Even performance wise? The Parhelia has the best image quality (with the FAA and so on) but it lacks in performance.
If the Parhelia was to have the same clock speed as the Ti 4600, would it be the best card up to date?
- Thread starter Emultra
- Start date
I read somewhere that a higher-clock version of the Parhelia (with 256MB memory) was going to come out this summer.
They can't just "boost" the clockspeeds on the cards, because they are using a .15 micron process (used in gf3/4s) though, Im sure Nvidia manages to get way better yields than Matrox can.
Also, As far as the 256mb version, I don't think its clocked any faster, just has more memory. Which... is a sorta win/lose. Parhelia has enough bandwidth, but rather, it uses it very inefficiently due to the lack of hyper-z technologies. More cost, but probally, not a drastic improvement as you might expect.
Even if they did manage to clock it higher, its going to be no where near Ti4600 speeds, and even if it DID get that high (very very very unlikely, 99.9%), Nvidia/ATi both have a new card coming out soon.
I personally don't see a market for Matrox this generation. Good luck next time.
Also, As far as the 256mb version, I don't think its clocked any faster, just has more memory. Which... is a sorta win/lose. Parhelia has enough bandwidth, but rather, it uses it very inefficiently due to the lack of hyper-z technologies. More cost, but probally, not a drastic improvement as you might expect.
Even if they did manage to clock it higher, its going to be no where near Ti4600 speeds, and even if it DID get that high (very very very unlikely, 99.9%), Nvidia/ATi both have a new card coming out soon.
I personally don't see a market for Matrox this generation. Good luck next time.
According to Anand's own review:
"Matrox will be offering a higher clocked version of the Parhelia with 256MB of memory later this year; a 64MB card will also follow."
Surely some of the parts to come out of the fab will support a higher clock rate, even using their current 1500Å process.
"Matrox will be offering a higher clocked version of the Parhelia with 256MB of memory later this year; a 64MB card will also follow."
Surely some of the parts to come out of the fab will support a higher clock rate, even using their current 1500Å process.
hi,
a couple of things to comment on, or to put it another way, correct.
in reply to "Emultra"
no they could'nt, if you compare the physical size and complexity of the chip compared to its rivals, its both physicaly larger, and complex, because of its added complexisity, it cannot be run reliably at high speeds. only 1 in a batch one around 1000 can be rated to these speed, so if they were to increase the speed, and use only that one chip, 999 would be wasted. not good business practice.
in reply to"heliomphalodon"
no there is not a high rated one being released. the only difference between the 128mb version and the proposed 256mb version is the memory size.
in reply to "Blurry"
if they moved to the 0.13micron process, then it would reduce cost, not increase them. DUH!. The reduced size means that more chips can be produced per silicon wafer. so, a wafer that could do say, 100 chips in 0.15, could maybe do 130 chips at 0.13. so as the micron process decreases, so does the cost. the only thing that goes up is the number of chips per piece of silicon.
as the micron sizes decreases, the heat generated by the chip goes down, & the price. the theoretical speed also goes up. thats why with a 486 the chips were huge, rated to 33-100mhz. the athlon actual size is very small, hence, the speeds of 600mhz-1.7ghz. now some will say, "but the p4s size is huge, and thats fast", wrong, the p4 itself is very small and only looks huges because it is benith a heat spreader.
as a side note to both "Emultra" & "Blurry".
if matrox moved to a 0.13 micron, the physical size would decrease, allowing a SLIGHTLY higher clock speed to be acheived.
no offence to anyone listed above, but get your facts right before you comment on something that you obviosly know little about.
a couple of things to comment on, or to put it another way, correct.
in reply to "Emultra"
no they could'nt, if you compare the physical size and complexity of the chip compared to its rivals, its both physicaly larger, and complex, because of its added complexisity, it cannot be run reliably at high speeds. only 1 in a batch one around 1000 can be rated to these speed, so if they were to increase the speed, and use only that one chip, 999 would be wasted. not good business practice.
in reply to"heliomphalodon"
no there is not a high rated one being released. the only difference between the 128mb version and the proposed 256mb version is the memory size.
in reply to "Blurry"
if they moved to the 0.13micron process, then it would reduce cost, not increase them. DUH!. The reduced size means that more chips can be produced per silicon wafer. so, a wafer that could do say, 100 chips in 0.15, could maybe do 130 chips at 0.13. so as the micron process decreases, so does the cost. the only thing that goes up is the number of chips per piece of silicon.
as the micron sizes decreases, the heat generated by the chip goes down, & the price. the theoretical speed also goes up. thats why with a 486 the chips were huge, rated to 33-100mhz. the athlon actual size is very small, hence, the speeds of 600mhz-1.7ghz. now some will say, "but the p4s size is huge, and thats fast", wrong, the p4 itself is very small and only looks huges because it is benith a heat spreader.
as a side note to both "Emultra" & "Blurry".
if matrox moved to a 0.13 micron, the physical size would decrease, allowing a SLIGHTLY higher clock speed to be acheived.
no offence to anyone listed above, but get your facts right before you comment on something that you obviosly know little about.
