MSI GeForce GTX 1650 Gaming X 4 GB

[R0H1T]

The 750Ti has 60W TDP, not 75W as is the case with x50/Ti variants that followed. The 1650 as well as 1050Ti peak power consumption exceeds 75W a fair bit, now I do recall that the GTX 950 was cut down to fit 75W but didn't realize it was hacked off so much. Besides do you remember why the RX 480 was criticized early at launch, when the power consumption is this much at load buyers should be worried!

The 6 pin power connector isn't just for show, the variants without them probably will have to be considerably slower so as not to exceed 75W peak power consumption.
Scratch that the lesser versions aren't that big of a deal, the only thing to look out for is the large performance delta between them especially if 1650Ti is also sold @75W TDP.

 

[jabbadap]

Cards without 6-pin power connector gtx1050ti/gtx1050 does not exceed that even on peaks. Thus it shows on lower clocks and of course lower performance as you said. I think W1zzard have taken those "stock" gtx1050ti numbers from Palit KalmX review.

And that is the reason why I'm eager to see numbers for that lurking Palit StormX OC W1zzard has. As ugly as it looks that is more real gtx1650 what people are looking for than these out off spec monstrosities.

 

[londiste]

Besides do you remember why the RX 480 was criticized early at launch, when the power consumption is this much at load buyers should be worried!

According to several sites measuring the power consumption directly reference RX 480 was consuming 80W or more from PCI-e slot 12V. While the entire coverage over this was overblown, exceeding spec at stock is something that one should generally try to avoid.

The situation with GTX 1050Ti is not the same. As much as these were tested, none of the GTX 1050Ti models without 6-pin connector consumed more than the allowed 66W from PCI-e slot 12V. There are models with 6-pin connector that have higher TDP and do consume more power but they do so within spec. Of course, models without 6-pin connector are slower as they are more restricted by power. GTX 1650 seems to follow the exact same formula so far.

ASUS Strix variants of GTX 1050Ti and GTX 1650 are both overclocked, have increased TDP limit and both have 6-pin connector to feed the power. There will be a performance difference between base models at 75W and OC models at higher TDP. Palit StormX OC review should give a pretty good indication of how large that difference is.

Edit:
As you brought up GTX 1050Ti numbers checked the TPUs reviews for different cards (Card - Peak Gaming/Furmark - Average/Median):
- Asus Strix GTX 1050Ti - 83W/104W - 1740/1772MHz
- Palit KalmX GTX 1050Ti - 58W/54W - 1657/1683MHz
Asus had 5% performance lead over Palit which is honestly less than I was expecting although it matches the clock speed difference.

 

[notb]

What's the pont of advertising 75W "no power connector" needed when nearly every card released thus far needs it

Most don't. It's just that most gaming websites, obviously, focus on the most powerful models.
Just look at TPU's history with MSI. "Gaming" and "Lightning" variants are covered, not much else.
https://www.techpowerup.com/reviews/?category=Graphics+Cards&manufacturer=MSI&pp=25&order=date&p=1

The last MSI ITX card TPU tested was a 760 back in 2014 - despite the fact that these cards are really popular. It's just not the client segment TPU attracts.

 

[A.Stables]

@W1zzard do you have a reference/75W card without 6-pin connector to review?
There is Palit StormX in some of the tables, is that going to get a separate review?

This all the commentators talking you don't need a PCIE but all the cards on the net reviewed are with PCIE power im guessing another 10% less perf

 

[64K]

According to several sites measuring the power consumption directly reference RX 480 was consuming 80W or more from PCI-e slot 12V. While the entire coverage over this was overblown, exceeding spec at stock is something that one should generally try to avoid.

The situation with GTX 1050Ti is not the same. As much as these were tested, none of the GTX 1050Ti models without 6-pin connector consumed more than the allowed 66W from PCI-e slot 12V. There are models with 6-pin connector that have higher TDP and do consume more power but they do so within spec. Of course, models without 6-pin connector are slower as they are more restricted by power. GTX 1650 seems to follow the exact same formula so far.

ASUS Strix variants of GTX 1050Ti and GTX 1650 are both overclocked, have increased TDP limit and both have 6-pin connector to feed the power. There will be a performance difference between base models at 75W and OC models at higher TDP. Palit StormX OC review should give a pretty good indication of how large that difference is.

Edit:
As you brought up GTX 1050Ti numbers checked the TPUs reviews for different cards (Card - Peak Gaming/Furmark - Average/Median):
- Asus Strix GTX 1050Ti - 83W/104W - 1740/1772MHz
- Palit KalmX GTX 1050Ti - 58W/54W - 1657/1683MHz
Asus had 5% performance lead over Palit which is honestly less than I was expecting although it matches the clock speed difference.

I don't know what to think about PCIe specs. Remember the R9 295X2 card tested here? It had two 8 Pin power connectors so that would be 300 watts and then 75 watts from the slot. It would have been within specs at 375 watts yet it averaged 430 watts power draw gaming with a peak of 500 watts gaming and then running Furmark it was drawing 646 watts.

 

[W1zzard]

that most gaming websites, obviously, focus on the most powerful models.

I keep telling companies to sample their cheaper SKUs, I'd love to review cards for the masses, but most don't get it.

 

[notb]

This all the commentators talking you don't need a PCIE but all the cards on the net reviewed are with PCIE power im guessing another 10% less perf

eteknix tested 4 different models already. 2 with a power connector (Gigabyte Gaming, MSI Gaming) and 2 without (ASUS Phoenix, Palit StormX).
Yes, the overclocked ones (with a connector) are slightly faster - usually by 3-5%.
E.g. Metro Exodus 1080p high settings: 30 vs 31fps.
https://www.eteknix.com/asus-gtx-1650-phoenix-graphics-card-review/10/

Despite all that MSI and ASUS cards manage to pull almost 20W more - simply because they could. Of course this includes the second fan as well (~5W).

Palit is significantly slower because of the smaller heatsink.

 

[tajoh111]

It's funny how long it takes for someone to point out a correction if it is in AMD's favor.

I guess that what happens when you have such a strong team red viral marketing team.

http://links.em.experience.amd.com/...zMzNjE4NDc4MzQS1&j=MTQwMTYxNzUxNwS2&mt=1&rt=0

*Offer available through participating retailers only. 18+ only. Following purchase, Coupon Code must be redeemed by April 6, 2019, after which coupon is void. Residency and additional limitations apply. For full Terms & Conditions, visit www.amdrewards.com.
Campaign period begins November 15, 2018 and ends February 9, 2019 or when supply of Coupon Codes is exhausted, whichever occurs first. Eligible AMD Product must be purchased during Campaign Period. Offer void where prohibited.

https://www.amd.com/en/where-to-buy...r&utm_medium=referral&utm_source=DonanimHaber

The promo page for the game bundle is also dead.

The promo has been mostly dead since February 9th(only obtainable if retailer had extra codes and you emailed the seller personally for the codes).

And 100% dead since April 6th. Look for a card that says the bundle is included on newegg and you can't find anything. See how quickly bad news comes for Nvidia cards, and it has now been nearly 3 weeks to over 2 months to point out a correction for AMD which definitely affect the value of their cards. The sad thing is I bet all these AMD fans know that it has expired but would rather this information not be posted or corrected for because it helps AMD sell cards.

 

[R0H1T]

And why does this need to be posted here?

 

[tajoh111]

And why does this need to be posted here?

Because all the reviews on techpowerpowerup for the gtx 1650 mention the rx570 comes with 2 aaa games when they dont in reality and have not been included for a while. Do you think its better for people not to be informed of the actual truth? This should have been corrected sooner. Wizzard is not biased whatsoever, but I would expect the community to inform him sooner..but I suspect those that do know don't want to say.

 

[W1zzard]

the rx570 comes with 2 aaa games when they dont

oh? let me look into that

edit: this has been fixed, thanks for bringing my attention to it!

 

[HenrySomeone]

Lmao, yet another case of AMD not delivering what they promise!

 

[AMX85]

WHY? why people don´t realizing that is using old AMD Beta Drivers, 19.2

actual performance difference between 570 and 1650 is higher

Greetings

 

[John Naylor]

I usually don't look at cards in the price range as never been asked to build a PC w/ a $150 GFX card before. I originally gave AMD an easy win for this tier ... but have now realized it's a lot closer than i had originally thought. I took another look and saw that the 1650 actually beat the 570 in TPUs OC test ...and with that much of an OC .... it should compete fairly well, which is a lot better than I expected on 1st read.

At 1080p, the 570 has a 109% / 100% advantage ... but with the OC....

MSI 1650 = 100% x 1.155 = 115.5
Sapphire RX570 = 109% x 1.103 = 120.23

Still a win for AMD but that makes it only 4% (120.23 /115.5) faster ....

The MSI $1650 is $160 on newegg with a $45 Gaming Bundle (real value to any individual will of course vary by individual).
The Sapphire Nitro + is $130 ... and the 1650 depending on how much ya value Fortnite is costing ya $115 to $160.

On the other hand .... the 1650 card has the edge in noise (29 / 31 dbA), power (80 / 180 watts) and heat load (65 / 74 C) ... the electricity cost savings kicks ya back $55.26 over 3 years at average US electricity cost. However most folks don't look at it that way ... most won't get beyond that 570 is cheaper and its 4% faster. So I still have to give the 570 the win", but it's by no means a big one considering less heat, less noise, a whopping 100 watts less power and $55 "cash back". In short, what my thinking is .... I don't see nVidia making any effort to get more competitive price wise in the near future.

