So you're telling us that 16 lanes to two GPU's on one slot is somehow significantly different than 16 lanes to two GPU's on two slots?Originally Posted by Caparroz
edit
No eyefinity numbers yet :/
So you're telling us that 16 lanes to two GPU's on one slot is somehow significantly different than 16 lanes to two GPU's on two slots?
edit
No eyefinity numbers yet :/
Last edited by STEvil; 03-07-2011 at 10:07 PM.
All along the watchtower the watchmen watch the eternal return.
580 SLI is a lot more expensive, though.
Where are the Eyefinity reviews?!? Testing these cards at 25x16 is a waste of time.
Murray Walker: "And there are flames coming from the back of Prost's McLaren as he enters the Swimming Pool."
James Hunt: "Well, that should put them out then."
Yup. Think latency...
One single card, be it 1, 2 or a thousand GPUs share information than dump it on a 16x lane bus. Two GPUs over an (much slower) 8x pcie lanes have to share information back and forth over a much narrower band... It's a completely different thing.
Edit: My question is if this back and forth comunication over a "narrow" bandwidth, given the computing power provided by today's fastest GPUs driving a very high resolution proves to be a bottleneck or not.
Last edited by Caparroz; 03-07-2011 at 10:45 PM.
Murray Walker: "And there are flames coming from the back of Prost's McLaren as he enters the Swimming Pool."
James Hunt: "Well, that should put them out then."
Last edited by onethreehill; 03-07-2011 at 11:00 PM.
you know that sli/crossfire connector thing.. its not only there for show.Yup. Think latency...
One single card, be it 1, 2 or a thousand GPUs share information than dump it on a 16x lane bus. Two GPUs over an (much slower) 8x pcie lanes have to share information back and forth over a much narrower band... It's a completely different thing.
Edit: My question is if this back and forth comunication over a "narrow" bandwidth, given the computing power provided by today's fastest GPUs driving a very high resolution proves to be a bottleneck or not.
the PLX chip isnt perfect either, and then there's that funny little feature that never saw the light of day since being mentioned for the 4870x2's...
I mean, if its that important why havent ATi been vocal about it?
All along the watchtower the watchmen watch the eternal return.
hmm interesting, what is the NDA date for these things?
I'm not sure if I'm following you right since English isn't my native language as you already probably noticed it...
All I'm saying is that no matter how much GPUs are in a single card, all of them is comunicating with the rest of the system as a whole with the given bus (pcie 1x, 8x or 16x).
I know there is not a magic interconnect chip speculated since before the 4870x2 launch. It's being speculated since before rv670x2 launch but I digress...
But saying a 2 Y GPU single card over a 16x pcie bus is the same as a 2 cards containing the same Y GPU over a 8x pcie bus is not correct (I know you didn't say this but the same principle applies). It doesn't matter if there's a CF/SLI bridge (wich have a finite bandwidth), they still have to comunicate with the rest of the system.
Last edited by Caparroz; 03-07-2011 at 11:32 PM.
Murray Walker: "And there are flames coming from the back of Prost's McLaren as he enters the Swimming Pool."
James Hunt: "Well, that should put them out then."
The NDA for the 6990 has already expired.
Disappointing msrp indeed. PassFor $700, it can sit on a store shelf somewhere until hell freezes over, or I change my underwear, whichever comes first.
And either way you go about it both GPU's have the same number of lanes going to the same PCI-E controller.I'm not sure if I'm following you right since English isn't my native language as you already probably noticed it...
All I'm saying is that no matter how much GPUs are in a single card, all of them is comunicating with the rest of the system as a whole with the given bus (pcie 1x, 8x or 16x).
I know there is not a magic interconnect chip speculated since before the 4870x2 launch. It's being speculated since before rv670x2 launch but I digress...
But saying a 2 Y GPU single card over a 16x pcie bus is the same as a 2 cards containing the same Y GPU over a 8x pcie bus is not correct (I know you didn't say this but the same principle applies). It doesn't matter if there's a CF/SLI bridge (wich have a finite bandwidth), they still have to comunicate with the rest of the system.
All along the watchtower the watchmen watch the eternal return.
Pretty much what I had expected. Initially (as in last year ) I figured theyd have stuck with a Cayman Pro configuration to keep power draw and heat output at bay so I was a little shocked when the first mention of 375watts came to light but all said and done they positioned the product were it belongs. With that in mind why most would chose one over 2 6950s is beyond me. I'm sure the odd bencher will eat up a pair of these as I can see the attraction there but given the power envelope things have reached the benifts of such products are shrinking. Its not like you can throw one of these in a mATX case and expect it not to explode( and yes I know it can be nice to have 1 less PCIe slot occupied for those who use a lot of add in cards - this is one of the few things I enjoyed about the dual gpu cards I've used as I use tv tuners, sound cards and other devices )
I'm curious to see how these fair on water with a proper motherboard ( I wouldn't dare try to seriously overclock one of these on the average motherboard, melting 24pin connectors... no thanks )
Last edited by Chickenfeed; 03-07-2011 at 11:56 PM.
Feedanator 7.0
CASE:R5|PSU:850G2|CPU:i7 6850K|MB:x99 Ultra|RAM:8x4 2666|GPU:980TI|SSD:BPX256/Evo500|SOUND:2i4/HS8
LCD:XB271HU|OS:Win10|INPUT:G900/K70 |HS/F:H115i
The pcie controller adresses devices not GPU's.