Can you provide a reference to the "fact" that the 256MB version will not have a higher clock?
My claim obviously came from Anand's review.
Well, there is no offical word on it yet, so I wouldn't stress it out. Anand most likely did not have a 256mb Parhelia at the time he wrote that, so his information is not 100% true. We've never heard an offical word from Matrox as far as the clock speeds and extra memory. So its more like, wait and see.
Not neccessarly true for all cards, but Nvidia as of last 2 generations (Ti200, Ti4200), more Ram = Slower Clock Speed. And if matrox does not lower the clock speeds, every single card with the "extra ram" option has never been clocked faster. TNT2, Geforce 2 MX/GTS/Ti, GF3 Ti200, GF4 Ti4200, Radeon 8500, etc etc, the list goes on.
Just wait and see.
Not neccessarly true for all cards, but Nvidia as of last 2 generations (Ti200, Ti4200), more Ram = Slower Clock Speed. And if matrox does not lower the clock speeds, every single card with the "extra ram" option has never been clocked faster. TNT2, Geforce 2 MX/GTS/Ti, GF3 Ti200, GF4 Ti4200, Radeon 8500, etc etc, the list goes on.
Just wait and see.
although anandtech is good for information, try looking at other reviews of it on sites such as toms hardware and others, there are at least 6 independant reviews and summerys of the card that ive seen.
also if you search around, there is an interview with matrox about the cards themselves.
in short, the parhelia is not designed to be a gamers only card. belive it or not, more that just games need lots of video memory.
In a business way, Matrox is similar to the old ATi, before they sold chips to 3d party vendors, they willl most likely release all three proposed versions, 64/128/256mb, as full and lite versions. similar to the "LE" range of ATI`s.
also if you search around, there is an interview with matrox about the cards themselves.
in short, the parhelia is not designed to be a gamers only card. belive it or not, more that just games need lots of video memory.
In a business way, Matrox is similar to the old ATi, before they sold chips to 3d party vendors, they willl most likely release all three proposed versions, 64/128/256mb, as full and lite versions. similar to the "LE" range of ATI`s.
BoomAM
A Northwood is around 140 sq mm
A Willy is almost 220 sq mm
Compared to a Palomino Athlon that's just under 130 sq mm, and a TBred Athlon thats around 80 sq mm.
Die size alone doesn't translate into MHz.
Well, actually the P4(both Willamette and Northwood) are larger than the Athlon family.
A Northwood is around 140 sq mm
A Willy is almost 220 sq mm
Compared to a Palomino Athlon that's just under 130 sq mm, and a TBred Athlon thats around 80 sq mm.
Die size alone doesn't translate into MHz.
i review site, i forget which, clocked a geforce4 to the same speeds as the parhelia, and the geforce4 still beat up on the parhelia. that should indicate what higher clockspeeds are going to do to the parhelia.
i never said that the die size alone equates the speed of the chip, and neither did i say that the p4 is bigger or smaller that the athons core.
what i was trying to say was that in order to increase the clock speed of the parhelia to a reasonable level, and retaining the complexisity of the chip, then a smaller micron size has to be reached to ensure safe running
Just like any other card that comes out, its going to take time for the drivers to mature.
Everyone seems to forget the 8500 was barely faster than a regular Geforce3 when it was released. The GF3 TI500 was smoking it at that time.
Everyone seems to forget the 8500 was barely faster than a regular Geforce3 when it was released. The GF3 TI500 was smoking it at that time.
good point, becides the technologies like HyperZ and occulation culling are driver baced, so maybe matrox will implement these at a later date.
Smaller die size does not automatically translate to lower cost.
Firstly, you can't just shrink something and expect it to work as well. Ask all the ladies here
Alot of R&D actually goes into testing it at a smaller size and to take out the bugs of shrinking it down.
Secondly, you have to make fabs for the newer and lower micron lines. You think the machines will just draw smaller transistors? This will mean practically changing the whole manufacturing line.