As far as drivers go ... the 570 is over 2 years old ... if it hasn't been tweaked by this point it never will ... Both cards are useless above 1080p, and no one needs more than 3 GB to run at 1080p.

 

[HenrySomeone]

Precisely, yet somehow almost everyone overlooks this angle. If 1650 was just a bit cheaper, it would simply be a flat-out better buy than the obsolete "gas guzzler" 570 that when OCed to the max can pull north of 200W, which is frankly insane for an entry level card...

 

[rruff]

a whopping 100 watts less power

Honest question... can AMD ever be in the same league for power efficiency in games without big changes to their architecture? They seem to do alright in mining and maybe some other compute tasks, but...

 

[John Naylor]

Who knows ... remember when AMD folks were making fun of the GTX 480 frying eggs

https://www.tomshardware.co.uk/gf100-fermi-egg-frying-gtx-480,news-33106.html

Green took a 10% hit in market share with that series as they spent time on a total architecture revamp ... Here's why i think that won't happen. Back them nVidia was chasing the console market, best thing that ever happened for their GFX card division was letting go of that market. Yes it's whole lot of revenue but margins are teeny and any bump in the road has to "get solved "off book".

As to the mining ... remeber that electrical costs also hits profits.

 

[londiste]

Honest question... can AMD ever be in the same league for power efficiency in games without big changes to their architecture? They seem to do alright in mining and maybe some other compute tasks, but...

It has been speculated a lot that GCN itself does put some limits to what AMD can do efficiency-wise. Radeon 7 showed that Vega (the current-latest iteration of GCN) needs a full process shrink to get to the same efficiency level as RTX cards. AMD can get into the same league for power efficiency but they either need a big change in the architecture or to compete with a bigger GPU against a smaller one as a lot of AMD cards seem to be a bit too far along the inefficiency curve.

 

[HenrySomeone]

Honest question... can AMD ever be in the same league for power efficiency in games without big changes to their architecture? They seem to do alright in mining and maybe some other compute tasks, but...

Almost certainly not without a major overhaul or well theoretically maybe on like 3-times smaller node (2-times smaller obviously doesn't cut it, as shown by the Laugheon 7 vs (16nm!) 1080Ti)

 

[jabbadap]

It has been speculated a lot that GCN itself does put some limits to what AMD can do efficiency-wise. Radeon 7 showed that Vega (the current-latest iteration of GCN) needs a full process shrink to get to the same efficiency level as RTX cards. AMD can get into the same league for power efficiency but they either need a big change in the architecture or to compete with a bigger GPU against a smaller one as a lot of AMD cards seem to be a bit too far along the inefficiency curve.

Hmh last time I looked Radeon VII is no where near the efficiency levels RTX cards have. I think you meant to say same performance levels or needs a another shrink from 7nm?

 

[londiste]

Hmh last time I looked Radeon VII is no where near the efficiency levels RTX cards have. I think you meant to say same performance levels or needs a another shrink form 7nm?

Undervolted Radeon 7 fairly regularly gets to around 220-230W of GPU-only draw without performance hit. Sometimes less, sometimes more. This is in the ballpark. From undervolting results it can be argued that AMD can get the same efficiency if they only reigned in their default voltage settings. The primary reason they do not is that they have a performance level they need to hit which today is RTX2080 and at this point in the efficiency curve the variability is large enough that they feel the need to overvolt a little bit by default.

 

[jabbadap]

Spoiler: Uhoh Radeon off topic, apologies

Undervolted Radeon 7 fairly regularly gets to around 220-230W of GPU-only draw without performance hit. Sometimes less, sometimes more. This is in the ballpark. From undervolting results it can be argued that AMD can get the same efficiency if they only reigned in their default voltage settings. The primary reason they do not is that they have a performance level they need to hit which today is RTX2080 and at this point in the efficiency curve the variability is large enough that they feel the need to overvolt a little bit by default.

Well yeah undervolting is form of OC, some chips can excel from it some don't(infamous silicon lottery). Then again Radeon VII is not the even full vega20 chip, with full fat Vega20 amd should not needed to clock card as high as it's now and binned voltage could have been a bit lower. Same can be said about RX Vega⁶⁴ it can be very efficient with lower voltages, but because of competition it's clocked way over it's optimum point on efficiency curve.

 

[rruff]

Undervolted Radeon 7 fairly regularly gets to around 220-230W of GPU-only draw without performance hit. Sometimes less, sometimes more. This is in the ballpark. From undervolting results it can be argued that AMD can get the same efficiency if they only reigned in their default voltage settings.

Isn't Nvidia in the same boat with this? They can be undervolted as well. Considering that they can usually be overclocked more at stock voltages, I suspect they could be undervolted more at stock clocks. I've never investigated it though. And people have less incentive to try since they are so efficient to start with.