Murray Walker: "And there are flames coming from the back of Prost's McLaren as he enters the Swimming Pool."
James Hunt: "Well, that should put them out then."
Reviews
http://www.bit-tech.net/hardware/201...-6990-review/1
http://www.legitreviews.com/article/1566/1/
http://lab501.ro/placi-video/amd-radeon-hd-6990-review
http://techgage.com/article/amd_rade...s_card_review/
http://www.kitguru.net/components/gr...hd6990-review/
http://nl.hardware.info/reviews/2028...hd-6990-review
Last edited by onethreehill; 03-08-2011 at 02:47 AM.
Wow, just finished reading most of those reviews. Cliff notes: expensive, incredibly loud, hot and you would be better served with two cooler running traditional single GPUs in X-Fire/SLI. Basically what I was expecting.
GPU: 4-Way SLI GTX Titan's (1202 MHz Core / 3724 MHz Mem) with EK water blocks and back-plates
CPU: 3960X - 5.2 GHz with Koolance 380i water block
MB: ASUS Rampage IV Extreme with EK full board water block
RAM: 16 GB 2400 MHz Team Group with Bitspower water blocks
DISPLAY: 3x 120Hz Portrait Perfect Motion Clarity 2D Lightboost Surround
SOUND: Asus Xonar Essence -One- USB DAC/AMP
PSU: EVGA SuperNOVA NEX1500
SSD: Raid 0 - Samsung 840 Pro's
BUILD THREAD: http://hardforum.com/showthread.php?t=1751610
viva la 6950 CFX
GPU's are not devices?
All along the watchtower the watchmen watch the eternal return.
The price is too high, even a pair of 6970's is cheaper and performs better, I guess this is due to low stock and NVIDIA is yet to respond with the GTX 590.
AMD Threadripper 12 core 1920x CPU OC at 4Ghz | ASUS ROG Zenith Extreme X399 motherboard | 32GB G.Skill Trident RGB 3200Mhz DDR4 RAM | Gigabyte 11GB GTX 1080 Ti Aorus Xtreme GPU | SilverStone Strider Platinum 1000W Power Supply | Crucial 1050GB MX300 SSD | 4TB Western Digital HDD | 60" Samsung JU7000 4K UHD TV at 3840x2160
+1 but for guys with only 1 x 16x slot and a 2nd 4x PCIE slot, this is a good option.
Now where is Nvidia with the 590 at 600 USD or less to keep things competitive?
the price will surely decrease, i think there's 2 reasons it's too high 1) they wait the 590 and so on will surely be able to kick it with a lower price, whatever is the perf of the Nvidia one. 2) availability ...
But after read some reviews, i really like the work have do AMD on it ... not completely perfect due to the 40nm... but well done.
5970 was pretty conservative on clock speed ( reference model ). The 6990 is all but not conservative. ( i want to believe this will maybe be a good lesson for next future high end cards )
Now AIB as Sapphire, Asus, Gigabyte, XFX have a nice support for work on it.. Im really curious to see what they can take out of this one.
( for peoples who want buy the card, it's urgent to wait and see what AIB's will bring )
Last edited by Lanek; 03-08-2011 at 02:49 AM.
CPU: - I7 4930K (EK Supremacy )
GPU: - 2x AMD HD7970 flashed GHZ bios ( EK Acetal Nickel Waterblock H2o)
Motherboard: Asus x79 Deluxe
RAM: G-skill Ares C9 2133mhz 16GB
Main Storage: Samsung 840EVO 500GB / 2x Crucial RealSSD C300 Raid0
We've just published our in depth tech article on the HD 6990 and we've also published a video review (swedish) on the card where we did some none scientific sound tests on tape. Our reference card has some pretty crazy coil whine which we dont know if consumer cards will suffer from but might be worth keeping in mind.
You can see the card in action here, Also hear it in action.
http://translate.google.com/translat...genomgang.html
Noisless room sound test at 15:15
ASUS GTX 580 DCII max load at 15:25
HD 6990 full load furmark at: 15:35
Notice the insane coil whine at 15:50
It remains to be seen if all cards have the same problems with coil whine during memory load. And it seems to be isolated to memory load and increase with memory frequencies.
Useless card ...
True to a degree.
[[Daily R!G]]
Core i7 920 D0 @ 4.0GHz w/ 1.325 vcore.
Rampage II Gene||CM HAF 932||HX850||MSI GTX 660ti PE OC||Corsair H50||G.Skill Phoenix 3 240GB||G.Skill NQ 6x2GB||Samsung 2333SW
flickr
Useless post ...
"Death is so sure of winning its race against us that she give us all our lifetime as an advantage"
Rig #1 "Core2Quadra"
Asus P5E
Intel Core2Quad Q6600
2x2Gb Corsair Dominator DDR2
XFX GeForce 285GTX
XFX GeForce 8800GTX (Physx and Triplehead Support)
Intel X-25M G2 80Gb
2xWD Caviar Blue 1Tb each (RAID 0)
AeroCool 750Horse Power
Coolermaster Cosmos S case
Cooling: all internal Dual Water Loop
CPU Loop: Swiftech MCR Drive + D-Tek Fuzion v2
GPUs Loop: Swiftech MCR320QP, MCP355, EK-Res, EK FC285GTX, D-Tek Fuzion GPU
EK & BitsPower fittings & Primochill Pro Black tubing
how can you claim "scientific" and "recording to be played back on user's system with different audio config and volume" at the same time ?
Posting Permissions
Reply With Quote
Bookmarks