Firstly, you can't just shrink something and expect it to work as well. Ask all the ladies here
Alot of R&D actually goes into testing it at a smaller size and to take out the bugs of shrinking it down. Secondly, you have to make fabs for the newer and lower micron lines. You think the machines will just draw smaller transistors? This will mean practically changing the whole manufacturing line.
in the long run, reduced micron size does reduce costs, considerably.
most companys that manufacturer the chips for others already have the "next gen" fab machines ready for implementation at a moments notice.
and unlike CPU manufacturers that have dedicated fabs for their chips, graphics chip developers send their designs to one company. so this one company could be producing ATI & nVidia chips, and have the new 0.13 micron machines ready for use. why do you think that nVidia announce the specs and micron size of their new chip months before it release? it because nVidia knows that the company that manufacturers their chips has the new machines ready, and so nVidia can advertise the fact that their new chip is the first of the smaller micron sized chips of that rating.
becides for all we know, the main fab plant/s for graphics chips could already have the 0.09micron machines ready.
most companys that manufacturer the chips for others already have the "next gen" fab machines ready for implementation at a moments notice.
and unlike CPU manufacturers that have dedicated fabs for their chips, graphics chip developers send their designs to one company. so this one company could be producing ATI & nVidia chips, and have the new 0.13 micron machines ready for use. why do you think that nVidia announce the specs and micron size of their new chip months before it release? it because nVidia knows that the company that manufacturers their chips has the new machines ready, and so nVidia can advertise the fact that their new chip is the first of the smaller micron sized chips of that rating.
becides for all we know, the main fab plant/s for graphics chips could already have the 0.09micron machines ready.
At 1600 x 1200:
In UT 2003 (Sun Temple), the Parahelia is 28.2/50.8 = 55.5% of a Ti 4600.
In UT 2003 (Antalus), the Parahelia is 23.5/37.4 = 62.8% of a Ti 4600.
In JK2, the Parahelia is 65.9/113 = 58.3% of a Ti 4600.
In SS2, the Parahelia is 37.3/51.7 = 72.1% of a Ti 4600.
Thus to get into the same ballpark as the GeForce4 Ti 4600, the Parahelia needs to close the gap by 1/72.1% or 1.386. This would get the Parahelia running neck-and-neck in Serious Sam 2 at 1600x1200. Thus the Parahelia needs to be runing at 220MHz * 1.386 = 305MHz. To exceed the Ti 4600 in all benchmarks, the Parahelia would need a speed of around 400 MHz.
A shrink to 0.13 micron, plus some process improvements, might do the trick for the 305 MHz speed. (Remember the speed bump ATI pulled from the Radeon to the Radeon 7500?) However, I doubt that such a shrink could occur before sometime in the fall. This puts it after the R300 and near the NV30.
In UT 2003 (Sun Temple), the Parahelia is 28.2/50.8 = 55.5% of a Ti 4600.
In UT 2003 (Antalus), the Parahelia is 23.5/37.4 = 62.8% of a Ti 4600.
In JK2, the Parahelia is 65.9/113 = 58.3% of a Ti 4600.
In SS2, the Parahelia is 37.3/51.7 = 72.1% of a Ti 4600.
Thus to get into the same ballpark as the GeForce4 Ti 4600, the Parahelia needs to close the gap by 1/72.1% or 1.386. This would get the Parahelia running neck-and-neck in Serious Sam 2 at 1600x1200. Thus the Parahelia needs to be runing at 220MHz * 1.386 = 305MHz. To exceed the Ti 4600 in all benchmarks, the Parahelia would need a speed of around 400 MHz.
A shrink to 0.13 micron, plus some process improvements, might do the trick for the 305 MHz speed. (Remember the speed bump ATI pulled from the Radeon to the Radeon 7500?) However, I doubt that such a shrink could occur before sometime in the fall. This puts it after the R300 and near the NV30.
NVIDIA all the way .. all the games are mostly supported by nvidia.... the best investment is in NVIDIA , i cant wait to see the new generation videocard they are workin on now ... I read a article about John Carmack commenting about the videocards and he didnt have any good words with the Matrox, NVIDIA IS GOD ! =)))
There is no question that the Parhelia will have the best image quality this seems to be a priority for Matrox. If you go into any video editing shop you will always find a Matrox. I know that some computer vision research labs use Matrox as standard HW. I would imagine anything from Blizzard would look amazing on the Parhelia.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter