Gambling on an unannounced (blower?) Gigabyte RTX 2080 Ti

[ratbuddy]

So, I've been buying early launch stuff from Bottom Line Telecommunications for years, since the E8600 came out. They always get stuff at least as early as anyone else, being a reseller for Ingram Micro, and using drop shipping directly from that particular behemoth of a distributor. They most recently got me my launch day Vega Frontier, and overnighted it even with my late order time.

I've been watching inventory on shopblt.com since RTX stuff was announced. Nothing showed up until a week or two ago, and at least one of the listings is plain old wrong, see the top one:

B5NQ079 GV-N2080GAMINGOC-8GC GEFORCE RTX 2080 TI GAMING OC 8GB GDDR6 3XDP1.4 $889.10
B5NQ649 GV-N2080TURBO OC-8GC GEFORCE RTX 2080 PCIE 8GB GDDR6 DP1.4 HDMI 1710MHZ $889.10
B5NQ648 GV-N208TTURBO OC11GC GEFORCE RTX 2080 TI PCIE 11GB GDDR6 DP HDMI USB C 1545MHZ $1,324.95

Clearly not a Ti at that price and with only 8GB. Google the UPC and you get this card, http://www.microcenter.com/product/...rclocked-triple-fan-8gb-gddr6-pcie-video-card

The third listing is interesting though: 11GB, and a model number that doesn't seem to exist - but wait a minute, the UPC and model of the middle card, the RTX 2080 GV-N2080TURBO doesn't seem to exist either. If you look at the known UPCs and model numbers for existing Gigabyte cards, it's these:

RTX 2080 Windforce OC: model GV-N2080WF3OC-8GC, UPC 889523014974
RTX 2080 Gaming OC: model GV-N2080GAMING OC-8G, UPC 889523014981
RTX 2080 Windforce OC: model GV-N208TWF3OC-11GC, UPC 889523014967
RTX 2080 Gaming OC: model GV-N208TGAMING OC-11, UPC 889523014998

The middle card in my quote above has a listed model GV-N2080TURBO OC-8GC with UPC of 889523015254
The bottom card shows of GV-N208TTURBO OC11GC with UPC of 889523015261

This might be meaningful.. Both 'unknown' UPCs are in pattern with Gigabyte's typical UPC structure, and come numerically later than the known four preorder cards. Both unknown cards have 'TURBO' in the model number, where we would see 'WF3' or 'GAMING' on the known cards. The dimensions for both cards show 10.7" x 4.4" x 1.5" which doesn't match any of the published dimensions for the existing cards. It's probably a blower card. One would expect the price to be lower in that case. One would hope for a lower price in that case.

Sadly, the in stock estimate for the unknown RTX 2080 Ti slipped from 10/9 (iirc) to 10/16. They show 100 incoming, with 0 preordered so far. I just ordered one, so that's 99 remaining.

It's a gamble for sure - if they are just RTX cards with blower coolers, the joke is on me. I'm 98% sure that's the actual case - see https://www.gigabyte.com/us/Graphics-Card/GV-N108TTURBO-11GD#kf for example, the TURBO in the model number is pretty telling. The 2% says maybe it's something neater than that. If not, you'll see the card in the classies here very quickly

For anyone who actually wants a blower RTX card, now's your chance to nab an easy preorder, their RTX 2080/Ti stuff is here https://www.shopblt.com/search/order_id=164321699&s_max=25&t_all=1&s_all=RTX+2080&search=Search

 

[Alaric]

Doesn't seem to matter what card you get, as far as performance and OC is concerned. Nvidia's (in)famous power limits are back with a vengeance.

 

[Maviryk]

I wanted a blower card, but the price VS performace of the new 2080's vs a used 1080 ti... 10-15% more performance at double the cost...

I made my last PC purchase last a decade, I can make the 1080ti last for at least as long.

 

[Alaric]

Rather than delete and screw up the thread continuity, I'll just say this post is wrong, factually incorrect. Post #16 explains where I screwed up and how I figured it out. With a friendly nudge from ED. As always.

Yeah, it looks like comparing the stock 20xx cards to OC 10xx cards is appropriate. The old argument "Well, you can OC the new *insert component here* , too!" was neatly invalidated by the power delivery on the 20xx, making the performance gap even smaller between 10xx and 20xx.

 

[ratbuddy]

Let me know which 10-series card I can buy that will come anywhere close to the 2080 Ti at 4k, OCd or no

Anyway, I changed this BLT preorder to an EVGA card, decided I just don't want to risk them shipping me a blower card - my case has insane airflow, I'd rather have a quiet card that takes advantage of it. The Aorus cards will be announced very soon, maybe I can snag one of those if the EVGA card doesn't ship by then.

 

[Alaric]

Nah, the 2080 Ti is The Card For 4k. Not even a question, there. The 2080 Ti is the highest performing card available, bar none. The gap is much closer between an OC 1080 Ti and a 2080, however.

 

[EarthDog]

Yeah, it looks like comparing the stock 20xx cards to OC 10xx cards is appropriate. The old argument "Well, you can OC the new *insert component here* , too!" was neatly invalidated by the power delivery on the 20xx, making the performance gap even smaller between 10xx and 20xx.

Patently false. The power delivery bits have literally ZERO to do with their overclocking capabilities or limits. It is software limits. The HW on Turing is more robust from the FE on up this time around (not that HW was a worry for pascal either).

Their limits (like AMD has also...) are always in place. They never left... it wasn't a return with vengeance. And to be quite honest, the FE cards have a HIGHER power limit than the MSI Gaming X Trio 2080 I am testing now (23% vs 10%). That said, both cards ended up around the same clockspeed in the low 2000s.

 

[Alaric]

Just going by the video. GN has always been fairly reliable. If he's wrong on this one I stand corrected.

 

[EarthDog]

He says in the video its a power limit... due to the software limits, not the hardware. He even states its the 'best PCB/VRM that NVIDIA has come out in.......forever".

The FUD machine with a side of Haterade is in full force.

 

[Alaric]

I didn't say the hardware was faulty. What are you reading? I said the power limits are the issue, and that's the BIOS. It's by design, I didn't say it's due to poor design.

 

[EarthDog]

"Power delivery" I assumed to be hardware (as that is what most would think when hearing that term). My apologies if you meant something else. Regardless, the power limits have been there for generations now. It is nothing new and not back with a vengeance.

Carry on!

 

[Alaric]

LOL. I'm getting dizzy bouncing between threads. Yeah, I meant their built in restrictions. Steve showed that no matter how much you cool it you aren't getting much, if any, improvement in clock speeds. That seems to be a difference compared to the 10xx series cards which benefited from better clock speeds with better cooling. The 20xx cards are still fast as $#@!, but better cooling gave a larger percentage of gain compared to the 20xx.
That lets the 10xx cards get closer to 20xx performance without the attendant gains available from the new cards.

 

[EarthDog]

You get a few bins with better cooling... just like Pascal. Nothing new, really.

EDIT: The highest overclock doesn't change because of that software power limit. Also nothing new.

 

[Alaric]

But does it mean the 20xx just gets better stock cooling so they maxed it out? Or it is more efficient so it's easier to keep stable at a greater percentage of its potential? The 20xx (according to one video, so, not The Last Word) in that example has less OC headroom as a percentage. Or did I misunderstand him and it just doesn't need higher end cooling, but has Pascal-like headroom?

edit: And yeah, I'll go back and read your review again. It's easily as reliable as any out there (and way ahead of some. LOL)

 

[EarthDog]

The cooler on the FE cards are notably better than the blowers they used in the past. Both the heatsink with its dual vapor chamber and fin array are notably better. The cards run cooler and quieter than Pascal FEs.

Those cards are not really temp limited is the point he is trying to make (or what I am taking away from it anyway). With the card running cooler, a user will get a couple/few boost bins higher during normal operation as that is the way Boost 4.0 (and lower) works. The peak clocks are limited by the software implemented power limits.... as it was with Pascal, and generations before that which used the power limits. As far as which overclocks further by percent........ who knows.

 

[Alaric]

Where I failed was watching the video without having the 10xx data to compare. Can't blame Steve, I should have done some actual comparison of numbers from both series. Rereading the excellent write up at OCF would have shortened this thread by a few of my posts, too....

You got a 23% bump in power from the card, whereas the utility for my card stops at 20%-advantage 20xx. You got under 100 MHz clock speed boost. My card, however, has boosted a whopping 101 MHz with a power bump (20%). Sooo, margin of error there. My card has a pretty good stock boost clock (2164 MHz) but the max clocks on the 1070 GPUs is pretty comparable across the AIBs. Maybe a higher percentage is available from lower clocked versions, but the actual numbers comparison between my 10xx and your review samples makes my initial premise go up in smoke. That's looking at percentages only, as clock speeds aren't otherwise relevant to my (incorrect) point.

So now I'll go back and point to this post for anyone who doesn't want to spend 10 minutes finding out how long it took me to figure out I was wrong. And thanks to ED for leading the slow kid into the light.

edit: Oh, and if anyone hasn't read ED's excellent review, it's here https://www.overclockers.com/nvidia-geforce-rtx-2080-and-rtx-2080-ti-review/