G5 used in the Mac Xserve only consumes around 25W, down from the already quite low 50W the 2Ghz 130nm G5). Though I've no idea what the manufacturing costs at IBM are.
Unless intel can iron out the power/heat issues, the revamped prescott core is a complete waste of engeneering money - it would have been much cheaper to just produce a slightly enhanced (SSE3, larger L2 cache) die-shrunk Northwood. What's the point of the prescott core if it could reach 5Ghz, but due to power/heat limitations can't - a die-shrunk northwood even if it would only allow for 4.5Ghz or so would be much better (due to less logic transistors it would certainly use less power).

 

[incurable]

Volari's definitely the bigger flop, which which chip was the biggest in proportion to its hype and expectations?

Volari on the other hand - XGI has very little experience compared to ATI and NV, neither SIS nor Trident had anything groundbreaking before, nobody said it would be a R300 killer but somehow everyone had high expectance of it and decided that if it couldn't kill the 9800XT card (which costs over 140 euros more than Volari over in Europe but never mind), then it had to be a flop.

Volari Duo (349 €) doesn't just not kill 9800 XT (410 €), it also gets seriously abused by the 9800 Pro 256 (310 €), 9800 Pro 128 (250 €), 9700 Pro (220 €) and even 9600 XT (130 €) in pretty much everything, sometimes even if they're cheating!

Prescott sure is very hot, as was NV30, but Volari easily takes the crown for biggest flop.

Personally, I sympathise with the poor memory chips that get to spend their life on Volari Duos, you know, they had so much potential!

cu

incurable

 

[Bjorn]

As for me, I was waiting for the Prescott to be released before deciding which P4 to buy...and it looks like I'll be getting a 3.2 Nothwood. (Now for about $260 street.) Mostly because even though clock-for-clock the Pescott is fine WRT performance, Northwood power consumption is considerably less than Prescott. Less heat=quieter system, and I'm shooting for as quiet a system as I can get.

If you're shooting for a quiet system, then how about waiting for a 90 nm Athlon 64 ? (Which is rumoured to run extremely cool)

 

[arjan de lumens]

If you're shooting for a quiet system, then how about waiting for a 90 nm Athlon 64 ? (Which is rumoured to run extremely cool)

AMD has specified a TDP (Thermal Design Power) of 89 Watts for its 0.13u Hammer parts; IIRC this number is increased to 103W for its 0.09u parts. So if noise is a concern, I wouldn't hold my breath for AMD's 0.09u chips.

 

[Tim Murray]

arjan, wouldn't that be to allow for headroom with regards to much faster clockspeeds, which in turn would require more voltage and more heat? or would a 0.09u A64 consume more power per clock because of leakage?