Results 1 to 25 of 95
Thread: (semiaccurate.com)
Hybrid View
-
29th July 2010 22:29 #1
(semiaccurate.com)
AMD FX 959o@47oo@default, NB@2oo@ 1.33 v.; Silver Arrow SB-E; ASUS Crosshair V Formula Z; 4x4Gb Corsair CMX4GX3M1A 16oo C7 @2133 10-12-11 2T 1.66 v.; Sapphire TRI-X R9 290X, Corsair TX85oM, LG34UM95-P etc.
-
29th July 2010 22:39 #2
...
Nvidia has no real future in consumer GPUs, products that account for approximately 60% of their income. By the time the company has a competitive architecture again, huge swathes of the market will be obsolete due to CPUs with integrated graphics. While there may be blips from here on out, Nvidia's future is not looking viable, much less bright
.
-
29th July 2010 22:42 #3
-
29th July 2010 23:04 #4
-
30th July 2010 16:38 #5
-
29th July 2010 23:10 #6
-
30th July 2010 08:31 #7
-
30th July 2010 09:11 #8Registered User
Join Date: Jan:2006
Location:
Posts: 577
: !
-
30th July 2010 10:02 #9
.. ( , ), . ... , , ... 1000$ 100$. , 30-40$ . , -. 2 , , . , , , . , .
-
30th July 2010 10:25 #10
/ , 20 . ? "" . , "" .
| Aorus X570 Master | Ryzen 9 5900X | Hellhound 7900XT | G.SKILL 32GB DDR4 3800 CL16 | Enermax Galaxy 1000W | SSD Corsair MP510 960GB & WD-RE3 2 x 1TB RAID0 | Pioneer BDR-209EBK | Creative X-Fi Titanium Fatal1ty & GigaWorks S750 | ViewSonic VX3276-2K-MHD |
-
30th July 2010 10:47 #11
-
30th July 2010 12:08 #12CPU:Ryzen 7 5800X3D + Arctic Liquid Freezer II 280; MB:Asus Crosshair VII; RAM:2x8GB GSKILL F4-3200C14D-16GTZRX; VGA:Saphire RX 5700XT Nitro+; HDD:Samsung 970 EVO Plus 1TB+860 EVO 500GB+860EVO 1TB; PSU:Seasonic X-760; CASE:FD Define R6 TG; DSPL:Samsung C32HG70; SPKR:Logitech Z906;
-
30th July 2010 10:52 #13
Join Date: Apr:2008
Location: Sofia
Posts: 15,560
, . , - , 2/3 $100 ( "" - ), . 28 . ATI 2900 , 55 3-4 , . , 100. . . 104 , . 106 - 100 104. 5700 , -. - , .
Bad Wolf
: ClasAuP^ , Dictamnus, kafka, LionS7, randomelectric, Superunknown, atscata, Dranplqs, racerx
-
30th July 2010 12:15 #14
-
30th July 2010 10:54 #15Registered User
Join Date: Aug:2008
Location: Varna
Posts: 291
. 20 . ?
-
30th July 2010 21:03 #16
?
(ooppsss ), FX, - R600 ( , ), "" . , ... ' JHH -... ... . ... .
:
💙💙💙
💛💛💛
-
30th July 2010 21:40 #17
, ,

-, - , , , ,
, ACDC
.
: ,AMD FX 959o@47oo@default, NB@2oo@ 1.33 v.; Silver Arrow SB-E; ASUS Crosshair V Formula Z; 4x4Gb Corsair CMX4GX3M1A 16oo C7 @2133 10-12-11 2T 1.66 v.; Sapphire TRI-X R9 290X, Corsair TX85oM, LG34UM95-P etc.
-
30th July 2010 21:46 #18
''
i9 9900k B-Die 970 Evo 3060 Ti PRIME Ultra Gold NZXT water Z906 4k120 curved
-
30th July 2010 23:57 #19
-
31st July 2010 00:42 #20
Join Date: Apr:2006
Location:
Posts: 8,666
-
31st July 2010 00:43 #21
, , , . -
.
, .EVGA X299 FTW K|i9-7960X@4.7|4x8 Patriot Viper Steel 4000|GTX 1660 Ti|970 EVO 1 TB|Seasonic Focus GX-1000|Xigmatek Elysium|
Rampage IV Extreme BE|E5-1680v2@4.7|4x4 HyperX 1866|Cougar Aqua 240|GTX 1050 Ti|970 EVO 1/4 TB|CM 850 SilentPro|HAF-X|
-
31st July 2010 01:16 #22
-
31st July 2010 06:36 #23
Last edited by BobyTT; 31st July 2010 at 08:22.
-
31st July 2010 21:01 #24
-
31st July 2010 21:20 #25
, , , . .
.SemiAccurate Was Wrong
Fermi is Better and Worse than Predicted, for Different Reasons
Last Friday was the culmination of months of speculation and rumors regarding the Fermi Architecture, NVIDIAs response to AMDs Cypress line of GPUs first launched October of last year. What really set the tone for this product launch was the level of vitriol; which reached its zenith early this month, when Theo Valich of BrightSideofNews referred to the Nvidia PR department as Nazis.
Reviews are now out and its time to see whether or not I owe Charlie Demerjian an apology. So lets fact check the last 8 months of rumors and find out.
Reviewer Access
Nvidia is doing its best to keep the GTX480 out of the hands of the honest. This means that only journalists who are known to follow the reviewers guide closely, are willing to downplay the negatives, and will hit the important bullet points provided by Nvidia PR will be the ones most likely to gain early access to these cards.
On February 20th, Charlie pushed the rumor that Nvidia would cut out honest reviewers like Anandtech and Kyle of HardOCP. This was 100% false. Both of them got two 480′s for SLI and one 470 and their reviews were posted at 7:01 p.m. on Friday night.
Frequency
we hear the A3 production silicon has a top bin of 600/1200MHz, and that is after an average of two shader clusters are turned off.
This was simply wrong. The GTX480 runs at 700Mhz, 17% faster than predicted by SA.
Shaders
if you are going to fuse off one, you are forced to fuse off a full set of 32. Since the weak transistors are scattered evenly throughout the GPU, fusing off two would mean that you lose not two but 64 shaders.
Charlie managed to predict both 448 shaders and 512 shaders and managed to get both of them wrong. The GTX480 has only 480 shaders. As Anand stated in his review, only 32 shaders were cut off, the ROPs, cache and memory bus remain intact.
Power Consumption
ATI draws less power for Cypress, 188W TDP vs 225W TDP for the Fermi
they have managed to reduce the TDP of the cards 5W to 275W.
Despite shifting his predictions from 22W, to 275W, to 290W; all of these predictions were wrong. TechPowerup tags the cards power consumption at 320W and most reviewers seem to oscillate around that number, finding that in general the card uses 150W more than the comparable 5870. Meaning Charlie was off by either 100W or 30W; of course given that hes made about 5 different predictions, its likely that the focus will be on whatever figure is closest.
Interestingly enough, NVIDIAs website states the TDP for the GTX480 as 250W which understates the power consumption of the GPU by a full 70W; which quite frankly is misleading.
Performance
The GTX480 with 512 shaders running at full speed, 600Mhz or 625MHz depending on which source, ran on average 5 percent faster than a Cypress HD5870, plus or minus a little bit.
Unfortunately for Charlie, Anand disagrees stating:
The GTX 480 is between 10 and 15% faster than the Radeon 5870 depending on the resolution, giving it a comfortable lead over AMDs best single-GPU card.
Although thats not particularly fair, Guru3Ds conclusion is more nuanced,
Even so, when we tally up the results out of all titles tested, the GeForce GTX 480 wins nearly everywhere except in Anno 1404 and ironically 3DMark Vantage. There are scenarios where the GTX 480 is very close to the 5870, but there are also scenarios where the GTX 480 completely and utterly kicks the Radeon HD 5870 in the proverbial nuts. What surprised me was the fact that say from 1280×1024 up-to 1920×1200 NVIDIA seems stronger than the competition. And at 2560×1600 things tend to equalize a much more.
The point being that to call it a flat 5% increase was wrong and misleading.
Voltage
On the current A3 silicon, sources tell us that Nvidia is having to use both fixes, fusing off at least two clusters and upping the voltage.
The GTX480 runs at .96 .99 volts, according to TechPowerup (the only site who tested the voltage), below the 1.1 volts the 5870 uses; the claim that Nvidia would overvolt Fermi to raise the frequency was wrong.
So Why do People Think SA Was Right?
Charlie has been playing the same trick since September with his 1.7% yield article. As stated before, this article was based on a Google mistranslation. The apologists will claim that while the specific number was incorrect, the statement about yields was accurate.
The problem is that TSMCs problems making 40nm chips had been known for months before the article was written. As early as January, a full 9 months earlier, it was being reported that it was difficult to make 40nm chips; problems which were continually mentioned in the press throughout 2009:
Dirk Meyer, AMDs President and Chief Executive Officer, said in response to a query on constraints at TSMC, We were heavily constrained in the quarter and we could have done a lot more business were we not so I dont want to get granular other than to say were seeing progress, both in terms of delivery of wafers and underlying yields. But were constrained today.
This has been the modus operandi of SemiAccurate; take information that was publicly available, add in either speculation, fabrication or sensationalism, then publish it as if it was some new and shocking development. This is seen over and over again. The famous broken and unfixable article, consists of a rewording of an article by AnandTech, retconned to make it look like he had known this information months before and that it somehow confirmed their previous claims.
Ditto for the February 20th article in which they claimed to have access to leaked benchmarks and claimed 5% improvements were a major disappointment. Unfortunately, as early as November 2009, the consensus was that 5 to 10% improvements over 5870 were the best to expect. The claims that A. the performance was disappointing and B. that this was somehow a scoop, are both false.
As for the strong performance in tesselation. This was again, already known, AnandTech had already reported on the strong tesselation performance in January.
Ditto for the tape-out of the GF100. Demerjian claimed to have scooped the story in July 29th, yet it had already been reported that the A1 revision of the chip had been taped out as early as June 16th.
SemiAccurate lowered the level of discourse in Tech Journalism while bringing nothing of worth to the table. As of today, the website is back to being listed as a Satire website by Google News and hopefully it will remain there.




:
Reply With Quote
. , , .

, ...
05.05.23 ., 14:20 in